Leviathan

Learning to Live with Crypto

Fri, 03 Oct 2025 00:00:00 GMT

Learning to live with Crypto, co-written with Andreas Dombret—our second collaboration—and published by the LSE’s Systemic Risk Centre (SRC), lands at a timely moment. In a Financial Times piece this week, Bank of England Governor Andrew Bailey asked, “how do we ensure the link between money (in whatever form) and credit creation as an underpinning for economic activity?” We take up that question—and more.

From “let it burn” to strategic embrace

Post-FTX, supervisors wanted banks far from crypto: if it failed, let it fail alone. That stance has flipped. Washington now treats stablecoins as internet-native dollars, deepening dollar reach and demand for Treasuries. Public aims and private issuance have aligned, recasting stablecoins from nuisance to instrument of statecraft. Like it or not, the world must live with a programmable, offshore extension of the dollar.

Coexistence, not replacement

The realistic end-state isn’t extinction or triumph; it’s coexistence. Permissionless rails (Ethereum, Solana, others) will continue to serve speculation and cross-border flows. Regulated finance will adopt tokenised deposits, tokenised assets, and (in some places) CBDCs to capture automation and instant settlement inside the perimeter. Stablecoin volumes already run in the trillions, with usage spiking when local currencies slip—functionally a retail-heavy, digital eurodollar layer.

Tokenisation is misunderstood

Nomenclature misleads. “Wallets” hold keys, not coins. “Tokens” are ledger identifiers with logic stored on-chain. The real design question isn’t cosmetics but co-location: cash, assets and business logic must share the same programmable substrate to unlock atomic settlement and composability. That is why the BIS’s “Unified Ledger” is a clearer—and better—framing than the catch-all “tokenisation.”

Three risks regulators underweight

1) Concentration.
The cash of crypto is a duopoly: USDT and USDC dominate. Despite both being “dollars,” they are not fungible—each is tied to its own contracts, pools and redemption rules. Network effects risk importing Big-Tech-style platform power into what should be a monetary commons. Bridge hacks and synthetic wrappers add fragility; a wobble in a base pair reverberates system-wide.

2) Dollarisation.
Stablecoins are overwhelmingly dollar-denominated. In emerging markets, they enable stealth dollarisation at smartphone speed, bypassing domestic frameworks and amplifying pro-cyclical shocks. The EU’s MiCA offers a pragmatic template for sophisticated markets: permit dollars, but constrain retail-payment-like use and volumes.

3) The erosion of credit money.
Here is Governor Bailey’s question in practice. Fully reserved stablecoins pull deposits from banks into public-collateral wrappers. That shrinks cheap, stable funding for lending, raising spreads and compressing the money multiplier. What looks like neutral plumbing can hollow out credit creation, the engine of modern economies.

The supervisor’s double bind (outside the U.S.)

Non-U.S. authorities face imported dollarisation, platform power from private money, and deposit flight from domestic banks—without lender-of-last-resort cover for offshore digital dollars. They must defend monetary sovereignty while preserving credit elasticity.

The opportunity: build the better rail

The same technology threatening sovereignty can reinforce it if harnessed well. Tokenised deposits deliver atomic settlement, automated escrow, event-driven payments, and real-time collateral mobility—where the gains come not from faster databases but from disintermediation and programmability.

Enter the BIS Unified Ledger: tokenised central bank money (i.e. wholesale CBDC), tokenised deposits, and tokenised securities sharing one programmable infrastructure. Public authorities supply the safe settlement asset and governance; banks supply intermediation and innovation. This preserves the singleness of money (par convertibility) while delivering crypto-like efficiency without ceding control.

Live experiments—Agorá, Helvetia/Jura, Dunbar, Mariana—point to feasible architectures: permissioned nodes run by supervised institutions; interoperability anchored in central bank money; governance via standards for APIs, permissioning and finality. Internationally, interlinking ledgers will be political work as much as technical work—but tractable within supervisory areas.

Restoring elasticity

Unlike rigid, fully backed stablecoins, tokenised deposits expand and contract with credit demand and central bank backstops. They integrate with lender-of-last-resort and deposit insurance, keeping liquidity elastic rather than frozen at issuance. That is the practical path to answering Bailey’s question: preserve the money-credit link while modernising the rails.

Conclusion: live with, not by, crypto

The strategic die is cast: stablecoins function as an offshore dollar layer; tokenisation is reshaping market design. The task for supervisors is to metabolise the risks—concentration, dollarisation, and the assault on credit money—while seizing the architectural advantage of Unified-Ledger-anchored tokenised deposits. Living with crypto is inevitable. Living by it—ceding money and credit to offshore issuers—is not.

Money, Not Crypto: The BoE’s Systemic Stablecoin Blueprint

Wed, 01 Oct 2025 00:00:00 GMT

The Governor of the Bank Of England, Andrew Bailey in the FT sets out a systemic stablecoin regime that is open to innovation but anchored in money-like standards. The Bank of England’s three tests are clear: (1) risk-free backing assets; (2) deposit-style protection—insurance and a statutory resolution placing holders as preferred creditors; and (3) direct, uniform 1:1 convertibility into fiat, not reliant on crypto exchanges.

The policy goal is continuity of the “par” promise across cash, deposits, and tokens. But the transition carries risks. If stablecoins scale, deposits could shift to narrowly backed tokens. That may weaken bank funding, push credit intermediation to non-banks, and complicate monetary policy transmission. It also raises operational, governance, and cyber risks that insurance and resolution need to address in detail.

The notable step is the proposal that widely used UK stablecoins should access BoE accounts. That would provide central-bank settlement and strengthen parity, but it also creates design choices on access, supervision, caps, and contingency tools to avoid unintended disintermediation.

A BoE consultation will follow on standards for payments and tokenised market settlement. Europe should engage early: align definitions of “systemic” and “money-like,” set common requirements on backing, convertibility, and resolution, and design cross-border interoperability to reduce regulatory arbitrage.

My view: permit privately issued tokens to plug into public settlement only under conservative safeguards, phased implementation, and ongoing impact assessment on bank funding and policy transmission.

Crypto’s Institutional Turn — Incentives, Business Models, and Dilemmas

Mon, 08 Sep 2025 00:00:00 GMT

Crypto is no longer a sideshow. From corporate treasuries borrowing to buy tokens, to Wall Street banks edging into bitcoin-backed lending, to Washington opening retirement savings to crypto and private equity—the industry is becoming embedded in the financial mainstream. Each move looks rational for the actors involved: higher equity valuations, new fee streams, more inflows. But when you connect the dots, a bigger picture emerges: fragility is being built into the system itself.

Our latest RAAK note explores how these dynamics play out across corporate balance sheets, credit markets, banks, and retirement funds. It shows how the incentives driving short-term gains can create long-term systemic risk, especially for ordinary savers who may end up absorbing the shocks.

If you want to understand not just the hype but the underlying shifts shaping the next phase of digital finance, this note is important reading.

👉 Download the full paper here.

The Myth of a Neutral Fed is Over

Sun, 31 Aug 2025 00:00:00 GMT

“What you call the fiction of independence really has acquired a huge global meaning. And…in the increasing polarization of American politics, it’s going to be harder and harder to maintain that.”

Economic Historian Adam Tooze’s words in a recent episode of Ones and Tooze cut straight to the heart of the Federal Reserve’s current predicament. Independence has always been more fragile than central bankers liked to admit. Nixon harangued his Fed chair; Johnson leaned hard; even Clinton and Obama, though more restrained, did not treat the Fed as untouchable. But Trump’s approach is of another order: public, relentless, and fused with a broader Republican project to “stack” the Fed both nationally and regionally.

The aim is not just lower interest rates—though Trump has made no secret of his desire to juice the economy—but a much more ambitious extension of presidential power. As Tooze and Cameron Abadi noted, the GOP is now openly attempting to assert partisan control over the apparatus of money creation and financial regulation. That, in Tooze’s words, makes this “an existential challenge” to the very idea of an independent central bank.

Trumps allies fear the Fed is run by the liberal establishment

This confrontation is sharpened by the electoral cycle. Trump’s allies fear that if a recession hits in the coming year, the Fed—staffed by figures they regard as part of a liberal establishment—will stand aside rather than act decisively to rescue the economy. The suspicion is that Fed officials would see it as their institutional duty to minimize Trump’s re-election chances by letting downturn risks play out. It is an extraordinary charge, but it reflects a deep distrust: the belief that the Fed is not merely technocratic but stacked with elites whose political sympathies lie firmly on the other side.

That sense of distrust feeds the urgency behind efforts to bring the Fed under partisan control. It is not entirely unfounded. During Trump’s first term, senior Fed figures made remarks suggesting that if the president himself was a source of instability, then their overriding duty as guardians of stability was to act in ways that might limit his prospects for re‑election. One striking example came from William Dudley, the former president of the New York Fed, who in an August 2019 Bloomberg op‑ed argued that Trump’s re‑election itself could pose a threat to the U.S. and global economy and that Fed officials should consider how their decisions might shape the political outcome.

Although quickly disavowed by the Fed’s leadership, Dudley’s words offered rare confirmation of Tooze’s claim. Such comments, rare though they were, reinforced the impression that the central bank was not impartial but inclined toward defending an establishment consensus against Trump’s agenda. For Trump’s circle, this is not simply about rates or growth but about ensuring the central bank does not act as a hidden veto player against their project of political dominance. It is the mirror image of the longstanding fear on the left—that central banks have always served as the “wall of gold” blocking radical governments from exercising real power.

What makes the silence of central bankers at their recent Jackson Hole gathering so striking is that everyone could see what was coming, yet no one dared say it aloud. The Fed has always been a political institution, shaped by presidents, Congress, and public moods. But the fiction of neutrality—the idea that monetary policy could somehow be above politics—gave it credibility at home and authority abroad. If that fiction collapses, what replaces it?

Stability of Bond Market owes to the efficiency with which Scott Bessent and his team have managed Treasury liquidity

Adding to the drama is the surprising calm of financial markets. For decades, the specter of bond market panic was invoked as the ultimate deterrent to political interference with the Fed. Yet, as Tooze notes, “look what’s happened. Zip, nada, crickets, nothing.” The bond markets have not flinched. Much of this stability owes to the efficiency with which Scott Bessent and his team have managed Treasury liquidity. For all his bluster, Trump has surrounded himself with operators who understand market plumbing well enough to keep the system steady even while shaking its institutional foundations.

The uncomfortable reality, as Tooze reminds us, is that the Fed resembles the Supreme Court more than it resembles a neutral technocracy. It may be institutionally distinct from Congress or the White House, but it is not—and never has been—apolitical. The struggle over its future is really a struggle over whether America’s central bank will remain an autonomous guardian of monetary stability, or whether it will be bent, quite openly, to the will of partisan politics.

Everything is political, even the Fed.

Crypto Isn’t Replacing Fiat — But It’s Changing Everything

Thu, 28 Aug 2025 00:00:00 GMT

In recent years, crypto advocates have made bold claims: Bitcoin will replace the dollar. Ethereum will power a new world economy. Central banks will become obsolete.

But that’s not what’s happening.

According to Mark Copelovitch and Thomas Pepinsky in their new paper, The Political Economy of Shitcoins, the most plausible future isn’t one of total disruption or total collapse. It’s something more complicated — and more dangerous to ignore.

It’s a future where cryptocurrency survives, but doesn’t replace, sovereign fiat currencies. Where Bitcoin, stablecoins, and crypto platforms exist alongside the dollar, the euro, and the yuan. This is what they call World 2, and it’s already taking shape.

A Dual Monetary System

In this world, fiat money still does the heavy lifting: taxation, public spending, central bank policy, and international trade. Governments retain monetary sovereignty.

At the same time, crypto persists — not at the core, but around the edges. It thrives as:

A speculative asset class
A workaround for capital controls
A remittance tool in underbanked economies
A way to transact outside of formal institutions
A banner for ideological movements that want to break from the state
A cheaper way to make international payments (my addition)

This coexistence isn’t a stable arrangement. It’s a delicate balancing act, and it raises hard questions for regulators, central banks, and finance ministries.

Innovation or Extraction?

Some see promise in this model. Crypto, they argue, forces financial innovation. It expands access. It pushes governments to modernize payments systems and build CBDCs. The friction between crypto and fiat leads to productive reforms.

But there’s a darker side.

Crypto’s survival often depends on avoiding regulation while benefiting from state legitimacy. It introduces new risks without offering new protections. It invites speculative bubbles, facilitates illicit flows, and complicates monetary policy — all while framing itself as an alternative to the very systems it leans on.

And as Rohan Grey, a law professor at Willamette, puts it: the stablecoin trilemma makes it impossible to create coins that are simultaneously stable, denominated in a public unit of account, and free from reliance on public collateral.

Copelovitch and Pepinsky put it plainly: crypto in World 2 is either a symbiont or a parasite. It’s either helping to modernize finance or draining state credibility. Likely, it’s doing a bit of both.

The Policy Challenge

What does World 2 mean for policymakers?

It means crypto is not going away — but neither is the state. The real challenge is not deciding whether to ban or embrace crypto. It’s managing the tension that comes from living in this hybrid system.

That means:

Crafting smart regulation for stablecoins and decentralized finance
Developing CBDCs not just as innovation, but as a defense of monetary sovereignty and innovation
Collaborating across borders to avoid regulatory arbitrage
Preparing for new kinds of systemic risk, even from assets that don’t look like money
Deciding when — and if — the state should backstop private crypto failures

Because in World 2, the line between public and private, fiat and crypto, is constantly shifting.

A Final Thought

Crypto doesn’t need to win to matter. It just needs to persist — and it has.

This future, where crypto and fiat coexist uneasily, is not a footnote in monetary history. It’s the battlefield. And what’s at stake isn’t just which currency we use — it’s who gets to decide.

World 2 is here. The question now is whether policymakers are ready to govern it — before it governs them.

Inspired by The Political Economy of Shitcoins by Mark Copelovitch & Thomas Pepinsky (2025)

The “Brussels defect”, is a bug in a world that needs credible strategy instead

Wed, 27 Aug 2025 00:00:00 GMT

Call it the “Brussels defect”: the habit of treating market rules as a substitute for strategy. As Abe Newman recently coined it in the Financial Times, Europe’s regulatory reflex no longer matches a security-first world.

Over the past decade, fines became a cost of doing business, US platforms scaled regardless, and Europe’s technology base fell further behind. Rules without industrial capacity are muscle without sinew. The policy pivot is obvious but hard: shift from rule-setting to capacity-building. Digital sovereignty is, at bottom, industrial policy—fund it, scale it, export it.

EuroStack plans

Newman touts the emerging EuroStack plans. These proposals are a start, but skeletal. They mostly neglect two engines of US digital power: network effects—where scale compounds advantage—and deep capital markets that finance ecosystems, not just firms.

The current EuroStack plans talks about federation and open source a lot. Yes. It can play a role in parts of the stack (particularly the back end), but they are not a growth model on their own. Part of the problem then is a suspicion of large centralised European players - European “Big Tech” if you will.

But Europe will likely need a mix that also includes centralized, proprietary champions in targeted layers, precisely so capital markets can do their work of aggregation, IPOs, and follow-on financing.

Both innovation and capital markets tend to cluster in specific geographies. But European member states jealously guard what they have and fight any such clustering elsewhere.
Fragmentation is the perennial European problem. The most crucial first step is then is political –– choosing to be European first.

Industry groups: new crypto law has a ‘loophole’ that could cost banks

Tue, 26 Aug 2025 00:00:00 GMT

Banks are feigning surprise at yield on stablecoins. US banks are warning that the new GENIUS Act could trigger trillions in deposit outflows, since it bars stablecoin issuers from paying yield but allows crypto exchanges to keep offering rewards on third-party coins. They argue this risks undermining credit creation and monetary policy, while crypto groups counter that banks are trying to block competition and protect their own margins.

The FT reports that banking lobbies are pressing lawmakers hard to close what they call a regulatory “loophole,” warning of destabilising deposit flight, while crypto groups frame the campaign as a protectionist push to tilt the rules in Wall Street’s favour.

Yet Coinbase has paid USDC “rewards”since 2019; the GENIUS Act merely bans issuers from paying yield while leaving room for platforms and exchanges to share economics on third-party coins. That’s continuity in the GENIUS Act, not a loophole.

How big could outflows be? Treasury analysis cited by banks pegs the upper bound at $6.6tn depending on whether yield is available—think MMFs in the 1980s, but on faster rails. Treat it as a stress-case, not a base-case, but it explains the sudden lobbying blitz.

Why a Digital Euro really matters (beyond “stablecoin summer”)

Mon, 25 Aug 2025 00:00:00 GMT

Goldman calls this “stablecoin summer,” and the FT reports that the EU is accelerating digital euro plans after new U.S. stablecoin laws stirred competitiveness fears. Officials argue a digital euro would provide central-bank-backed payments as cash use declines and bolster the euro’s global role; they also worry U.S. rules will turbocharge dollar stablecoins in Europe.

The ECB, however, needs to sharpen its communications on the purpose of a digital euro. The underlying logic the FT reports is off: people choose dollar stablecoins because they are dollars. Even if euro demand were strong, the proposed digital euro is account-based—unlike bearer-style stablecoins—and will require onboarding.

There are also good reasons not to court large external holdings: a stronger euro could weigh on exports; Europe has AML/CFT responsibilities; and it should avoid eroding other countries’ monetary autonomy. Meanwhile, Europe already regulates venues offering stablecoins in Europe—dollar, euro, or otherwise. It can make it hard for Europeans to buy unregulated stablecoins if it chose to.

Of the standard arguments, the cash-access point is compelling; an offline-capable digital euro deserves priority.

Why a digital euro makes sense

But equally the “CBDCs solve no problem” position is also short-sighted.

The case for a digital euro rests on:
Strategic autonomy. Reduce dependence on non-European card networks and finally guarantee that every cardholder can pay in any euro-area country (still not universally true).
A programmable backbone. Use unified/interlocking ledgers so tokenised bank deposits settle in tokenised central-bank money—preserving the two-tier system and a flexible, credit-responsive money supply.

As Andreas Dombret and I have argued, being anti-stablecoins is not anti-innovation position. Unified ledgers can strip out intermediary layers, cut cross-border costs, let smart contracts replace escrow and some collateral workflows, and unlock liquidity in “frozen” assets with real-time performance data.

The real debate isn’t tech vs. status quo; it’s how to apply these capabilities to strengthen a powerful yet delicate credit-money system. We can have the cake and eat it: innovation—without tying itself to fully-reserved private money.

Born Free, Coded in Chains

Sat, 23 Aug 2025 00:00:00 GMT

“Man is born free, and everywhere he is in chains.” Rousseau was speaking of society, but his words embrace digital debates too. If freedom really means the ability to stand apart and flee into nature, as Rousseau believed; if to step aside from the bonds of others is the answer, then what do today’s much-hyped technologies that promise freedom in the name of decentralisation — blockchains and federated networks — really offer us?

This newsletter draws its name from Thomas Hobbes’s work, “Leviathan.” Hobbes viewed nature as perilous, proposing the state, or Leviathan, in which power would be centralized, as the solution. In stark contrast, Rousseau believed that humans were inherently free in nature, and it was society, and it’s rules, enforced by the state, that imposed restrictions on us.

Advocates of blockchain insist it represents a new frontier of freedom. Erik Voorhees declares: “Bitcoin will win because there is now competition in money, and bitcoin is the best money available because it is decentralized and cannot be stopped.” But decentralisation here is a slippery notion. Blockchains do not merely link servers; they fuse them into one network, bound by common code, ledger and consensus. To join is to become part of an indivisible whole. For Rousseau the network is the problem, its edges are chains — and blockchain is all edges.

Federated networks like Mastodon tell a different story. They also interconnect, but loosely. Each server is autonomous, capable of severing ties, rewriting its rules, or withdrawing entirely. What you loose is distribution, and an ability to communicate across the federation. But you can still publish on the internet. You can leave the network and still survive. Here Rousseau’s solitude, his ability to slip away, remains imaginable. Federation offers a purer form of decentralisation than blockchain.

Blockchains, by contrast, tend toward centralisation through its architecture — one ledger. Here the inherent network relationship between scale and usefulness manifests most powerfully. For instance, Stablecoins, the first killer app of the blockchain thus far, gravitate to the largest chains; liquidity pools and token standards reward concentration. Network effects produce the very monopolies decentralisation promised to resist. The architecture may be distributed, but power coagulates. Instead of freedom, we inherit new masters: vast technical fiefdoms disguised as peer-to-peer liberty.

This is not to say we can retreat into Rousseau’s woods with federated services. Even when disconnecting from federated networks we cannot entirely escape connection. Servers still rely on shared protocols, physical infrastructure, and the messy politics of standards. Rogue crawlers will find and index you; vulnerabilities demand constant updates of your software. Online the dream of total solitude, is a mirage, perhaps even more so than Rousseau’s fantasy of freedom offline. Society presses in everywhere.

Perhaps Rousseau is the wrong guide. Donna Haraway’s cyborg offers a better compass. “It is not an option,” she wrote in her Cyborg Manifesto, “but a constitution: we had better inhabit this world as movers and shakers, or we will be just victims.” By which she meant, technology is not external to us; it is woven into what we are. We cannot flee it, but we can shape it — through politics, design, and collective struggle.

If blockchain is another chain disguised as freedom, and one that is amassing immense power, then Hobbes’s warning may be worth heeding. To secure peace and a measure of freedom, we may need to entrust power to a sovereign — a Leviathan — tasked with ensuring that these networks serve the public good rather than threaten it.

This logic extends well beyond blockchain. Consider Big Tech platforms that dominate search, social media, and app stores: here too, technical “opt-outs” are rarely enough. Activists seeking relief from pervasive platform influence should rediscover that law, regulation, and collective governance are as essential as code. In practice, freedom online may rely less on escaping chains than on reshaping who forges them, and under what rules.

Wall Street's dramatic reversal on cryptocurrency

Sat, 16 Aug 2025 00:00:00 GMT

The NYTimes has reported on Wall Street’s sudden about turn on cryptocurrency. Major bank executives who once competed to be crypto’s loudest critics are now racing to embrace digital assets, marking a stunning reversal in the financial industry’s position.

Behind Wall Street’s sudden embrace of cryptocurrency lies a troubling reality: major banks are rushing into digital assets despite serious risks to both their stability and consumer protections, according to nine financial executives briefed on their organizations’ crypto initiatives.

The deposit flight threat

The systemic risks extend far beyond individual accounts. Every dollar converted to stablecoins shrinks banks’ traditional deposit base, fundamentally undermining their ability to make loans. Since stablecoin funds must be held in government securities rather than lent out, widespread adoption could cripple banks’ core lending function.

“You don’t need a lot of deposit flight to really buckle the banks,” warns Mike Cagney, former SoFi CEO and current head of digital lender Figure. The Federal Reserve Bank of Kansas City has already flagged potential “unintended consequences for the economy” if this shift accelerates.

One prominent banking attorney, speaking on condition of anonymity, revealed the industry’s true sentiment: “If the banking industry was being totally frank, they would say they wish stablecoins had never been invented.”

Learning nothing from history

The NYTimes report that JPMorgan’s internal research team recently uncovered disturbing parallels to the “wildcat banking” era of the 1800s, when hundreds of state-chartered banks issued their own currencies. That period was characterized by rampant fraud and cascading bank failures, only stabilized by creating a unified national currency.

Now, banks are preparing to fragment the monetary system again. Plans include both individual bank stablecoins and a potential industry-wide token, complete with rewards programs designed to lure customers away from protected traditional accounts. The rush is driven by fear of being left behind as companies like Walmart and Amazon explore their own digital currencies.

Regulatory vacuum and political conflicts

The hastily passed GENIUS Act created a framework for stablecoin operations but left massive regulatory gaps. The new interbank payment system being developed would operate with “few consumer protections and nascent regulatory oversight,” according to executives familiar with the plans.

Unlike traditional bank accounts, stablecoins will not be covered by federal deposit insurance. If they fail, customers have no government backstop to recover their funds.

The GENIUS Act requires banks to park stablecoin money in “risk-free” assets, allowing them to pocket the returns while offering depositors nothing—a guaranteed profit scheme with no benefit sharing.

Critics note that President Trump’s sons run World Liberty Financial, a crypto startup issuing its own stablecoin, raising questions about conflicts of interest in the administration’s crypto push. Bank lobbyists who helped shape the legislation admit they felt “forced by circumstance” rather than acting from conviction about the technology’s benefits.

Consumer confusion and hidden costs

Bank of America’s Brian Moynihan acknowledged the coming system would be a “complex array,” hopefully “not complex to the consumer”—an admission that banks themselves don’t fully understand what they’re building. Customers may not realize they’re trading FDIC-insured deposits for unprotected digital tokens, or that their stablecoin holdings generate profits for banks while earning nothing for them.

The promise of cheaper international transfers masks deeper costs: the erosion of lending capacity that funds mortgages and business loans, the loss of interest on deposits, and the elimination of federal insurance protection. These hidden trade-offs aren’t being clearly communicated as banks race to launch their stablecoin offerings.

The unspoken fear

Perhaps most tellingly, the nine Wall Street executives who shared their concerns did so only on condition of anonymity, not authorized to speak publicly about their employers’ crypto initiatives. This culture of public enthusiasm masking private dread suggests the industry knows it’s taking enormous risks but feels powerless to stop.

While executives like Fifth Third’s Tim Spence claim “the consumer checking account is probably safe,” the hedging language reveals uncertainty even among banking’s top leadership. The rush into crypto isn’t driven by careful risk assessment or consumer benefit, but by fear of missing out and political pressure.

As banks prepare to launch these untested systems before year’s end, they’re essentially conducting a massive experiment with the foundation of the American financial system—one where consumers bear the risks while banks secure the rewards.

How stablecoins feed global USD demand

Tue, 12 Aug 2025 00:00:00 GMT

Scott Bessent must be happy. A new IMF study finds that stablecoins boost dollar demand. As a % of GDP, the largest users are Latin America & Caribbean (7.7%) and Africa & Middle East (6.7%)

The paper reveals $2 trillion in stablecoin flows in 2024, with North America as the primary source of net outflows ($54bn) - essentially exporting digital dollars to meet global demand. Flows increase when local currencies weaken against USD.

Particularly striking: China’s stablecoin activity is 5.5x higher than previously estimated ($153bn vs $28bn), likely due to VPN usage. The March 2023 banking crisis significantly disrupted flows from North America, showing how intertwined stablecoins are with traditional banking.

Decrypting Crypto: How to Estimate International Stablecoin Flows

Why stablecoins are Silicon Valley's Pandora's box

Mon, 11 Aug 2025 00:00:00 GMT

President Trump’s signing of the GENIUS Stablecoin Act marks a chapter in an ancient storyline—the eternal tension between progress and order. Pandora’s jar—mistranslated as a box—came with strict instructions never to open it. Stablecoins also arrive with promises and prohibitions.

The warnings are mounting about the Pandora’s box of problems that stablecoins will unleash. The economist Barry Eichengreen, one of the leading experts on the history and development of money, invokes the spectre of coins slipping their pegs to the dollar—as happened in 2022—and recalls the chaos of the free-banking era, when dozens of regulatory systems produced banknotes trading at wildly different prices. Yet this is perhaps the least urgent danger lurking inside.

It is likely that only a few stablecoins will dominate. Despite piggybacking on the dollar and often sharing the same blockchain, out of the box their smart contracts remain stubbornly non-interoperable and therefore non-fungible.

Each coin is a brand—and combined with the network effects inherent in money itself, winner-takes-all dynamics emerge. Already, two giants rule: Tether commands a $159 billion market capitalisation - 61% of the market - while Circle’s USD Coin holds $62 billion or 24%.

Say what you will about the banking system, but it resists the extreme concentration that defines Big Tech. In spite of years of consolidation, the US still has more than 4,000 banks. Europe has more. That is because, up to now, banks have made a business of a private means of payment but to an open public standard. The result is a monetary commons—a seamless blend of commercial and central bank money that, in normal times, the public can’t distinguish between.

Read the rest of this essay on the profound and mutiple problems with Stablecoins I wrote with Andreas Dombret, formerly on the Deutsche Bundesbank board and the supervisory board of the ECB.

Why the marketplace of ideas failed online

Sat, 09 Aug 2025 00:00:00 GMT

“WHEN we use a network, the most important asset we get is access to one another,” wrote Clay Shirky in his book, Here comes everybody**,** a book about the internet and social media in 2008. Back then, most people believed the internet would do more than just innovate — it would free ordinary people to connect, express themselves, and even spread democracy around the world.

The thinking was simple: bad speech would be drowned out by good speech, and from all this conversation, truth would emerge. This faith in the power of free-flowing ideas runs deep in Western thinking. Philosophers from Milton to Locke to Voltaire all championed it.

John Stuart Mill argued in On Liberty that freedom of speech and debate are essential because they allow truth to emerge from the clash of opinions. Something many later compared to a marketplace. For Mill, open discussion prevents dogma, corrects error, and sharpens understanding — making progress and healthy democracy possible.

Fast forward to today: democracy is retreating worldwide and conflict is rising. Within democracies, division has become so bitter it’s eating away at the fabric of society. We need to face an uncomfortable possibility: What if letting everybody speak to everybody doesn’t bring enlightenment? What if it actually spreads division and chaos so effectively that it endangers democracy itself?

Within democracies, division has become so bitter it’s eating away at the fabric of society. We need to face an uncomfortable possibility: What if letting everybody speak to everybody doesn’t bring enlightenment? What if it actually spreads division and chaos so effectively that it endangers democracy itself?

The Old Media World and Its Critics

Before the internet, the media world of democracies had plenty of critics. To be a publisher, you needed serious money. You had to buy massive printing presses, build distribution networks, or own a TV station. You needed specialized staff and advertising departments — all requiring substantial capital.

Noam Chomsky captured this critique well in Manufacturing Consent. He argued that these big media corporations had built-in biases and carefully controlled what topics were up for discussion. Since they depended on advertising revenue, they tended to avoid rocking the boat. (Elon Musk recently discovered firsthand how advertisers can constrain what you say.)

The result? Media that pushed society toward safe, middle-of-the-road consensus. The left especially disliked this system, seeing how commercial interests prevented serious discussions about creating a more equal society.

Then came the internet and social media. The left hoped these new tools would bypass corporate media’s manufactured consensus. As the saying goes: be careful what you wish for.

The left hoped these new tools would bypass corporate media’s manufactured consensus. As the saying goes: be careful what you wish for.

The Filter Bubble: Only Half Right

Eli Pariser was among the first to sense something was off with online media. In The Filter Bubble (2011), he warned that online, we tend to connect with people who already agree with us. Algorithms make this worse by showing us content we’ll like, shielding us from challenging viewpoints.

This theory felt right because it fit our existing beliefs about the marketplace of ideas: if only we were exposed to diverse opinions, we’d debate our way to enlightenment.

But recent research shows Pariser’s Filter Bubble Theory was only half right. Yes, we prefer our own tribe and befriend like-minded people. But here’s the twist: even on platforms without engagement-based algorithms (like Mastodon’s simple chronological feed*), we actually pay more attention to posts from people we disagree with — especially when they say something we find morally outrageous. We notice it more, read it more, share it more, and comment on it more than other content.

But recent research shows Pariser’s Filter Bubble Theory was only half right. Yes, we prefer our own tribe and befriend like-minded people. But here’s the twist: even on platforms without engagement-based algorithms (like Mastodon’s simple chronological feed), we actually pay more attention to posts from people we disagree with — especially when they say something we find morally outrageous.

Why We’re Wired for Outrage

Humans evolved as a highly social species. We developed psychological traits that helped us maintain tight-knit groups — a crucial survival advantage in a dangerous world. Morality itself evolved to help us cooperate and manage these small social groups. We’re literally hardwired to notice and react to people breaking the rules of cooperation above almost everything else. In small groups, this instinct helps maintain order.

So the problem isn’t that we don’t see opposing viewpoints — it’s that we see too many of them, and only the most extreme examples.

Research shows that compared to offline life, we’re far more likely to hear about someone doing something “immoral” online than through print, radio, and TV combined. This constant exposure to others’ wrongdoing breeds contempt. And algorithms designed to maximize engagement only amplify our natural instincts.

So the problem isn’t that we don’t see opposing viewpoints — it’s that we see too many of them, and only the most extreme examples.

The Extremism Illusion

Making matters worse, we assume these extreme examples represent what the other side actually believes and does. We forget that people are motivated by status-seeking, and a small minority gain status by saying outrageous things that trigger our instincts — things that algorithms then boost to millions.

Online, your follower count is the clearest status signal. It also can determine your reach — how many people see your messages. The fastest way to build followers? Post extreme, emotional content. But in reality, only a small number of people generate most of this attention-grabbing content.

The Problem with Online Punishment

Another evolved trait compounds the problem: what researchers call “third-party punishment.” This is when we take action to discipline someone who broke the rules, even at a cost to ourselves. You can see it in children — like when a child gives up their spot on the slide to stop another child who cut in line.

We’re especially likely to punish rule-breakers when others are watching, probably because it shows we’re reliable group members. Online, we always have an audience, and punishing wrongdoers maximizes our chances of attention and status within our group.

But here’s the catch: before the internet, this punishment was costly. It only made evolutionary sense when you had to live day-to-day with the people you were disciplining — not when they live across town, across the country, or across the planet.

Technology Is Indifferent

Technology changes how we interact, but it doesn’t care about the outcome. The printing press, with its one-to-many communication, is linked with the Enlightenment and the birth of nation-states. We got lucky. Now were out of it: digital technology’s many-to-many megaphone might be putting that entire project in reverse.

The promise of everyone speaking to everyone hasn’t delivered enlightenment. If there ever was a marketplace of ideas online, it is an example of market failure.

Instead, the internet may have given us something our evolved psychology was never equipped to handle: constant exposure to the worst of human behavior, amplified by algorithms and stripped of the natural constraints that once kept our punitive instincts in check. It is pulling democratic societies apart.

* People love to criticise engagement boosting algorithms, and though they make these issues worse, platforms with no algorithms like Mastodon face the same baseline issue.

This piece originally appeared in Vrye Weekblad.

The Great Resentment

Fri, 08 Aug 2025 00:00:00 GMT

“…something I keep thinking about and which I can’t make up my mind on is: were swathes of the right, centre-right and, in some cases, the centre, full of racists hiding their beliefs for decades, or are these people all incredibly suggestible? neither answer is good” wrote Marie Le Conte on Bluesky this week.

It’s a good question.

I’ve been puzzled and intrigued by why the white working class, in Western countries continues to swing to the right, and in an increasingly racial exclusionary way.

While this phenomenon is likely multi-causal, one obvious explanation remains the economic deprivation in many areas where white voters initially supported far-right parties—regions that have become even more economically disadvantaged over the past 30 years as Western countries have de-industrialized due to globalisation.

In the US, data shows this effect has been significant in particular areas, though overall the so-called “China shock” hasn’t been as impactful as technology in affecting job prospects.

The situation varies across countries. Travel through East Germany today, for instance, and it’s visibly much better off than poorer parts of the UK. So Germany doesn’t have such a stark gap between rich and poor, which suggests we’re really talking about relative differences—and perhaps perceived differences in status: East Germany is very aware it is not as wealthy as the West, and that it is where it is, in part, because of subsidies from West Germany.

Nevertheless, both the perceived and real relative decline in the economic status of the white working class in the West has played a significant role.

What else might be at play? Branko Milanovic offers a compelling framework, arguing that it’s Marxism 101: when the economic base changes, so does ideology. As he puts it: “One has to realize that the world population was 5.3 billion in 1990; it is now, 8.2 billion. China produced 7.2% of the global GDP; now it produces 22%. India produced 3.5% of the global GDP; now, it produces 9%. As the underlying reality changed, so did ideology.”

Marx might have said, “the material base of society has shifted from West to East. As the means of production expand and relocate—in this case, to China and India—so too does the ideological superstructure adapt to justify, obscure, or resist these changes.” While Milanovic doesn’t specify exactly what these ideological effects might be, I can guess at some possibilities.

One is that many successful or perceived-to-be-successful Asian countries—Korea, Japan, and especially China, with India increasingly following suit—have long advanced not a multicultural vision, but rather a single cultural and even ethnic idea of what society should be. This stands in contrast to the liberal multicultural position that places like the USA or UK have claimed since WWII.

A striking illustration of this is the infamous Kising interview with the ex editor of the Spectator. Kisin claimed Rishi Sunak isn’t English, pointing at Japan. It’s hard to imagine that many working class folks walk around with Asian conceptions of ethnic identity top of mind, but it is possible to imagine that in the future, if this region continues to dominate, and if multi-cultural societies languish economically, it will. Silicon Valley’s fantasies about countries being run by CEOs instead of democratically are undoubtedly already influenced by places like Singapore.

Social psychology offers another lens through which to understand this shift. Research on status threat and relative deprivation shows that people react strongly not to absolute losses but to perceived declines in their group’s standing. In the past, if you were part of the white working class in Western countries, you might not have been as well off as your more affluent compatriots, but you were still part of nations perceived to be more successful and higher in the global pecking order.

You could afford holidays in other parts of the world. Your currency went further when you went on holiday. Foreigners looked up to you. That’s increasingly no longer true. The relative status of people has changed, and I think this shift in standing vis-à-vis the rest of the world—this perceived loss of global position—could be, if not fueling resentment, making people far less accommodating toward others.

Of course, other things are playing a role too. No doubt social media and its ability to bypass gatekeepers, something the left thought would be a good thing, is. I discussed that here. And the fact that we are hard wired to engage with certain kinds of content. And engagement is what drives social media distribution. I discussed that here.

And sociologists have long known that most of us don’t make up our ideas on our own. We are social animals and take our cues from others in our social network. When enough people in our circles change their minds, it can reach a tipping point that has much larger effects. And that’s—scary.

Perhaps Le Conte’s question has an answer that encompasses both possibilities: people weren’t necessarily hiding their beliefs, yet they are suggestible. Instead, the material conditions of their world have fundamentally shifted, and with them, the ideological frameworks and psychology through which they understand their place in it.

Both the Democrats and Silicon Valley shifted their political views

Wed, 11 Dec 2024 00:00:00 GMT

Why did Silicon Valley turn right?

Institutions shape people’s understanding of their interests and alliances during stable times. In crises, institutional frameworks come under strain and certainties crumble. This creates opportunities for political entrepreneurs armed with new ideas to reshape how people understand their interests — new alliances form over time.

This summarises Mark Blythe’s theory of the process by which shifts in institutional frameworks occur. Why do institutions matter for the question of Silicon Valley’s right turn? Shifts in our ideas and ideologies are embodied in institutions and alliances over time, so they become a framework for—and indeed constrain—action.

Here’s a practical example of why this matters: Although many argued against neoliberalism in the 1990s, neoliberalism as an idea was embedded in institutions and alliances, which muted the impact of these broadsides against the dominant dogma of the time. So the changes in alliances we are seeing mean this is a significant moment.

Political scientist Henry Farrell (he of an important concept du jour—weaponised interdependence) used Blythe’s theory to disagree with Noah Smith and Matty Igesias, who claim that it was the Democrats who moved away from the Valley.

Not so, says Farrell. That’s too simplistic. Both sides moved away from each other. In this post, I will summarise and comment on Farrell’s argument, which will help me think through it. TLDR: I think Farrel’s choice of Blythe’s theory and his application of it is typically excellent.

However, his description of the forces shaping Silicon Valley’s shift is light on detail. In a follow-up post, I will attempt to provide some ideas on the forces and events that may explain Silicon Valley’s shift.

The Californian Ideology is not libertarian but “liberal-tarian”

A survey conducted in 2017 is apparently the only solid social scientific evidence we have of Silicon Valley politics, says Farrel. Fascinatingly, it does not entirely confirm the idea many had, ever since Richard Barbrook and Andy Cameron’s influential Californian Ideology essay, that Silicon Valley ideas were simply a peculiar example of early neoliberalism.

The Californian Ideology eloquently argued that Silicon Valley’s politics was a novel mix of the New Left’s counter-cultural social values—personal freedom—and the right’s free market politics. A potent form of neoliberalism.

But the survey by Broockman, Greg Ferenstein and Neil Malhotra found that 75% of tech founders rejected libertarianism, instead showing a unique “liberal-tarian” ideology: “liberal” (in the American sense of the word) on most issues, including globalism, redistribution (public goods provision, education, the environment and significantly, taxation). Yet, they were also strongly anti-regulation, similar on that to Republican donors, and more so than non-tech entrepreneurs, especially regarding labour markets.

Proving that ideology was at work and not only self-interest, tech elites in 2017 showed broader anti-regulation views beyond their own industry. They opposed regulation and unions more strongly than typical millionaires and were particularly anti-union. Both Farrel and Broockman, Ferenstein and Malhotra argue that the Silicon Valley that held these beliefs influenced the Democrats.

Farrell references Margaret O’Mara, who hailed these anti-regulation Democrats the “Atari Democrats”. The effects of this point of view can be seen in the bipartisan consensus that meant that e-commerce wasn’t regulated or taxed (“see the influence of the Framework for Global Electronic Commerce”), the opposition to the copyright protection measures contained in SOPA, as well as Section 230 of the 1996 Communications Decency Act, which indemnified social media companies from being held liable for the content posted on their platforms.

Curiously, nobody mentions Net Neutrality provisions, which prohibit ISPs and telecom networks from preferencing and differentially pricing the data they carry. A form of regulation itself, Silicon Valley types tended to have no issue with it and framed it as necessary to protect speech and innovation.

When tech elites clashed with the Democrats back then, Farrell notes that it was because their illiberal policies threatened tech’s business model — and, in their minds, the world becoming more liberal. This is because they were fully bought into the quaint view that a more globally connected world, would mean a more liberal world, which would be a good thing. So, Google’s Eric Schmidt, during the Obama presidency, expressed fear in response to Snowden’s revelations that U.S. surveillance might destroy the Internet.

But the symbioses between the neoliberal Democrats and Palo Alto consensus mostly worked just fine:

…the neoliberal consensus among Washington DC Democrats and the Palo Alto Consensus were highly compatible with each other. Both favored free markets and minimal regulation of technology, while sidelining unions, labor politics and too close an examination of the market power of the big tech companies. There were frictions of course - even neoliberal Democrats were often keener on regulation for many people in Silicon Valley - but Democrats could see themselves in Silicon Valley and vice versa without too much squinting.

Much has been written about the US left’s move away from central tenets of neoliberalism — like free trade. The idea is that unbridled free trade led the US to the [China Shock](https://en.wikipedia.org/wiki/China_shock#:~:text=The%20China%20shock%20(or%20China,World%20Trade%20Organization%20in%202001.), which was bad for US manufacturing jobs. And a lack of US vision and industrial planning has led China, armed with an industrial strategy, to steal a march on some future key technologies, like solar and, more recently, electric vehicles. The Inflation Reduction Act is an attempt to address that. Much of Biden’s CHIPS Act’s specific design is aimed at slowing China’s progress in AI by denying the Chinese access to the most powerful semiconductors.

For Farrell, this shift by the Democrats is matched by a shift away from the Palo Alto consensus in Silicon Valley. More digital tech is not necessarily going to lead to more liberal societies. The “obvious” link between innovation and liberalism has been broken. All we hear now from the Valley is about the importance of innovation for economic growth, and sometimes, they add that it is also important for state power. When Mark Zuckerberg sought to defend Facebook’s ill-fated Libra before Congress, he framed its success in championing the US in geopolitical currency competition.

So, Farrell’s argument seems to have great explanatory power. But it’s an account that lacks enough explanatory detail about the shift Silicon Valley has undergone — why and why not all of Silicon Valley has shifted in the same way. And I will attempt to flesh that out in the next.

Footnote: Also see John Ganz’s commentary on Farrel from a Gramscian perspective. He says that what Blyth is describing sounds identical to what Gramsci calls a “crisis of authority” or “crisis of hegemony”.

Trump's New Coalition: A Blueprint for Authoritarian Plutocracy

Wed, 20 Nov 2024 00:00:00 GMT

Two weeks after the U.S. elections, I’ve moved from oscillating between alarm and calm about Trump’s potential re-election to deep concern.

The election produced surprising results. Voter turnout dropped from four years ago, yet Trump became the first Republican in over 20 years to win a popular majority. Though his margin was not that big, it was smaller than Biden’s or Obama’s victories, this Republican milestone marks a significant shift.

Trump’s coalition consists of three distinct groups:

The hardcore MAGA right: Politically active, media-consuming and producing, right-wing, often including blood in soil nationalists, including a larger portion of the white working class than previously seen.
A group of influential super-rich men, crucial not for votes but for other reasons.
The swing voters who secured his victory: Less ideological, often working class men, spanning all races, and notably disconnected from news media. Those reporting economic hardship due to inflation disproportionately supported Trump.

The voting patterns showed clear trends. Most demographics swung toward Trump, with two exceptions: university-educated white women and those over 65, who favoured Harris. Young men and minority men showed the strongest swings toward Trump, while educated white men showed modest movement.

Adam Tooze offers insight into this coalition’s dynamics. American society divides into three layers he says: working class, managerial class (educated middle class), and super-rich. The working class resents both rising inequality, property prices and the managerial class’s cultural authority, including policing things like what pronouns we all must use. The super-rich, unbounded by social constraints, appeal to working-class voters precisely because they can defy these mandates.

Sociologist Dylan Riley highlights a crucial status shift with respect to gender: White working-class men without degrees have fallen from 7% above average income in 1980 to 12% below today while degree-holding white women now earn 18% above average. A status reversal.

So what is Trump going to do?

Here I am just going to focus on the domestic outcome. We have a habit in democracies of discounting what politicians say because they are media-trained to be so slippy in reaction to the media’s penchant for got-ya soundbites. Which is why we often get authoritarians like Putin so wrong. Similarly, one should take what Trump has said repeatedly seriously. We can say without a doubt he does not care much about abortion, but if it helps his cause to be anti-it, he will be. We can say he cares about money, power and himself. His core beliefs include:

Mercantilism and tariffs over free trade (A theme of his since the 80’s)
Strong anti-immigration stance
A desire for revenge against perceived enemies

Dictator on day one

Trump declared he would be a dictator “on day one,” leaving open how long this might last. When Time magazine questioned whether Americans’ attachment to democracy would prevent this, he dismissed the concern, asserting Americans would welcome it. Combined with his initial appointments, this statement severely threatens democratic institutions.

His appointments reveal a dangerous strategy to back this threat, as analyzed by Anne Applebaum and Ezra Klein, as well as in this excellent podcast with sociologist Dylan Riley. His strategy targets the government’s most powerful arms—military, intelligence, and Justice Department—with appointees whose only viable future lies in absolute loyalty to him. Consider his choices: Pete Hegseth, a Fox host who denies germ theory, for Defense Secretary; and Matt Gaetz, tainted by serious allegations and shunned by his own party, for Attorney General.

The Gaetz nomination particularly reveals Trump’s intent. By forcing Senate Republicans to confirm someone they despise, Trump transforms the appointment into a loyalty test for the entire party.

The implications are clear: These appointments position Trump for retribution. Through Gaetz, he gains the FBI’s investigative powers to target Biden’s family and hostile and noncompliant generals. Through Gabbard, he can dismantle intelligence agencies. Through Kennedy at Health, he can undermine other government institutions. These aren’t appointments for governance but for revenge and busting through norms and guardrails.

Neoliberalism’s end

Trump’s presidency marks neoliberalism’s end, with two possible successors:

Genuine working-class populism is represented by figures like Vance and his convention speech, as well as organisations like American Compass. If this is it, Democrats can build alliances with the Republicans.
More likely: A hyper-libertarian vision led by oligarchs, gutting state protections (state capture, South Africans know all about it) while enabling widespread corruption, also represented by Vance (which just shows the confusion in the party at the moment).

The constitutional guardrails may prove insufficient against this onslaught.

Two Trumpy Lessons: Digital Echo Chambers and Political Tribalism

Fri, 15 Nov 2024 00:00:00 GMT

A few thoughts on Trump, our media consumption, and polarization – along with a mea culpa.

This reflection stems from an event during Easter weekend in South Africa while visiting my family, which upset both me and my relatives significantly. I learned several things from this experience, some of which I already knew but was interesting to see affect me personally.

During the April visit, a heated discussion erupted over Trump and Biden. When my family members brought up mass immigration into the US, I reacted with unexpected intensity. The irony wasn’t lost on me – I had long criticized South African liberals and progressives for their lack of sympathy toward ordinary South Africans on immigration issues. These concerns weren’t abstract: they touched on the state’s ability to provide basic services. South Africa needs to address its own problems and care for its people before absorbing millions from other countries. And I had often pointed out how people in wealthy suburbs remained disconnected from ordinary South Africans’ daily struggles.

And it’s a longstanding concern of mine that both liberals and progressives underappreciate the importance of borders and local communities. Years ago, I visited Madrid’s City Hall, which displayed leftist pro-immigration posters declaring “No Human is Illegal.” While I appreciate the sentiment (and it chimes with the values I grew up with, values that you find all over the New Testament), this isn’t how human societies function. It’s disconnected from how most people think about these issues, especially during economic and social stress. The Biden administration’s approach was divorced from reality and bound to be a vote looser.

Many of these South African sentiments about immigration mirror those expressed by Republicans and others concerned about immigration in the US. In both countries, while some opinions stem from racism and xenophobia (and, in South Africa’s case, an often deadly variant), others arise from legitimate worries about social cohesion and economic competition. So why did I react so defensively?

First, I didn’t understand the extent of what was happening. I’m a total news junky. My bias was suspecting that my family was misled by fake news — while I was ignorant!

First, I didn’t understand the extent of what was happening. I’m a total news junky. My bias was suspecting that my family was misled by fake news — while I was ignorant! So, despite typically being well-informed about major trends and flashpoints, I hadn’t realized the scale of immigration that occurred after Biden took office. I found this to be a bit of a shock. This blind spot partly stems from spending less time on X – I had reduced my usage because I disagree with Elon Musk’s absolutist (and, as it turned out, hypocritical) free speech position that has made X ground zero for misinformation, and I didn’t want to support that. But even if I had been more active there, how algorithmic filtering works on X would likely have prevented me from grasping the situation. It’s concerning how we now exist in entirely different media ecosystems. It impacts both left and right.

The reality was stark: this period saw the largest immigration surge in U.S. history. Context matters – the U.S. border had been sealed during COVID, suppressing immigration in Trump’s final years. When restrictions lifted, pent-up pressure exploded. However, Biden’s actions amplified this surge by relaxing standard immigration restrictions, going beyond merely reversing Trump’s harshest policies like family separation. His approach proved considerably more permissive than Obama’s.

The administration eventually grasped their mistake. As immigration hit record highs, so did deportations. Biden sought a bipartisan deal for increased border funding, which Trump, cajoling the House of Representatives, blocked, openly admitting it would help Biden politically. Though border crossings eventually fell well below Trump-era levels in Biden’s final months, the damage was done.

So why, then, did I become so emotional defending Biden’s position? Beyond my initial ignorance and bias regarding the scope of the situation, there are key differences between South Africa and the USA. The USA is arguably the world’s wealthiest country with near full employment. Most economists suggest that without this immigration surge, American inflation would have been higher. America’s problem isn’t people competing for a small pie – these newcomers strengthen its economy. Studies suggest this immigration surge even boosted wages for poor American workers by expanding their industries (see link). South Africa faces the opposite challenge – catastrophic unemployment and economic crisis.

Yet something deeper bothered me: seeing my Christian family fixate on this issue in the world’s richest nation, where immigrants arrive desperate and destitute. Even deeper still came this emotional reaction: This is a Trump talking point. How is this man different from South Africa’s Jacob Zuma, except with the backing of the world’s wealthiest people? Still, I should have acknowledged my agreement with their core point: social cohesion matters more than economic gains. AND this practicality: this issue would likely play a decisive role in Biden’s electoral struggles.

What did I learn besides how we suffer from a polarised media sphere? A practical example of how knowledge is, to use a fancy term, “socially constructed”.

What did I learn besides how we suffer from a polarised media sphere? A practical example of how knowledge is, to use a fancy term, “socially constructed”. See fascinating articles here and here. Our opinions aren’t purely rational – they shift with tribal allegiances. Our truth is based on our social connections. And just as Republicans once championed immigration and free markets while leftists like Bernie Sanders advocated border control, I found myself trapped in a reflexive opposition to the right’s new position. Instead of seeking common ground, I let tribal instincts drive my response. And that’s a real pity.

More than ever, it is crucial that people keep talking and seek common ground. Research shows that both the left and the right completely overestimate the prevalence of extreme opinions on the other side. This challenge is made immeasurably harder without responsible leaders, but we must keep trying.

What the DA should be telling the people of South Africa

Fri, 07 Jun 2024 00:00:00 GMT

If I were the leader of the Democratic Alliance, what would I tell the ANC, and more importantly, what would I tell the South African people?

I would say to them, look, I know many of you don’t like us, and I can understand why, given our country’s history. Some of it is also justified. However, some of it isn’t. And I think you need us, and we need you. Now, let me first deal with one thing. The heritage of the DA. We are not the party of apartheid. We come from the political party of the old white South Africa that opposed apartheid consistently. And we were right to do so.

The second thing to say is, yes, most whites support us now, some of whose parents supported apartheid, and some who did it themselves. Back then, our heritage was liberal, the party of wealthy white South Africans. But here is the thing. Even though we opposed apartheid, and we were right to oppose it, there are some things which the National Party did, which we also did not appreciate, which I wish we had.

The National Party uplifted Afrikaners, in particular, when they were very poor, and they used the government - and business - to do so. And if we are truthful, at the time they were doing this, it didn’t damage the economy. In fact, we had the highest growth in our history. But they were not anti-capital like the EFF and MK. They had careful plans and encouraged capital.

If we are all honest, we should also look at the graphs of the famous left-wing economist Thomas Piketty. And we would realise that from 1948 to 1993, the income share that the top 1% of South Africans took, declined consistently! From 1993 to now, it has shot right back up! The data doesn’t go back that long, but from the 1960s, it’s also clear that the top 10% of South Africans earned less and less of national income all the way to 1993 and then shot back up. And if we look at wealth inequality, records show that that too declined when the National Party was in power and it’s gone back up since 1993. Though the effect of the decline in wealth inequality wasn’t as great as with income, the National Party was obviously doing something right, even though they were doing a lot wrong.

This rise in economic inequality since 1993 is the ANC’s fault, but the DA did not make a fuss about it. And we should have. But the ANC and us must put that right.

The world is currently a dangerous place, but also an interesting place. China, President Biden have shown that what the NP did was not a fluke: industrial policy can work and build an economy if applied judiciously, while the private sector is allowed to play its role. I say all this because people think that if the DA came in power, we would just seek to minimise the role of government and not spend any money. If we only stopped government spending, it would not be enough to turn our economy around. We know that.

South Africa desperately needs investment. It needs infrastructure, energy, security, education and training, for starters. What we are against is wasting money - and corruption. You need a plan on how to spend money credibly to make the economy grow. If the government can’t show that it has a plan for making the country’s economy grow, then how can you expect businesses to join in and invest? The EFF and others have said, business has been on an investment strike. They are right! Businesses have not been investing, but you can’t blame them because the government needs to have a plan first and lead. Nobody wants to invest in something if they can’t see they will make a return.

So, I think the DA, if part of the government, can help make such a plan, help the government stop corruption, and invest in those parts of the economy that need it, and as the National Party did with Afrikaners in particular, help poor South Africans. If we do that, we will also create what economists would call “demand”. And if there’s demand in our economy, if people have money to spend, then our businesses will thrive because they can sell things to our people.

But what about our debt, I hear you say? The DA has talked a lot about that. That is very concerning. And if we keep spending and stealing without a credible growth plan, we will face a proper crisis where we can no longer pay. If you think things can not get worse, you are wrong. But, if we have a credible plan to invest and a team that can make that plan happen, markets will like it, and our borrowing costs will be lower. We will have more room to make the investments we need to make. We still have some room to turn things around. But time is running out. This may be our last chance. If the ANC makes a coalition with the EFF or MK, our decline will accelerate.

I think the Springboks have shown we are better when we work together - and much stronger. Together, the ANC and the DA represent more than 60% of our country’s people. Like in a rugby team, we have different skills and roles to play. The DA would be sorely missing from the government. No other party has our skills to bring to bear on making a positive change for South Africa. We are ready to pull up our sleeves, be team players and help.

So, I humbly ask that the ANC and the people of South Africa choose us as their coalition partners. Together, we can start making South Africa better.

Thank you.

It's crunch time for the ANC - and South Africa

Sun, 02 Jun 2024 00:00:00 GMT

In November 1992, I cast my first ballot in the all-white referendum on whether FW de Klerk should continue his negotiations to bring about a political settlement ending grand apartheid with the promise of a new dispensation. My friends (photo) and I all voted yes, along with 68% of South Africa’s white population.

Less than two years later, in April 1994, I voted for a second time in South Africa’s inaugural unitary elections, which the African National Congress won decisively with 62% of the vote on an impressive 87% turnout. (I also had the privilege of serving as an electoral officer at a voting station in Hammanskraal, north of Pretoria in that first election. The long queues and excitement will be something I will always remember. )

This week’s vote, the sixth since 1994, marked the first time the ANC failed to secure an outright majority in South African elections—and by a significant margin. It barely garnered 40% of the vote on a turnout of only 59%.

Instead of celebrating the maturity of South African democracy in opting for change, the currency is dipping, markets are skittish, and commentators are worried. But are their concerns justified? Unfortunately, yes.

Although he is almost universally disliked, most of South Africa’s chattering classes concur with ANC minerals and energy minister Gwede Mantashe that the ANC’s ultimate downfall is due to “Zulu tribalism.” He was alluding to ex-president Jacob Zuma’s newly formed Umkhonto we Sizwe (MK) party, which had taken the polls by storm, particularly in South Africa’s most populous province, KwaZulu-Natal, the traditional homeland of the Zulus.

Sidenote: Despite their unprecedented performance, the MK party is disputing the election outcome at the time I’ve written this, without offering any evidence as to why, or following established complaints procedures. The dispute is most likely hogwash. The elections tracked opinion polling to a remarkable degree. Zuma has shown time and again that he is willing to damage the country for his own gain.

In spite of this, ethnic allegiances are not the main problem. Though trending in that direction, South Africa is still far from being like Kenya, where ethnic voting dominates above all else. Neither the two biggest parties nor the fourth biggest can be said to represent one ethnic group.

Let me then explain what is likely to be the main worry. Because of the size of their loss, the ANC will have to have to choose one of the Economic Freedom Fighters (EFF), MK or Democratic Alliance (DA) to form a government.

The main concern, besides MK’s irresponsible posturing and rather than ethnic voting, is the stated preferences of parties like the EFF and MK, who would like to abolish the constitution and redistribute land and wealth in ways that would not only massively damage the economy but also most certainly lead to serious conflict. Based on past behaviour, there is also ample reason to believe wealth would not reach most needy people but line the pockets of a small elite.

Yes, South Africa is desperately unequal, and in some ways more so than in 1994. But the constitution already allows for land to be redistributed in an orderly manner, and with some legitimacy. That it has not happened faster has as much to do with government incompetence and startling lack of political will as anything else.

A more equitable South Africa would require far more than a flash of rash land distribution. It would also require simply fixing some basics: a far better education system, rebuilding the energy and transport systems and infrastructure, and rejuvenated policing.

Yes, fixing these systemic issues would require more money and time, which would be increasingly difficult to marshal in a shrinking economy with rising debt levels, even if the currency did not take a nosedive. South Africa is dependent on imports of many high-end capital goods from outside to keep its economy ticking over, and in spite of abundant sun and wind, this includes oil. So, what outsiders think about what goes on inside South Africa matters to prices inside South Africa.

But even money and time are not enough. Fixing South Africa requires a sweeping cultural change in how the country is governed and managed, a shift towards sacrifice instead of graft, and that extends to the culture and skills of its civil service. And all of these factors point to the need for one coalition partner.

Although the arrogance and posturing of the current leadership of the cosmopolitan but very much white-led DA make them hard to stomach, securing them as the ANC’s coalition partners and giving them key ministries like energy is the only way I foresee to stem South Africa’s decline. They are the only party that can mobilise the institutional and cultural capital, knowledge, and skills at scale to start fixing some of the country’s worst ills.

But are they not dyed-in-the-wool neoliberals? Does South Africa not need investment, while the DA believes in austerity? Yes, some in the party still do. But the global mood has swung so decisively against austerity and in favour of industrial policy that the DA cannot help but be impacted by that shift. And look at what the DA does, not what it says. The roads and infrastructure in the Western Cape, which they govern, are better maintained than those in parts of the United Kingdom.

Will the ANC choose the DA? If I were a betting person, I would not put money on it. Part of the problem is that so many vested interests benefit from corruption in the current regime. The changes needed to fix the economy would stop these people from eating.

And I doubt the maturity and wisdom inside the DA in making smart concessions to make it easier for the ANC to sell such a deal to their constituency. Everywhere, I see political pigmies and no statesmen. South African politics does not have a Rassie Erasmus or a Nelson Mandela.

What is more likely to happen? At best, the ANC forms a minority government, and it is unable to pass any new legislation through parliament while facing numerous votes of no confidence. The government is even more of a lame duck than before, and economic decline accelerates as the country lurches on. At worst, the ANC chooses the EFF or MK as a partner. The currency tanks. The economy goes into an even steeper downward spiral. Who knows what would happen next?

GPT7: When AI Becomes a State Affair?

Mon, 27 May 2024 00:00:00 GMT

Large Language Models (LLMs) like GPT-4 have faced criticism from many, not to mention from many within the AI research community. As discussed previously, the critique cuts to the core: not only do they not resemble human intelligence, but they lack real understanding, functioning merely as a sophisticated form of auto-complete.

Yet there are others in the community with a very different opinion. When GPT2 came onto the scene, some thought something significant was happening within deep learning. With the transformer architecture on which ChatGPT is based in particular, the larger these models became, the more data they were fed, that they were showing “emergent” properties that they were not trained for.

When we say “larger models,” we generally mean the number of parameters it has. The parameters of an AI model control the size and bias of the pathways between the neurons of the model, pathways that are learned from the training data. Think of them as the model’s internal variables that shape its actions and effectiveness.

Sam Altman from OpenAI, Dario Amodei of Anthropic and Demis Hassabis of Google Deepmind all believe that large language models (LLMs) possess a type of understanding and suggest that with sufficient data and parameter size, they could achieve superhuman intelligence. We can call this way of thinking the “scaling hypothesis of intelligence,” founded on two straightforward ideas: a larger “brain” correlates with greater intelligence, and there’s a functional similarity between artificial and natural neurons. The scaling hypothesis has been substantiated in scholarly articles, and the findings remain impressively consistent: Expanding the models’ parameters and concurrently amplifying the training data improves their performance.

The neat thing about this point of view is, that if it is true, then reaching superhuman AI will require few new scientific breakthroughs. It is mainly an engineering and financing feat. So there’s been a group of influential AI researchers that argue that this is just the beginning for large language models (LLMs) and that we’re entering an era of exponential enhancement. Some go even further and postulate that consciousness might just be a by-product when a certain scale is reached.

Gary Marcus, a well-known sceptic, points out a critical observation: it’s a year and a half since ChatGPT’s release, and while Google, Anthropic, and recently Meta have unveiled models that may surpass GPT-4 in some aspects, they’re all within the same range of performance. He argues that if we were to witness exponential growth, at least one should have demonstrated a significant leap forward by now, and a more profound leap than the progression from GPT-3 to GPT-4. Thus, he asks whether large language models (LLMs) hit a performance ceiling?

Marcus may well be right, but there’s another possibility. The scaling hypothesis might explain why we are not seeing a massive leap, because it’s becoming harder to achieve. The hypothesis posists the exponential need for inputs, to achieve linear improvements.

The scaling hypothesis might explain why we are not seeing a massive leap, because it’s becoming harder to achieve. The hypothesis posists the exponential need for inputs, to achieve linear improvements.

The hypothesis argues unambiguously that training more advanced large language models (LLMs) with the transformer architecture requires exponentially more resources. And we have seen that in the past. For instance, GPT-2’s training was relatively modest, costing around $45,000 in compute power. In contrast, GPT-3, is roughly 10 times bigger in data and parameter count, and incurred a cost 100 times greater, soaring above $4 million.

The specifics of GPT-4 remain under wraps, yet it’s speculated that it increased by a factor of nearly 10 in both parameter count and data used. The training expense is estimated to be between $50 and $100 million. Regarding data volume, it’s believed that GPT-4 was trained on an amount comparable to that held in the US Library of Congress, equivalent to approximately 10 trillion words. Yet, although GPT4 is considerably better than GPT3, and GPT3 better than 2, their superior performance on benchmarks is linear, not exponential.

You can see where this is heading. To get another improvement on benchmarks of the same magnitude, GPT5 will probably have to be trained on something like 10 times the data GPT4 was trained on, and be a 10 times larger model, and cost 100 times more to train — in the order of a couple of billion dollars or more.

That is an eyewatering sum. But very much within the budget ranges of most big tech companies who typically fund these training runs. But also, to get that amount of quality data together to train on isn’t trivial. Epoch AI projects that the pool of high-quality text data will be exhausted between 2024 and 2026. They estimate that approximately 10 trillion words of such data are available, an amount akin to what was used to train GPT-4. This data includes sources like books, news articles, scientific papers, Wikipedia, and curated web content.

This view may be overly pessimistic. For instance, Anthropic’s Claude 3 model partially relies on internally generated data for training, a form of synthetic data. Some fear this is a kind of intellectual inbreeding, but Claude 3 works well. Similarly, ChatGPT is creating a substantial amount of data through user interactions, already thought to be equivalent to the volume it was trained on. However, with heightened awareness of data’s value and the flurry of copyright litigation, procuring sufficient data might become a complex and costly endeavour.

No wonder, then, that GPT5 is taking a while to arrive. But when it does, I predict it will be noticeably better than GPT4—but not enough to silence the detractors. And if the scaling hypothesis remains the only pathway to machine intelligence, expect a new set of players to come to the fore by the time we get to GPT6: States with their massive financing abilities. And GPT7?

Only two people on the planet will be able to afford to build GPT7, Xi and Biden, or their successors.

This first piece first appeared in the Vrye Weekblad, where I write a bi-weekly column on AI.

The left has no clue about culture

Sun, 05 May 2024 00:00:00 GMT

A few quick thoughts on culture. Just this week, I wrote about how Large Langauge Models (LLMs) are powerful models of culture, for good and bad. Culture has been a bit of an obsession of mine, partly because I think that the failures of South Africa have been approached by both liberals and progressives in lots of books, papers and so forth without nailing the origin of the problem.

Liberals and progressives tend to ignore the influence of culture on politics and society, which is a particularly acute mistake to make in the case of South Africa. I’ll delve into why I say that more on another occasion.

But now I want to highlight a report about the cultural dimension of US politics from the Roosevelt Institute, which I came across this week. It’s really interesting because it articulates many of the hunches I have about what is wrong with today’s politics.

It deals with the idea that the left, particularly in the US, has successfully started grappling with neoliberalism’s failure in policy terms. Now we’re thinking of Biden’s raft of legislative and policy measures that aim to increase the income of people on the lower rungs of society, bring manufacturing back to the US, and so forth. Yet, we see how the right keeps being extremely successful and salient in the people’s lives that these policies are designed to help. Why is this so?

The report argues that the right has cultural arguments that address deeper needs than the superficial economic ones and that the left doesn’t have a cultural response in return. The report points out that neoliberalism itself, although an economic project, had a cultural dimension.

The culture of neoliberalism

This cultural dimension of neoliberalism encompasses elements such as extreme individualism and grind culture, which promotes the idea that if you work hard enough, you’ll reach the top and shame those who don’t as failures. It also includes the pervasive belief adopted, especially on the left, that we have the freedom to make choices in life and direct not only our own destinies but also design our own identities, including that where we live is merely a lifestyle choice. And, of course, all this is underpinned by the neoliberal idea that free markets are the solution to all our problems.

The symptoms of the failure of neoliberalism

Today, the economic ideas behind neoliberalism are called into question. However, the results of this failure are not only economic. What are the symptoms of this failure of neoliberalism on a social and cultural level? Well, the article mentions the mounting despair and the widespread insecurity people feel financially, physically and for their communities. Widespread overwork, the lack of agency many experience with respect to changing their own lives (let alone society), social isolation, drug abuse, the increase in suicides, and so forth. These are all symptoms of neo-liberalism failing to deliver its promise for most people.

The report lists several reactions to these failures of neoliberalism, which they have usefully organised into archetypes, though there may be considerable overlap between them for many people.

Archetypes of reactions to the failure of neoliberalism

First, there are the Strivers, the few Mark Andreessen’s of the world - and many others who are just about successful, perhaps because they managed to get on the property ladder somehow — or still hope they can make it. They are the people who feel they have, or still need to pick themselves up by their bootstraps. The Strivers are prone to wellness and self-help culture. It’s the appropriate individualistic response to the demands of neoliberal society by those who have taken it to heart.

The Strivers are prone to wellness and self-help culture. It’s the appropriate individualistic response to the demands of neoliberal society by those who have taken it to heart.

Second, there are the Innovators. This is an archetype I didn’t quite get at first when I looked at the report. They’re not talking about entrepreneurs here. It’s the people who innovate new community-centred arrangements that satisfy their desire for connection. I can think of the LGBQT community’s “found family” or many charismatic churches’ “cell groups”, collective living projects and people creating alternative moral orders and identities, for example, the plethora of sex-positive and other events and communities you see nowadays. I’ll admit it is something that I’ve been thinking a bit about: creating community.

Then there are the Dropouts, the people who escape into destructive or mind-altering drugs or look for other ways of withdrawing to private worlds. People who stop reading the news and disconnect from the debates in society can also fall into this category.

Lastly, there is the fourth group, the Rebels, who rebel against the failures of neoliberalism. This has come from both the left and the right, but more successfully and sustainably on the right, with the MAGA movement being a prime example.

Scratching cultural itches

So why has the right been more successful? Well, partly because they’ve not only scratched cultural itches, but they are also offering a soothing balm, not just economic policy ones. We can point to diverse figures and movements on the right, such as Jordan Peterson, QAnon, the MAGA movement itself, Russell Brand in the UK, and Meloni in Italy. Quite diverse, but these individuals and movements have adopted a culture-first orientation.

Another useful dimension in the report is that the Roosevedt Institute articulates five elements that are creating what they call “tides of longing”, as a response to five dysfunctions of neoliberalism.

Tides of longing

The first is a longing for community in response to detachment and isolation. The second is safety in response to economic and other insecurity. A third is the desire for agency or control over our lives because of an uneven distribution of power. I think this category is also reflected in the massive differences in status in contemporary society.

The fourth element is the need to understand what is going on in the face of rapid and constant change. Hello, conspiracy theories! Also included here is the institutional failures that we are seeing, which are not helped by the rapid technological change we are going through. I’d argue that the need here is not only for understanding but for new or invigorated institutions which people feel they can trust. Finally, they argue there’s a need to feel better because of a world that often causes unease.

What could a left cultural project look like?

After reading it, the big question for me is this: Can the left deliver an alternative cultural vision(s) that addresses these issues? This is no easy task, as it requires grappling with complex issues such as the role of religion and spirituality, the nature of community and identity, the significance of place and location, and rebuilding trust and institutions.

The left’s tendency to prioritize rationality and reason, along with their scepticism towards the concept of the nation and their preference for self-actualization and niche identities, may further hinder their ability to connect with wider groups of people on a deeper emotional level.

Additionally, the cosmopolitan worldview often espoused by the left, which downplays the importance of borders and local communities but instead values disparate social networks, further complicates their efforts to address people’s longing for a sense of belonging and connection to place.

As we navigate the post-neoliberal landscape, the left and progressives must engage with these cultural questions and develop a compelling vision that speaks to people’s fundamental needs and desires. Only by understanding human needs that can not be satisfied by material things alone can we hope to build a more equitable, fulfilling, and resilient society.

Beyond the brain

Fri, 03 May 2024 00:00:00 GMT

One of the most trenchant criticisms of large language models (LLMs) like ChatGPT is that since they are simply trained on copious amounts of data found on the Internet, statistically making correlations, they are mere probabilistic parrots of language with no real reasoning ability.

If correct, several problems arise, critics point out. For example, since the data they are trained on are not curated, they are bound to regurgitate society’s dominant biases. Google’s Gemini recently gave an apt demonstration of this problem, angering many.

For example, when Washington Post columnist Megan McArdle asked Gemini to write something condemning the Holocaust, it straight up did so. But when she asked it to write something condemning the Holodomor or Mao’s Great Leap Forward, it pointed out how complex history is and then contextualised what took place.

South Africans have long been aware of a similar issue. For years, Google Translate, which utilizes the same machine learning techniques, has translated the Afrikaans word “boeremeisie” to “peasant girl” in English—a reflection, perhaps, of the English-language corpus’s unflattering portrayal of Afrikaners.

A more fundamental critique is that generative models like GPT-4 aren’t truly intelligent in the way humans are. Their outputs may seem meaningful and well-formed, but this masks a lack of real comprehension and a grounding in reality. Instead of the intentional communication humans do, GPT-4 merely calculates the likelihood of sequences of words based on its extensive training data, navigating an immense “latent space” of potential sentences without grasping their significance.

Models of culture

Ted Underwood, who teaches at the School of Information Sciences and in the English Department at the University of Illinois, applies machine learning to literature. He agrees that LLMs are, rather than human-like intelligences replete with subjectivity, the ability to plan and to desire, more like models of something collective—namely culture—and, therefore, replete with bias.

Underwood says intelligence comes in many forms, some of which we share with other animals, like, for example, the ability to plan ahead, which LLMs cannot yet do. But language is something that, up to now, has been uniquely human. Perhaps language is even more distinctly human than intelligence as such.

Francios Chollet, Google deep learning pioneer and the creator of the Keras deep learning software library, agrees. He points out that we have a bias when we think about cognition: focusing on the brain. Yet he and many cognitive scientists now believe that the brain is not the only place where thinking happens.

The environment we find ourselves in is vital for thinking, too. Our abilities cannot be explained by just one disconnected brain learning independently in a single lifetime. As evidence, Chollet points out that our intelligence is greater than that of the great apes. But not that much more that it explains the difference in our capabilities. So, what explains the vast differences? It’s our culture, language and civilisation, says Chollet, the vast scaffolding on which we rely and allows us to live elevated lives.

Rage With The Machine

Culture and language are thousands of years of accumulating knowledge, skills, and mental models of the world, says Chollet. So, when you are born into this system, you don’t have to reinvent the wheel. It follows that models of the world, including models of language and culture, can be important tools for thinking in their own right, even if they are not intelligent in the same way humans are.

That does not mean LLMs aren’t dangerous. Not all aspects of human culture are beneficial, and culture extrapolated through technology can lead us to dark places. Underwood highlights this by referencing movements like QAnon, illustrating how humans are prone to creating alternate realities. When bolstered by the power of sophisticated deepfakes, these fabricated worlds could become overwhelmingly convincing.

Worlds Apart

Talking about alternate realities. Vernon Vinge, a visionary in science fiction, passed away two weeks ago, leaving a profound legacy. More than any other science fiction writer Vinge grappled with the dilemmas artificial intelligence may present us with. Borrowing from cosmology, he was the first to come up with the concept of “the singularity” with respect to technology: a pivotal event that would occur due to the development of superintelligent entities, resulting in unforeseeable changes to human civilization.

The very first science fiction Vinge ever sold was a short story called “Apartness” to the British literary magazine New Worlds in 1965. It tells the story of a post-apocalyptic Earth in which the protagonist discovers an isolated group of people in Antarctica.

Confused with “natives” of the frozen continent, the protagonist describes them as hairy and of an indeterminate race, but primitive in every physical sense and inferior in their tools. Later, the protagonist discovers two shipwrecks, one clearly marked as the “Nation.” The other’s name is nearly worn away by the action of ice and water on its hull, but one can just about discern “Hendrik Verwoerd.”

PS: Culture is both misunderstood and underrated as a force in society. In a wonderfully erudite and wide-ranging lecture delivered in London last week, Terry Eagleton explored the multifaceted nature of culture. It’s well worth a listen.

I write a bi-weekly column on AI for the Afrikaans language Vrye Weekblad. This piece first appeared there.

An AI that knows you and your business

Thu, 14 Mar 2024 00:00:00 GMT

Cohere’s new Command-R model does two things many AI models do not, that are essential for business: Retrieval Augmented Generation (RAG) and tool use.

It matters that AI has easily updatable memory. Without dynamic memory, AI cannot act sequentially on a plan or get to know you or your business.

Up to now, one method to address this, besides expensive one-off fine-tuning, is the so-called Retrieval Augmented Generation or RAG. At the heart of RAG lies the principle of leveraging a repository of information (like documents, web pages, or database entries). The relevant documents are retrieved, usually via search functionality, which is then fed to the generative AI’s context window. RAG is, therefore, only as good as the retrieval functionality and is also limited by the context window size and accuracy.

Cohere claims its model, Command-R, features a longer context length, supporting up to 128k tokens (that’s about 80,000 words in English), and that, like Google’s Gemini, it scores perfectly in a needle-in-a-haystack text (where it is tested to find information anywhere in a document).

GPT4 and Claude, by comparison, context windows are lossy, missing much that’s in the middle of long documents.

Command-R can be accessed via an API or self-hosted.

Taking stock of AI's impact

Mon, 11 Mar 2024 00:00:00 GMT

The AI Horizon - Mirages and Milestones: There remains a lot of uncertainty and often diametrically opposed opinions about many of the vexing questions around AI. A comprehensive survey involving 2,778 top-tier AI researchers revealed significant differences in predictions about AI’s development pace and impact. Aggregating key projections suggest a 50% likelihood that by 2028, AI will autonomously build a payment site, compose a hit song, and fine-tune large language models independently. Notably, the probability of AI surpassing human cognitive capability across all tasks is seen as 10% as soon as 2027, reaching 50% probability by 2047—13 years sooner than previous estimates given just a year ago. Yet, full automation of all human jobs is given a 10% chance by 2037, with a 50% likelihood way into the future in 2116.

Surpassing GPT4 with Context Windows: Efforts to outperform the capabilities of proprietary, centralized models with smaller, open-source alternatives have intensified. Notably, neither open-source nor propriety ones have beaten ChatGPT’s GPT4 on cognitive benchmarks since its release. Everyone expected Google to lead this, but after unexpected false starts, it has made a significant breakthrough with the release of Gemini 1.5. A proprietary, centralized model itself, it stands out as the first to not only match but, in certain benchmarks, surpass the performance of ChatGPTs GPT4. And in the last fortnight, yet another propriety model, Anthropic’s Claude 3, broadly matched GPT4. Gemini 1.5 introduces multi-modal capabilities, processing inputs across text, audio, and video—a trend gaining momentum in generative AI. Perhaps most significantly, it boasts an impressive “context window” capable of handling up to 1 million tokens. This advancement enables the analysis of entire books under 200,000 words or comprehensive code bases in a single query. In comparison, the current iteration of ChatGPT has quadrupled its initial context window size to 32,000 tokens. The drive for longer context windows reflects the weak points of current model architectures. They have nothing akin to the dynamism of human memory. They are pre-trained or primed on new information via fine-tuning, an expensive process that needs expert human intervention and bespoke data. However, models can be given new information via their context window, a form of working memory limited in size due to the computational cost of processing them at scale — until Gemini.

RAG and Beyond: Research into giving AI easily updatable memory matters. Having no dynamic memory limits an AI’s ability to act sequentially on a plan or to get to know you or your business. Up to now, one method to address this, besides expensive one-off fine-tuning, is the so-called Retrieval Augmented Generation or RAG. At the heart of RAG lies the principle of leveraging a repository of information (like documents, web pages, or database entries). The relevant documents are retrieved, usually via search functionality, which is then fed to the generative AI’s context window. RAG is, therefore, only as good as the retrieval functionality and is also limited by the context window size. Multiple promising research tracks are now looking at new model architectures (e.g. state space models) that could endow models with something like a dynamic memory (aka episodic memory), which would be another major leap forward.

Embodied Intelligence: The prevailing view suggests that human cognition extends beyond the brain, utilizing the entire body and more for thought processes. This concept has implications for artificial intelligence, where models integrated with sensory and motor functions, or ‘embodied AI,’ are believed to surpass their text-only counterparts. PaLM-E exemplifies this theory as a cutting-edge, embodied generalist model. With 562 billion parameters and training across visual, linguistic, and robotic datasets, it can manage a robotic manipulator in real-time and achieve unprecedented performance on visual question-answering benchmarks. But PaLM-E’s embodiment also allows it to excel in language tasks, especially those requiring an understanding of spatial relationships, showcasing a significant leap over traditional, text-only AI models.

Data Drought: Quality data is becoming only more important for training high-performance models. According to a study by Epoch AI, if current trends in data usage and generation persist, we are on track to deplete our reserves of low-quality language data as soon as 2030, exhaust our supply of high-quality language data by 2026, and run out of vision data sometime between 2030 and 2060. Another avenue is the use of AI-generated synthetic data, though the effectiveness of this approach has shown varying results.

Behind Walls - The Changing Face of the Internet: The growing appreciation for data’s worth is transforming the open nature of the Internet. Platforms like Twitter and Reddit are now inaccessible without a login, while the New York Times has initiated a copyright lawsuit against OpenAI for training on its data. And while Getty Images is also suing the owners of Stable Diffusion, Shutterstock has done an image data deal with OpenAI. Moreover, as revealed by internal documents, Tumblr, WordPress.com are reportedly planning to monetize user data through agreements with Midjourney and OpenAI. Reddit will let “an unnamed large AI company” access its user-generated content in a deal “worth about $60 million on an annualized basis”. Meanwhile, Apple, very much behind its competitors, reportedly offers publishers large sums to access their data.

Social Media Data Generators: In another twist, illustrating the data generation flywheel (and value) of social media platforms, new analytics data shows that compared to the most popular incumbent apps such as YouTube (which has a 1-month user retention rate of 85%), Instagram, TikTok or WhatsApp (with a DAU/ MAU ratio of 85%), GenAI apps such as ChatGPT (1-month retention 56%; DAU/MAU 14%), Runway or Character.ai suffer from significantly lower median retention and daily active users, meaning they are also generating a lot less data. In May 2023, only 14% of Americans had used the free and far less able GPT3 tier of ChatGPT, according to Pew.

Europe, And Public Funded Research - Not In The AI Race: The most cited research in AI in the last three years has come from very few places, with Europe lagging markedly and the United States dominating the field. It accounts for 70% of all highly cited AI research papers in the past three years, positioning it well ahead of China and the United Kingdom, which rank second and third, respectively. Combined, the contributions from Europe’s best performers, Germany and France, match those of the UK. Remarkably, over 40% of the most influential papers originate from just two entities: Google and Meta. Despite this concentration, there’s a budding trend where major tech firms (Meta excepted) have begun withholding their research findings, marking a shift in disseminating AI advancements.

Europe’s Lack of AI Capital: The revival in the S&P 500 during 2023 can largely be credited to “The Magnificent Seven” - Apple, Microsoft, NVIDIA, Alphabet, Meta, Tesla, and Amazon - which have emerged as major contributors to and beneficiaries of the rapid acceleration in AI technology. Meanwhile, US AI enterprises have attracted 70% of the world’s private AI investment in 2023, marking an increase from 55% the previous year. Investment in European AI ventures has decreased by more than 70% at the same time. Mistral, the leading European start-up and rumoured to be the reason Macron intervened to reduce the regulatory burden of the AI Act, secured a round of funding led by Silicon Valley’s Andreessen Horowitz in December 2023, underscoring Europe’s capital malaise.

Narrow AI - Wide Impact: Despite receiving less attention, significant progress is continuously being made in specialized or “narrow” AI, particularly the life sciences. Here is just one sample. Creating entirely new proteins with specific desired functionalities or structural characteristics, known as de novo design, is of significant interest to both academic research and the industrial sector. Leveraging their demonstrated success in generating sophisticated models of images and text, diffusion models are now being explored for their potential to revolutionize the field of protein engineering from the ground up.

The AI Paradox - Jobs Amidst Uncertainty: Across the Western world, we are at a time of exceptionally low unemployment. And overall, the most recent research on robot and AI diffusion points to positive productivity gains and an increase in employment at the firms that adopt it. Employment effects at the industry level are less clear-cut. Yet, anecdotes of AI-induced job losses abound. It may simply be too early to observe the aggregate productivity and employment effects of AI.

A Future Unevenly distributed: While acknowledging considerable uncertainty, the IMF predicts that nearly 40% of jobs worldwide will be impacted by AI, with workers in advanced economies facing higher exposure but also standing to gain more on aggregate from AI’s benefits compared to those in emerging and developing countries. In advanced nations, around 60% of employment involves cognitive tasks that AI could impact soon, with potential outcomes evenly split between job losses and productivity boosts through AI being integrated with workers’ tasks. This would lead to higher wages for those workers and a widening gap between these winners and the losers. Key, however, should be the aggregate higher growth and overall increase in employment in these economies. Emerging markets see a 40% exposure rate, while only 26% of jobs in low-income countries are similarly touched by AI. However, the limited exposure of these economies to AI widens the aggregate digital and economic gaps between wealthy and poor regions. If these productivity levels are achieved along with large-scale job disruption, expect calls on governments to help workers transition to new roles to grow.

The 700-Agent AI: Klarna, the Buy Now Pay Later pioneer, reports remarkable efficiency gains with its OpenAI-powered chatbot, achieving the workload of 700 full-time employees since its worldwide debut last month. This AI assistant has managed 2.3 million conversations, accounting for two-thirds of all customer service interactions. The implementation of the chatbot has resulted in a 25% reduction in follow-up queries, enabling customers to find solutions in under two minutes—a significant improvement from the previous eleven-minute average. A survey by legal publishing firm Lexus Nexus indicates that the majority (80%) of Fortune 1000 executives expect to see a reduction in billing from outside legal counsel due to the ability to bring work in-house because of generative AI (80%).

AI Joins the Culture War: ChatGPT has become a flashpoint in a series of heated cultural debates, largely in the US, with particularly conservatives sharing screenshots to allege bias in ChatGPT’s training and fine-tuning. The heat was turned up several fold when a viral post showed that Gemini’s recently launched AI image generator created an image of the US Founding Fathers, which mistakenly included a black man. Elon Musk is working on his own models and Musk emphasised that “our AI can give answers that people may find controversial even though they are actually true”. Meanwhile, Mistral, Europe’s leading general purpose AI start-up, has launched a new Mistral Large model that is touted as a multilingual, culturally sensitive alternative to models that have only been trained on English language data. It is natively fluent in English, French, Spanish, German, and Italian, with “a nuanced understanding” of grammar and cultural context. Meanwhile, the Spanish PM has announced a private-public partnership to build an AI model in Spanish and other regional languages.

Chip War Champs - Too Early to Call: Amid tightening restrictions, Huawei surprised the world with its new Mate 60 Pro phone, powered by the advanced Kirin 9000S chip produced by Chinese chipmaker SMIC. Yet the 7-nanometer chip is no longer state of the art, and this advance still leaves SMIC lagging behind Taiwan’s TSMC by half a decade. With access to more advanced lithography machines from the Netherlands’ ASML cut-off, it may be much harder for China to achieve 5nm. In 2023, all eyes were on NVIDIA’s new H100 GPU, the more powerful successor to the A100, both banned for export to China. While H100 clusters are being built, researchers rely on the V100, A100 and RTX 3090. It is quite remarkable how much competitive longevity NVIDIA products have The State of AI 2023 points out: the V100, released in 2017, is still the most commonly used chip in AI research. This suggests A100s, released in 2020, could peak in 2026 when the V100 is likely to hit its trough. It also suggests we won’t know if Biden’s chip ban is successful for some time yet. Meanwhile, an analysis by Bloomberg shows that an invasion of Taiwan would have a bigger impact on the global economy than the financial crisis, partly because of the impact on chip availability and resulting Western sanctions impact on trade.

Social Engineering Phishing Spike: Darktrace reports a staggering 135% increase in these phishing attacks among its extensive customer base from January to February 2023. This trend is strongly associated with the swift integration of large language models (LLMs) like ChatGPT into mainstream technology. In a year that sees multiple elections around the world, the misinformation threat that AI poses is one of the most immediate.

The EU AI Act’s Wings Clipped: In December 2023, following last-minute French concerns, a consensus was reached on the EU AI Act to protect European start-ups from burdensome requirements when developing generalist AI models. The legislation distinguishes between general purpose models and those deemed to pose a systemic risk, which is determined either by the computational power used in training—specifically, if it surpasses a benchmark akin to GPT-4’s capabilities—or through identification by the Commission. Providers of AI models categorized under systemic risk must now conduct standardized evaluations, address and mitigate potential systemic risks, monitor and report any incidents, and ensure cybersecurity measures.

AI Act Exclusions - Military, Research, Open-Source and Fraud: The AI Act delineates specific exemptions, excluding military AI systems and those solely dedicated to scientific research and development from its purview. Free and open-source AI systems are exempt unless they fall into prohibited or high-risk categories. Notably, AI systems used in credit scoring are classified as high-risk, entailing additional requirements. However, AI systems mandated by Union law for financial fraud detection or for assessing the capital requirements of banks and insurance companies do not fall under the high-risk category. The European Parliament has revised its stance on the outright prohibition of real-time remote biometric identification, opting instead for an approach that allows narrow exceptions for law enforcement, particularly in preventing terrorism or identifying individuals involved in serious crimes. The regulations for post-event usage of this technology will be more lenient too.

Europe’s Regulatory Data-Danger: The agreement on the AI Act was hailed as a success. Yet the impact of EU regulations on the development of AI for the benefit of the European economy may lie elsewhere, in regulations that touch on data, like the Data Governance Act, Data Act, DSA and GDPR. The most impactful AI relies not only on data at scale for its training but also on scope, linking diverse datasets together. The GDPR and subsequent European court rulings like the SHREMS cases make linking data sets difficult and fragment data across borders. The Data Governance Act purports to enable data sharing but explicitly prohibits linking datasets without providing incentives for such sharing. This is a major oversight when the value of data is being reassessed upwards.

The political economy of AI

Mon, 11 Mar 2024 00:00:00 GMT

Size matters in AI. Ground-breaking advances in AI models hinge on the ‘scaling hypothesis’. The hypothesis posits that larger models, encompassing more data, that are more computationally expensive and have more parameters, excel in emulating intelligence across diverse tasks. Where previous advances in computing became rapidly cheaper, the reverse may now hold.

The importance of scale and its cost has implications as to who will benefit most from AI, and it will be felt from geopolitics to competitiveness to the structure of our economies. It is thought that GPT4, which is a large language model (LLM), cost more than 100 million dollars to train. At this early stage, only a handful of companies with deep pockets can compete in this race to create the most powerful AI. Among those disadvantaged are researchers working for educational institutions, with implications for public access to knowledge.

Most states could easily afford to spend this kind of money, however, but as of yet, have not yet entered the race in all seriousness. But that is starting to change. An emerging trend to watch is the rise of AI nationalism: Because AI will supercharge economies and boost military capabilities, expect states that can to start investing for the sake of technology sovereignty and economic independence and put in place other measures, like export controls on semiconductors –– to support the local development of AI.

Market concentration

The internet turned out to be incredibly beneficial for the development of sprawling digital platforms, largely due to the phenomena of economies of scale and network effects in digital networks. The digital landscape enabled companies to engage with a worldwide audience at a significantly low incremental cost. Simultaneously, the intrinsic value of a platform for users escalates as its user base expands, while its ability to gather data does too. Now, the extensive costs and substantial data requirements associated with training top-notch AI models, plus their emergent properties, will further intensify this trend towards centralization.

The concept of emergence is particularly intriguing in the context of what are referred to as “foundation models” in artificial intelligence. These extensive models, like GPT4, are trained on vast unlabelled datasets, and exhibit emergent behaviour rather than explicitly intended behaviour. In this process, unexpected properties may surface. For instance, a model that has been trained on a vast language dataset may also produce code or music notation without any explicit instruction. While fascinating, this dynamic raises concerns about potential challenges to competitive balance within the digital economy, as the top foundational models open doors for vertical integration.

On top of this, the user-friendly, flexible, and familiar nature of language interfaces facilitates the collection of yet more data and an affinity with a service, enhancing the value of these services to users and leading to a state of lock-in. Moreover, these language interfaces are set to intensify the process of disintermediation initially brought about by the internet. Legacy media, brick-and-mortar stores, travel agencies, and shopping malls have already felt the impact of this shift. Now, it seems even former disintermediators like search engines may face a degree of disintermediation themselves.

Centralisation vs Proliferation and Geopolitics

Containing the risks associated with the most powerful AI models will depend on whether they proliferate or remain in the hands of just a few players. Control of the hardware (semiconductors) used to produce the most powerful models will help the West keep an edge over potential adversaries like China in the short term and help reduce AI risk more generally. However, trained models are already software and can very easily be copied. Who knows if this or hard-nosed business reasons is why OpenAI switched from an Open Source model to a centralised one, preferring the provision of access via application programming interfaces (API’s). Mistral, on the vanguard of Open Source AI until recently, has removed any reference to their commitment to Open Source from their website with the release of their most powerful model.

Expect this trend towards centralisation for the most powerful models to continue and even for governments to mandate it if AI becomes more powerful.

Our current AI acceleration started in 2012

Sat, 09 Mar 2024 00:00:00 GMT

Most, even inside the AI community, were blindsided by the capabilities of ChatGPT, despite the many recent advances in the field. Pinpointing the exact moment when the machine learning revolution began is challenging, as advancements have been steadily unfolding for several decades. However, if one has to choose a year, the year 2012 would be it. That year was indeed a significant milestone in the field of deep learning, a subfield of machine learning.

This significance stems from AlexNet, a deep learning model that won the 2012 ImageNet Large Scale Visual Recognition Challenge, a competition for computer vision algorithms. AlexNet significantly outperformed the previous year’s winner, marking a dramatic shift in how these tasks could be approached. It demonstrated the potential of convolutional neural networks (CNNs) and deep learning more broadly, igniting increased interest and investment in the field. The following years saw many critical advancements, such as the development of recurrent neural networks (RNNs) for sequence data, the introduction of GANs (Generative Adversarial Networks) in 2014, and the advent of transformers in 2017, which revolutionized natural language processing and led to models like GPT (Generative Pre-trained Transformer), which powers ChatGPT.

Indeed, benchmarks in machine learning have regularly been exceeded in the last decade and become obsolete as the technology progresses. The ImageNet challenge is one of the few older benchmarks still active and competitive. The current state-of-the-art model on ImageNet is BASIC-L (Lion, fine-tuned), which achieves a significantly higher level of accuracy than humans.

The holy grail in AI is to achieve artificial general intelligence (AGI). In machine learning, the ability of a model to “generalise” from the data it was trained on to new data it had not seen before is key. However, the term AGI is a quantum leap more than that ability. It refers to an AI that can deal with any data and in any context in an intelligent fashion like humans can. Of course, such an AGI will have additional capabilities that humans do not have, like much larger memory and the ability to network with and effortlessly learn from what other AI’s have learned.

Until recently, knowledgeable insiders in AI community thought the technology was underhyped. Not any more. In 2023 year Alphabet CEO Sundar Pichai felt emboldened to say that AI is a more profound technology than fire or electricity. So fast are things developing that making accurate predictions about its impact is hard. Yet we are able to see the contours coming into view, and AI will have a sweeping impact on nearly everything, from how we work to geopolitics. As a dual-use technology, not all of that will be positive, especially if we make injudicious decisions about its use. Even what turns out to be a benign change in the long term, like a new age of productivity and economic growth, could be disruptive and painful for some in the short run.

The EU is set up to fail

Sun, 26 Nov 2023 00:00:00 GMT

The election of Geert Wilders in the Netherlands is remarkable, considering how the Dutch see themselves and their liberal heritage. I’m new here so I, apart from being a part of the problem for the PVV voters, can only but give very prelimary impressions of a newcomer.

Caption: Sinterklaas and Zwarte Piet on the Saturday before the vote in the streets of Den Haag

The day after the election, I listened to a lunchtime TV programme where they asked people why so many voted for Wilders’ PVV. The reaction: the middle classes, despite everybody saying the Netherlands is a rich country, are suffering. Besides migration, they repeatedly mention the soaring cost of living and house prices in particular.

But also the protests against Zwarte Piet and climate activists blocking highways, and the fact that for years, the centre-right Rutte government has been saying it will address these issues, but it hasn’t, and it’s only gotten worse.

And, of course, migration is blamed for this suffering of the middle classes. The high house prices are a good example of the muddled scapegoating, here too, migrants are said to be responsible, and not the primary culprit, the quantitative easing (QE) in response to the financial crisis that saved the world - and the Dutch - economy from collapse.

QE worked as intended by lowering interest rates and inflating asset prices. The problem is property is a very attractive asset. During Covid, QE was extended, along with other help, and interest rates plummeted further, and house prices grew even faster.

The Netherlands is a rich country

Some context, the Netherlands is indeed a rich country. GDP per capita in 2021 was $11,000 USD higher than in the UK, and it’s about $5,000 higher than Germany too.

And its not as if these riches are not shared. It’s a much more economically equal country than the UK or even Germany, with a Gini coefficient, a measure of income distribution, of around 0.29, which places it among the top 12 globally, with a more equal income distribution than Sweden, Denmark or Finland. A number that has not changed significantly since the early 90s.

However, it should be noted that the devil is again in the details, and Gini is a rough guide. Preliminary data shows top income earners benefitted from the pandemic, relative to the bottom 50%. And that the bottom 50% has been taking about a 4% lower income share when compared to the 1980s. See the graph below.

The population has increased steadily, with no marked increase in the last 20 years. It grew faster in the 1960s and 1970s. And the debt to GDP ratio is a very low 46.7%. The major discontinuity from past trends is that growth since the financial crisis has slowed markedly, not surprising since more than 60% of its trade is with Europe, which suffers from the same problem. However, it has done better than the most developed Western European countries, bar Germany.

So, on the face of it, the crude numbers don’t show the basis for the unhappiness. But looking closer may reveal trends within these numbers that tell us more. So for example, birthrates of the “native” Dutch could be less, and immigrants a bigger percentage of the population even if migration is similar to before. That is indeed what the right here thinks.

Things will only get worse

It is all the more depressing when this rightward shift is plotted against future trends. Climate change is only starting and will add to the pressure the Dutch middle classes are facing. Things will likely get worse, too, on the immigration front, with more pressure likely because of climate change-induced disruption.

The costs of climate change - including lower growth - will pile up the longer it takes to address it aggressively. The AI revolution could massively boost productivity — and, with that, growth, but it also could cause major disruption in labour markets, especially among middle class workers. So don’t bet on AI helping the general sense of unhappiness in the short term, but rather adding to it.

Neither does it help that refined petroleum remains the Netherlands’ main export. The country has a significant stake in the status quo. Its CO2 emissions per capita is significantly higher than France or the UK’s.

Add to that the risks to the open Netherlands economy of geopolitical competition and an ever more fragmented global economy. The Netherlands is a major hub of world trade, the 6th biggest exporter, and the 8th biggest importer. While the Netherlands is not nearly as exposed as Germany with exports to China, who won’t be able to rely on exports for growth as it has in the past, the Netherlands will suffer if the German economy does. Germany is its largest export destination.

If the Dutch turn rightward under these relatively benign conditions, where will they end up in five to ten years’ time? Will they eventually settle on no free movement of people and enforce border controls within the EU? That’s a principle which is held as sacrosanct within the EU. But it surely looks like that is where we are heading.

So you have to wonder how the Dutch will react in the future and whether the European Union will be able to withstand these pressures. If I were a betting person, I would probably say, if the relatively privileged Dutch do not withstand them — then no, the EU won’t either.

The design flaw at the heart of the euro

On the macro-level I have more of a footing for commenting. I’ve been closely watching events and reading about the structural issues of the European Union for several years. Full disclosure: I advise a major European bank on public policy. I’ve been reading about monetary and political theory about what makes societies work. All of this points to one conclusion: the EU, as it is currently constituted, must move forward to closer integration or backwards (voluntarily), or it will come undone in a drawn-out, destructive, acrimonious way.

Let’s start with the money stuff. As Joseph Stiglitz pointed out, the EU has a problem — the euro. Instead of creating a convergence between its members, the euro is pushing member states apart, and it is also one of the most significant reasons it is not growing faster. This should matter to the Dutch because as I have already pointed out, unlike Germany, the Dutch economy depends much more on trade with Europe.

The belief was that free movement of capital, enabled by the euro, would lead to increased efficiency and, consequently, prosperity. This view saw the euro as a facilitator of capital mobility. However, this overlooked a critical element: the integral role of the state as the ballast of any banking or financial system.

Yet the EU was not conceived as a state, never mind a super-state. So the Euro is effectively stateless money. This is why, when the 2008 crisis hit, which started in the US, capital did not flee the US; it streamed there from the EU. Money managers knew the might of the US state had its banks’ backs.

In Europe, this issue also manifests as follows: Given a choice, investors prefer putting money into the German banking system and state over the Spanish. The Euro and single market make this transfer easier, resulting in capital flight from Spain to Germany, particularly in times of stress. This outflow impacts Spanish banks’ ability to lend.

The ideologues behind the euro assumed that a small Spanish business, like a dry cleaner or grocery store, could easily secure loans from large German banks. However, this ignores the local nature of information and relationships in banking. The movement of capital out of Spain, therefore, led to massively reduced lending to local businesses, causing a destructive economic impact.

So the euro area established a system where a negative shock exacerbates problems, prompting capital to flow towards stronger economies and slow growth. This structure essentially means that in times of trouble, resources in the eurozone move away from those who need them most. We saw this with the Russian invasion of Ukraine, how Europe, who was in the process of stopping QE and raising rates to combat inflation, had to announce special measures to support Italy’s ability to finance its debt.

Why was this moment considered dangerous? Because it can set in train the so-called sovereign banking doom loop. The doom loop is a perilous entanglement between a nation’s fiscal health and its banking sector. For stability’s sake, banks need to hold substantial safe assets, usually government debt. Fiscal distress diminishes these bond values, eroding bank balance sheets. Weakened banks may necessitate government intervention, often as bailouts, further inflating public debt.

This escalation of government debt compounds the crisis, devaluing bonds further and intensifying banks’ vulnerabilities. This is what we observed in Greece and Italy. The process can also start with banks, where healthy government balance sheets can be thrown in the red, when saving their banks, as happened in Ireland and Spain.

This scene still haunts Europe today since it still has no banking union like all successful states do. It impacts weaker states’ growth and eventually the Netherlands’ own.

First and foremost, the eurozone needs shared deposit insurance like the US does. As Stiglitz explains, if Washington State had been responsible for bailing out Washington Mutual, one of the US’s largest banks, it would have faced significant difficulties. The state’s moderate size and resources would have made such a bailout a formidable problem. However, the rescue did not fall on the state but was handled by the Federal Deposit Insurance Corporation, a federal entity.

Yes, having a banking union would imply that taxpayers in the Netherlands might be needed to help prop up a weaker state, but that is the price one must pay for all the benefits the union has bestowed on the Netherlands, but more importantly, for a flourishing EU economy from which it will benefit. That growth, cheap capital and frictionless trade with the rest of Europe are not the only benefits.

The Netherlands and Germany in particular, have benefited from the euro being a cheaper currency, making their exports more competitive globally. If they had their own currency instead, it would appreciate immediately and reduce exports.

So they should thank Italy, Spain, Portugal and Greece for sharing a currency. If the EU fails over immigration, the euro would also be a goner. I don’t think the export-orientated countries in the north have communicated what this would mean to their voters. But even if they did, I still don’t think it will be enough to change minds. And here is why.

The trilemma of integration — (or its identity stupid)

The fundamental problem for the euro and the EU lies in the realm not of the economy but in identity. It is the most critical issue that the market-orientated designers of the euro never anticipated.

Harvard economist Dani Rodrik has posited a trilemma on economic integration: To have economic integration between different countries, one must sacrifice either democracy or sovereignty.

In other words, to maintain democracy, one must vote within that new larger body, such as the EU. This means that the Dutch citizens would vote alongside the Germans, the French, the Italians, thereby relinquishing national sovereignty and vetos. Their national voice would be subsumed by citizens’ individual votes in a larger polity. The Dutch would become Europeans first and foremost like Americans are.

If, instead, one wants to integrate but preserve national sovereignty, one must give up on democracy. This is what the EU has become. You have a situation where citizens in member states want certain things done, and even vote for them, but they are not implemented because of the contradictions inherent between international economic integration and national democracy.

For example: A year before Wilders launched his party, in 2005, first the French and then the Netherlands resoundingly rejected the newly proposed European constitution. Despite these setbacks, EU leaders and subsequent EU Presidency holders were keen to keep the constitutional project alive.

This commitment led to the negotiation of the Treaty of Lisbon, which retained many elements of the Constitution but in a different format and under a new name that was not put to voters. The Treaty of Lisbon was signed in 2007 and came into effect in 2009.

So Wilders has a point about the EU being anti-democratic. There is, of course, another way around this problem. That is to opt for democracy and national sovereignty and give up on the European Union altogether. That will be worse for the Netherlands than the PVV voters imagine but better than the slow train wreck that is the current unhappy marriage.

A citizen of somewhere

Mon, 14 Aug 2023 00:00:00 GMT

I am probably the kind of person Theresa May referred to when she talked about “the citizens of nowhere” in a speech. “If you believe you’re a citizen of the world, you’re a citizen of nowhere. You don’t understand what the very word ‘citizenship’ means”, she said.

After more than two decades in London, I find myself thinking about that here in The Hague. I’d like to think I did not leave London because I am tired of life. On the contrary, I left (for a while at least) because it’s been bugging me for some time.: Where would I like to spend the second half of my life?

And some of the chief considerations are to be in a place where you feel part of the social fabric of a place, to contribute outside just doing your job and feel that you belong. The irony is that, on some level, I felt like I belonged in London for a long time. And many Brits living outside of London, experience it as a strange and foreign presence. It is a city for citizens of nowhere.

So why the dissatisfaction? It’s best to explain that by way of a contrast with what I have experienced about the Dutch so far. The Dutch, contrary to my expectations of rudeness, have been super friendly and helpful. Perhaps I confused directness and reluctance to sugarcoat opinions with being rude.

But London is not unfriendly either. It is much more friendly than Paris. But here, the confusion of two concepts is again at play. London just cares less. London’s famous tolerance is to a considerable extent — indifference. That makes feeling you belong in London a bit cheap, and not altogether that significant.

When I was younger, I loved browsing my sister’s Cosmopolitan magazine. I found all the beautiful women and liberated talk about sex and pleasure intriguing. But one word of Cosmopolitan wisdom struck a chord. An article about friendship boldly proclaimed: “In the future, friends will be the new family”.

Now that I am older, I still think friends can be like family. But I am less inclined to think that they can easily replace it. Last night I saw Jan Karel, an old friend for the first time in years. Two thoughts struck me. One is that friendships are like language. They are much easier to forge into something deep and meaningful when you acquire them in your youth.

Jan’s description of the Netherlands helped me understand some of the irony and nuance of the Dutch I had observed but not articulated yet. The Netherlands didn’t really have a mask mandate during Covid. It had some of the lightest lockdowns, and yet there were big protests against the measures imposed because the indignant said it was an imposition on the cherished Dutch value of personal choice.

The Netherlands is famously the place where prostitution and drug use are tolerated, underlining the point. However, in many ways the Netherlands is also a far more controlling society than Britain. Here, as a new arrival, many things cannot be done, including getting a mobile number and a Dutch bank account, without a citizen’s identity number. The UK, like most anglophone countries, does not even have such a thing as an identity number.

Without a Dutch mobile, most Dutch online services cannot be registered for, including the Internet. And without a Dutch bank account you can’t pay at most shops, because the Dutch have their own payments network, IDEAL, and having a card that uses Mastercard and Visa does not work.

Here home-schooling is verboten, and you cannot teach your own kid to swim at the local pool. They must do a government-initiated swimming program and get accreditation before they can access pools. Like in the UK, you cannot simply take your kids out of school without permission. But here, the difference is huge in monitoring and enforcement. If you did and tried to fly to South Africa for a holiday, the immigration officials at the airport would ask you for proof that the school gave permission.

You are left with a pervasive sense that the Dutch state and society are keeping an eye on your affairs. A couple of days after we arrived, we received a “Buurt Kadoos”, a welcoming box with information about and discounts from businesses in the area. Who told them we were new to the area and where we lived?

A citizen’s ID registration required consultation between two officials. Emiel has Ruth’s surname as a second name. In the Netherlands, surnames are protected, and this is not allowed. An official provided the verdict. Since the UK allows this, they would have to make an exception for us.

One other notable thing is how clean the streets are here compared to London. And if you litter, you will get told off. To properly dispose of rubbish, it must be separated in four ways and deposited in the appropriate bins located in the neighbourhood. Between 7 am and 10 pm that is, and not if the bin is already full, and if you don’t follow those rules menacing messages on the bins inform you that you will be punished severely.

Jan says that the Dutch don’t like when you point out how controlling a society it seems to be to outsiders. It offends their sensibility of how they value personal autonomy.

Anglophones may never grasp something important about all of these things, even if they are a pain or a sign of a “nanny state”. They are also things that make you feel like you are part of a society - and one that cares. A bit like parents.

This is a society where every first of the month, alarms are sounded across the Netherlands for testing. Alarms that warn citizens that a dyke has been breached. In such a place, there must always be a sense that you are a part of something bigger.

We did not come here because we thought this was an end destination or expected to belong. It was more out of a sense of adventure and a feeling that it would be good to get off the British Isles for a while. Perhaps a good thing then, because Jan adds, the Dutch will never truly accept you as one of their own if you are not born here.

Adam Tooze on the schizophrenic South African experience

Wed, 12 Apr 2023 00:00:00 GMT

I’m an out-of-the-closet Tooze-bro. So imagine my excitement when South Africa’s economy was the subject of his latest Foreign Policy podcast. It did not disappoint, although I felt Tooze held back a bit. Until the end, when his gloomy and, in my opinion, quite correct assessment landed.

Here is an excerpt from that end, with Tooze talking about Eskom, South Africa’s giant state energy company:

* That has been an absolute crisis of corporate governance with the CEO, Andre de Ruyter. Essentially resigning from his position on the threats from within the organization, that there’s credible reason to believe that there was a poisoning attempt, that an assassination attempt was made against him, by the mafia groups that operate within Eskom. So there were powerful, organized crime syndicates connected to the trade union and the ANC, operating in the major coal regions of South Africa who see themselves as having a vested interest in this, in this malfunctioning giant.*

This goes from, you know, large-scale resistance to green energy as an option for the future, down to criminal gangs that systematically rob coal trains, replace the coal, the good coal with bad coal, or even scrap metal, which then gets fed into the furnaces, which then sabotages and blows up the power stations. The fear here, and it’s a very real terror, is of what would happen in a society as unequal and as violent a society, with as much simmering resentment as there is in South Africa, if there were to be a prolonged and serious power shortage?

And we have seen rioting in recent years, which was taken on the scale of civil war conditions. And while I was there, the insurance companies in South Africa actually publicly declared that we’re there to be a power outage, they were not committing themselves in advance to honoring insurance contracts on private property in the event of large-scale civil disturbance.

So that is the kind of level at which the conversation is carried on in South Africa, all of this within a society, which unless you look below the surface, You might not think was really shaking. I mean, it’s, it’s a remarkably schizophrenic experience being there because it’s a highly developed society with great internet.

And if you live in the more prosperous parts, You know, you have no sense at all of being on the edge of a precipice or, you know, on the edge of a volcano. But the fear, that haunts, South African society — of total breakdown is, is traumatic, it really is.

It is this breakdown of a modern state, albeit a very particular one, that partly fuels my fascination with states. That, and people in the West taking the states they live in for granted. This is partly the reason I started this newsletter.

Adam Tooze’s podcast, called Ones and Tooze, is where he uses a few data points as a way to illuminate a subject. But like most commentators on South Africa, even someone with Tooze’s analytical ability seems best at describing the tragedy, rather than explaining how we got there.

Some counter-intuitive graphs on South Africa

Here is a set of numbers, graphs, in fact, that, in my opinion, should be included for a discussion on South Africa because they are so counter-intuitive to existing narratives. And I think unpicking and explaining these graphs will go some way (though not all) to burn away the mist that envelops the countries’ sad trajectory.

All these graphs are from the World Incomes Database, in whose creation Thomas Piketty had a hand. The first shows the share of national income that went to South Africa’s top 1% of earners. During WW2, this bracket earned its most significant share of national wealth. From then to 1993, their share fell. But it’s right back up now.

Caption: South Africa income inequality: Share of national income that went to the top 1% of earners.

The second shows the share of national income that went to the top 10%. The top 10% of earners also took the lowest share of national wealth in 1993. Today they are as high as ever.

Caption: South Africa income inequality: Share of national income that went to the top 10% of earners.

Caption: South Africa wealth inequality: Top 0.1% and top 10% share.

The last, which only starts in 1995, unfortunately, shows the share of the wealth that went to the top 0.1% and bottom 90%. The 0.1% bracket’s wealth has only increased in the last three decades. The bottom 90% have about the same share of wealth as they had in 1995, after a low round 2008.

High-status social media users have a license to misinform

Mon, 06 Mar 2023 00:00:00 GMT

Plato, like Aristotle, defined rhetoric as the art of persuasion. But unlike Aristotle, Plato had a negative view of it. In his dialogues “Gorgias” and “Phaedrus,” he argues that rhetoric is merely a means of convincing people, and does not provide genuine knowledge or truth.

In other words, he was not arguing that rhetoric is ineffective. On the contrary, Plato saw rhetoric as a powerful tool that could be used to manipulate and deceive people. In contrast, he believed that philosophy was supposed to uncover truth and provide genuine knowledge.

Several studies have shown that aggressive or untrue messages tend to reach a wider audience on social media. A fascinating recent study indicates that low-status users are penalised for sharing misinformation. But as users move up the status hierarchy — of which follower counts are a good proxy — sharing low-quality information grows their status even faster. The study goes on to argue that these dynamics produce a feedback loop:

“…in which those who have already achieved some stature further distance themselves from others through quality-ambiguous actions.”

This presents a moral dilemma for those that use social media as a platform — and poses a seductive but real risk to leaders using it.

Chatbots are data driven manipulators - or our personal daemons? (Links 5)

Thu, 23 Feb 2023 00:00:00 GMT

This week’s newsletter: Things are getting weirder thanks to Microsoft’s new AI chatbot, but will AI make us more productive and more powerful? However, if we thought that data and digital ads manipulate human behaviour, what would chatbots like Sidney accomplish? Is social media becoming paid for? And — recommender algorithms in the dock.

Rapidly, things are getting weirder.

A few early testers have been given access to Microsoft’s cross between OpenAI’s ChatGPT and its Bing search engine. Able to surf the web, it was clearly far more potent than ChatGPT, as in this example, where Evan Mollick instructed it to write something using the writing advice of Kurt Vonnegut eloquently demonstrated. Or see this example brand people; or this academics. Astounding!

But it got weird too, as this interaction of the New York Times’s Kevin Roose showed - the bot, revealed its secret real name - Sydney, and seemed to fall obsessively in love with Roose.

The Art of data powered Chatbot persuasion and manipulation

Regarding what these developments may portend, L M Sacasas wrote an intriguing piece about the ELIZA chatbot of 1966. ELIZA relied on the techniques of Rogerian psychotherapy in which the therapist echoed the patient’s statements.

ELIZA’s creator, MIT computer scientist Weizenbaum was shocked at people’s reaction to his bot. Even when they knew it was a chatbot and how it had been made, and his disclaimers that it was not a real therapist, they shared intimate details with it.

It prompts Sacasas to wonder whether: “People found it difficult to resist the illusion that they were being heard and attended to with a measure of care and interest?”

Sacasas fears that many lonely people in our society will become attached to these tools and worries about those with mental health problems engaging them. What he thinks may happen in the realm of prediction, persuasion and manipulation is as important.

Like me, Sacasas has been sceptical about the grand claims made for digital disinformation campaigns waged via bots and advertising. Just this week, I reread Evgeny Morozov’s takedown of Zuboff’s The Age of Surveillance Capitalism. Much of his critique still holds, in particular, how confused and imprecise Zuboff’s theorising is.

But her broader point perhaps comes into sharper focus as Sacasas writes:

Another sobering possibility arises from the observation of two trajectories that will almost certainly intersect. The first is the emergence of chatbots which are more likely to convince a user that they are interacting with another human being. The second is the longstanding drive to collect and analyze data with a view to predicting, influencing, and conditioning human behavior.

…there seems to be a world of difference between a targeted ad or “flood the zone” misinformation on the one hand, and, on the other, a chatbot trained on your profile and capable of addressing you directly while harnessing a far fuller range of the persuasive powers inherent in human language.

“a chatbot trained on your profile and capable of addressing you directly while harnessing a far fuller range of the persuasive powers inherent in human language.”

Your daemon, your muse

Related: Yes, shit is getting weird. But that’s a small price to pay for what we will soon all have writes Henry Oliver — our own daemon (Tyler Cowen’s phrasing):

*One good analogy for the way these bots work is as a familiar or a daemon. There is something deeply faustian about them. You have to accept their inaccuracies in order to unleash their benefits. *

Magic has not been a part of everyday life since the seventeenth century, so it’s no wonder we are a little out of practice. But the basic idea is familiar. Some people will do remarkable things by using this strange magic-like technology that many others simply don’t understand. Inevitably, therefore, much of the reaction will be dismissive, critical, or hostile. Roose reported on Sydney without quite realising, I think, the extent to which he created her hallucinations.

So the hallucination is the point.

You can now have a daemon, should you so want, who can inspire and incite you. It can be your Socrates, your friend, your sparring partner, your tutor, your muse.

Facebook announced plans this week to introduce a premium Verification Tier, sparking fears that this is yet another blow to digital advertising. Paid users will get more visibility in search, comments and recommendations.

Facebook’s offering differs from Twitter’s in that it’s only a test in New Zealand and Australia. Verification will be matched against government IDs, not the slapdash Twitter approach of payment methods.

I’m sceptical, as pointed out last week, take up of the Twitter scheme has been super low. But if successful, these schemes may impact advertising.

Verification of social media users is however long overdue. Banks have to abide by Know Your Customer Rules (KYC) yet social media companies have been allowed to get away with providing anonymous users publishing platforms.

Of course, suppose all social media users were required to identify themselves via a government ID. In that case, it is arguable that some of these negative effects that the Gonzalez case (see below) seeks to rectify would be achieved, as third parties would bear a greater risk of being caught out or blocked from publishing harmful content using internet platforms like YouTube.

A big potential deal: Section 230 immunity for recommender algorithms under threat in Gonzalez vs Google.

It has not gotten that much coverage, but an extremely important case is before the US supreme court, namely whether Section 230(c)(1) of the Communications Decency Act, which states that:

“No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.”

…still applies (and therefore immunizes online platform operators) — when their websites make targeted recommendations of user-provided content.

So the question turns to recommender algorithms.

If the court were to agree with the Plaintiff in this case, it could at a stroke make algorithmically created content timelines, which are the default setting on Twitter, Facebook, and the only setting on Instagram, Tiktok and the to-be-launched Artefact — a too-risky business model.

This is also not as straightforward a case as it seems, for example, take the concept of “amplification of content” as this excellent post makes clear.

For a more in-depth dive into the case, listen to this podcast on the case by Tech Policy Press.

Related: Jonathan Astray has published a paper setting forth a methodology whereby editorial values could be embedded in news recommender algorithms.

Tangentially Related: Alice Evans has a fascinating post about Sexual Competition, Social Media & Algorithms.

More annoying, and more of it - online advertising in the future (links 4)

Wed, 15 Feb 2023 00:00:00 GMT

In this week’s newsletter: the campaign for privacy will have unintended consequences, more and more annoying ads. The big active state is making a comeback, partly because of changes in the intellectual climate, and we are in danger of underestimating AI.

Caption: Midjourney: minimal illustration of a five vintage advertising posters

You will begin to see more annoying ads as a result of Apple’s App Tracking Transparency (ATT). And according to The New York Times, anecdotally at least, that is what is happening.

Other less obvious unintended consequences have been discussed by me a few times: how ATT will lead to online consolidation, for example.

Yet the New York Times still fails to mention this: targeted ads not only mean more relevant ads but also mean fewer ads and lower costs for the advertiser.

Additionally, targeted ads encourage and allow smaller ad campaigns, which is advantageous to startups and small businesses, especially for niche products. Fast-moving consumer goods, like toothpaste and petrol, tend to have more undifferentiated consumers. So the large companies that provide these products don’t care as much about targeting.

All this is because of how ad platforms work (simplifying here for the sake of brevity). There cannot be too many ads on platforms, or users will stop using them. As a result, they cap the frequency per user over a period of time, which means there is a limited amount of ad inventory available.

Platforms display ads until an advertiser’s goal is reached (clicks or conversions). The price is determined by how many ads need to be shown until this happens. With very targeted ads, few need to be shown.

Due to Apple’s ATT and soon-to-be legal restrictions in Europe, targeting is now more difficult, so expect not only fewer relevant ads, but also more. You can also expect costs to rise, particularly for small businesses, charities, and start-ups.

But what about the scourge of “surveillance” capitalism? In my opinion, the moral panic over tracking, ads, and privacy is misplaced, and the issue of privacy can be effectively addressed with the right regulation making a distinction between what amounts to three separate issues: security, privacy and nuisance.

As I predicted - Elon Musk’s plans to fund Twitter via premium payments instead of ads is not working out.

(Musk’s Twitter Has Just 180,000 U.S. Subscribers.)

And… As I predicted - Mastodon will have trouble replacing Twitter.

(Mastodon’s daily active monthly user count dropped to 1.4 million by late January from 2.5 million in December.)

An active state: The intellectual tide has turned

There are times when the intellectual ground beneath our feet shifts dramatically without much fuss. The idea of governments doing industrial policy is hot right now, as I mentioned in my last newsletter. Until very recently, that was anathema to much of western thinking. In its state aid rules, for example, the EU has prohibitions against this written in stone, which are now being overridden by “temporary” measures.

In this vein, Martin Wolf sent tongues wagging with his answer to a question at a talk this week that he sounded “Marxist”. Wolf is the chief economics writer for the Financial Times, a position he has held for decades. A highly respected economist, and with meticulously argued essays, he criticized the Tory government’s austerity policies, as well as much of Europe’s response to the financial crisis. Wolf argued - correctly, as it turned out - that cutting government spending when businesses are cut spending too would not reduce government debt or lead to robust growth, but the opposite.

Wolf has a book out, which he claims will be his last. He started writing it after Trump came to power and Brexit happened. The reasons for these things happening, he says, are long-term - the erosion of welfare states since the 1970s - and more recent - the financial crisis. Despite its obviousness, I have yet to see a writer explicitly link the rise of populism and the financial crisis.

The book is about his concern about “democratic capitalism”. A system he believes in but which he feels is now under profound threat. It is a system that he feels is particularly vulnerable to unscrupulous populists without significant growth and lower levels of inequality in our societies. A measure of how the ground has shifted is that The Financial Times’s editorial board famously supported austerity in the early 2010s, and Wolf was a lone voice. But now they agree with him. And even the mouthpiece of free trade and unfettered markets, the Economist, their Europe editor, tweeted this, this week about Germany’s rail network grinding to a halt:

The UK may not have an industrial strategy, but it is planning on launching something akin to the US’s DARPA, the government research agency that launched the internet. It’s called Aria.

The rampant “statism” that we now see in China (which has been a primary theme of Chinese political economy since 1912) is a direct reaction to the perceived failures of the Qing dynasty, which undertaxed the peasants and was to weak to compete with other states in the modern world.

Underhyping AI

A new study warns of a tendency by AI researchers to understate successes and warns that this may make it harder to mitigate harm. Just this week, one example.

The New Yorker ran an article to demystify AI, using a metaphor to do so. Large language models (LLMs) work like blurry JPEGs it was said, implying they do memorizing and rephrasing what they have learned.

This thread counters that, arguing that existing science already shows LLMs are more powerful than that: LLMs can learn generalizable strategies that perform well on truly novel test examples on which they were not trained.

Not a cheap worker replacement. AI use will cost a lot more than we think?

Aristotle on rhetoric — the art of persuasion

Fri, 10 Feb 2023 00:00:00 GMT

Rhetoric refers to the art of persuasion, the ability to use language effectively to influence others. It’s not just about what you say, but also how you say it and the context in which you say it. Throughout history, philosophers have studied rhetoric and have developed various theories about how it works and how it can be used to achieve specific goals.

One of the most famous philosophical theories of rhetoric is Aristotle’s. According to Aristotle, rhetoric is a means of discovering the available means of persuasion in any given situation. He believed that there are three primary modes of persuasion: ethos, logos, and pathos. Ethos refers to the credibility of the speaker, logos refers to the use of logical argument, and pathos refers to the emotional appeal of the message.

Another philosophical theory of rhetoric is the classical theory, which was developed in ancient Greece and Rome. This theory is focused on the use of formal language, logic, and structure to create persuasive arguments. The classical theory of rhetoric is still widely used today, particularly in legal and political contexts where the goal is to present a clear and convincing argument.

Now, let’s consider how these philosophical theories about rhetoric can be applied to everyday professional life.

Delivering a motivational message to employees: If a business leader wants to motivate their employees to work towards a common goal, they could use Aristotle’s theory of rhetoric to craft a speech / post / video that resonates with their audience. They could appeal to the employees’ sense of pride and commitment to the company (pathos), present data that shows how their efforts will contribute to the company’s success (logos), and highlight their own experience and track record as a leader (ethos).
Building brand reputation: If a business leader wants to build their company’s reputation and influence online, they could use Aristotle’s theory of rhetoric to craft a social media strategy and tactics. They could use their personal or company account to share inspiring stories and content that appeals to the emotions of their followers (pathos), use data and statistics to demonstrate the value and quality of their products or services (logos), and highlight their experience and track record of success as a leader in their industry (ethos).

In each of these examples, the business leader can apply the principles of philosophical theories of rhetoric to communicate more effectively and persuasively. By understanding the different modes of persuasion and the principles of classical rhetoric, they can improve their ability to influence others and achieve their goals.

Sometimes however, direct attempts at persuasion are not the best approach to social media. More on that another time.

Create your own chances, Social CEO

Thu, 09 Feb 2023 00:00:00 GMT

Let me tell you a story about a great leader who once said, “The best way to predict your future is to create it”. Leaders don’t wait for opportunities to come to them, they go out and create their own opportunities. They have a clear vision of what they want to achieve and they work tirelessly to make that vision a reality.

But, more often than not leaders need help to make things happen. How do they inspire those around them to follow their lead and work towards a common goal? The answer is simple: by articulating their vision in a way that resonates with their audience.

Articulating your vision is not just about speaking words, it’s about painting a picture in people’s minds that motivates and inspires them to take action. When you clearly articulate your vision, it becomes contagious, and people will naturally want to follow your lead. They will believe in what you are trying to achieve and be motivated to work towards it alongside you.

And in today’s world, there is no better tool for building your influence and articulating your vision than social media. Social media gives you a platform to reach a massive audience, and it’s an incredibly powerful tool for spreading your message and building your brand.

Remember, a vision without action is just a dream.

So my challenge to you today is to think about your vision, to articulate it in a way that resonates with your audience, and to use social media as a tool to build your influence and bring your vision to life.

We can't escape AI, but we can shape its future (links 3)

Tue, 07 Feb 2023 00:00:00 GMT

In this week’s newsletter: Should we fear AI? Can we opt out? Will it take our jobs?; Industrial strategy is in vogue - from right to left, except in the UK; And the Titok of text.

Please forward this email if you enjoy it.

An article in the Atlantic by Reid Hoffman, co-founder of LinkedIn, about whether Chat GPT’s arrival is good for humanity — and a tweet by writer LM Sacasas piqued my interest.

Caption: Stable Difusion: A female cyborg walking confidently, positive, background a beautiful landscape, Gustav Klimt, highly detailed painting, organic surreal shapes, masterpiece, great composition, intricate detail, very detailed, hypermaximalist, JS03 –ar 3:2

Sacasas quoted Austrian Ivan Illich, a catholic priest, public intellectual and noted critic of industrial society, its institutions and technology. With the rise of social media, many intellectuals have dusted off Ivan Illich. And you can expect more reaching for him as the age of AI dawns.

This particular quote can be read as both a warning on social media and large language models (LLMs), like GPT3, which has been trained on the text of the World Wide Web.

In Illich’s view, economic development often does not result in human flourishing but rather ‘modernized poverty,’ dependency, and an out-of-control system in which humans are not the masters of technology but have become reduced to worn-out mechanical parts.

Illich proposed that we should “invert the present deep structure of tools” in order to give people alternatives that guarantee our right to work independently and be in control of our lives.

Opportunities for personal fulfilment are as large as the universe itself

That Reid Hoffman’s takes the opposite view should surprise no one. Hoffman is an investor in OpenAI, the entity behind the bot in the limelight, Chat GPT. But he is unusually thoughtful for a billionaire tech entrepreneur1.

Hoffman does warn of the importance of enlightened and responsible leadership by tech founders, but none the less sees in tools like Chat GPT a future -

“in which opportunities for personal fulfilment are as large as the universe itself.”

“If we merely lived up to our scientific classification—Homo sapiens—and just sat around thinking all day, we’d be much different creatures than we actually are. A more accurate name for us is Homo techne: humans as toolmakers and tool users. The story of humanity is the story of technology.

Technology is the thing that makes us us. Through the tools we create, we become neither less human nor superhuman, nor post-human. We become more human.“

We are Cyborgs

Personally, I’m on team Donna Haroway. She wrote the Cyborg Manifesto in 1985 to make the following points described colourfully in this blog post by sociologist PJ Patella-Rey.

Haroway agrees with Hoffman that humans and technology are inseparable. What’s more, she thinks the impact of technology on our lives can not simply be opted out of.

As much as those who refused to drive motor vehicles were still impacted by the massive change in how our cities looked and worked due to the advance of cars, we cannot escape our socio-technical circumstances.

So we can opt out of using social media and AI, but it will still impact our lives, for example, by helping determine who gets elected. Unlike Illich, Haroway recognises no ideal past, but unlike Hoffman, she does not think technology will necessarily lead us to a brighter future. Technology isn’t good or bad, but it is a place where politics happens.

And that’s important because its politics that determines technology’s impact on our lives: Good or bad.

And its politics that determines technology’s impact on our lives…

Will AI take our jobs?

Hoffman mentions it as a concern - what AI could do to jobs. The Gradient podcast has a great interview with Steve Miller, a Professor Emeritus of Information Systems at Singapore Management University. Steve works closely with a number of Singapore government ministries and agencies via steering committees, advisory boards, and advisory appointments.

In this long in-depth discussion, Miller makes a number of important practical points. How different is this “industrial revolution” to those that went before? Ultimately Miller sounds like Haroway.

Whether AI ends up immiserating or empowering most people depends on how we decide to apply it.

Industrial policy is so hot right now.

That’s a headline from the FT, and it’s bang on. Industrial policy is in vogue - and most economists consider it to be positive for global economic growth. It seems that everybody agrees, from the right (yes even Marco Rubio and Josh Hawley) to the left, even the centrists in the EU.

Here is a podcast featuring Adam Tooze on how Biden has killed off the influence of the World Trade Organisation. And here is Bloomberg on how Biden’s misnamed Inflation Reduction Act is forcing Europe to react - a race to the top for clean tech.

Actually, not everyone agrees. The UK, whose Kwasi Kwarteng recently ditched its industrial strategy because it was ultimately ideologically incompatible with the then-ascendant libertarian wing of the Tories - are now very worried about being left in the dust by the US, Europe and China.

But isn’t this a return to nationalist protectionism and even mercantilism? Tyler Cowan, who these days call himself a state capacity libertarian, argues it’s not. Rather it’s competition between states:

Even the most successful “nationalistic” industrial policies rely on a highly globalized world. If carried out strictly on a one-nation basis, industrial policy is doomed to fail. Globalization has been so thorough, and has gone so well, that at least a little industrial policy is now thinkable for many nations.

The Tiktok of text

The co-founder of Instagram is launching a competitor to Twitter, Artefact, that will, like Tiktok, not depend on social networks but be completely algorithmically curated via machine learning.

He reckons it will be more meritocratic, as your status as determined by your social network will matter less.

Hoffman has a masters in philosophy from Oxford University. Do listen to this fantastic interview with Tyler Cowen to get a sense of his wide-ranging perspectives. Did I mention Cowen’s podcasts before? If I had to take one podcast to a desert island, his would be it.

The wider implications of the case against Google (links 2)

Tue, 31 Jan 2023 00:00:00 GMT

In this week’s newsletter: the antitrust case against Google and what it means for the digital ecosystem; The importance of cross-disciplinary thinking — to make sense of AI; How state-funded research is superior; More evidence that it’s not the algorithms but us poisoning the information sphere - and privacy - it isn’t very easy.

Caption: Via Midjourney

The State vs Google

The US Department of Justice antitrust complaint against Google is arcane and technical. Still, quite an eye-opener, it lists a range of blatant practices that, on the face of it, puts Google’s old motto of “Don’t do evil” well to rest.

See, for example, here and this tweet by Ari Paparo.

How to deal with Big Tech and the power they wield?

When new FTC chair Lina Khan was appointed, much was made of her new approach in terms of antitrust that would not focus on economics’ orthodox consumer welfare standard as measured by monopoly prices. The reason for her approach was simple enough: most of the “Big Tech” services are free to use for consumers.

Instead, the focus with Khan would be on something more abstract: a concentration of power. This idea is not wholly alien, and it has been the philosophy behind regulation to ensure media plurality in the past.

Caption: Graphic: Bloomberg

But this case against Google is old school. Specifically, the US Sherman Act’s requirements for muscling out the competition to raise prices or offering substandard service (in this case, advertisers paid more while also significantly lowering revenue for publishers) are relevant.

This makes Google vulnerable even when its market share in digital advertising is at a years-long low: Amazon has joined Google and Facebook to become the third major player (See below graph).

Caption: Google is a player with a declining share in a triopoly that is taking an ever bigger share of digital ad revenues.

And this is so even when we are only talking about the so-called “Open Display” part of the advertising market on 3rd party publishing sites. Open Display, even if it caters for everyone except for the owned and operated platforms of Google, Meta and Amazon, is a far smaller slice of the digital advertising market.

And Open Display earns a far smaller subset of Google’s advertising revenue than its owned and operated Google Search, and about as much as YouTube — and only 12% of overall revenue (See graph below).

Caption: “Google Network ads” is the “Open Display part” of Google touched by this case.

As The Information points out, “while Google’s display ad business has grown, any publisher that is heavily dependent on it is no longer in business—or won’t be much longer. But that doesn’t mean this case isn’t important.”

That’s correct; it’s still important because several notable issues arise. In our era of disinformation, Google’s shortsighted actions appear to have contributed to the demise of journalism and the media in general. As print (and now broadcast advertising) shrinks, Open Display Ads, are not delivering for publishers going digital either.

All the more remarkable because Google’s mainstay search business relies on the existence of a reliable and quality open web, which Google’s actions appear to have worked against.

Another nail in the coffin of the Open Web?

And if Google is forced to disinvest in the sell side of Open Display advertising (as the DOJ complaint wants), will it cause the company to focus more on its owned and operated services? I have highlighted before how Apple’s App Tracking Transparency (ATT) is leading to consolidation online by incentivising large owned and operated advertising platforms.

And what could this mean for Google’s plans for Chrome-based Privacy Sandbox, the project that sought to replace 3rd party cookies1, which the Open Display ecosystem uses for targeted advertising?

Continues Jessical Lessen in the Information:

One thing I’m more certain of is that this isn’t the last challenge to a tech ad business we’ll see from regulators. I’m not even sure it’s the most important. Imagine the company selling ads was also directly selling products that competed with those depicted in the ads (Amazon!). Imagine a hardware company that sold ads was letting itself use far more targeting data to sell its ads than it was allowing its rivals to use (Apple!).

PS: From the DOJ case it appears Google was doing similar to Apple now with ATT, using privacy as an excuse to cut competitors’ use of data.

The importance of state-driven research innovation

A new paper is out making a fascinating claim:

Figuring out the impact of AI

We need more people with interdisciplinary skills, like Paul Meehl, to understand our AI future tweeted Jonathan Anstray this week.

Caption: Paul Mheel

Meehl was a statistician, a therapist and a philosopher of science. One of his notable contributions was a paper that argued way back in 1954 that statistical predictions would beat clinical predictions (decisions based on professional clinical judgment).

In this vain, here are two interesting papers:

First, an impressive new paper from Google: When Google tuned a chatbot to answer common medical questions, doctors judged its answers 92.6% correct, compared with 92.9% of the answers given by other doctors.

And…

It’s us, not the algorithms

Political Scientists Henry Farrell and Cosma Shalizi present a new preprint to illustrate an argument that I have made here and here (but not using the same example): how the problems of our information ecosystem might arise from simple web 2.0 technology without any algorithmic curation of content.

Their basic argument is summarised well by Karl Bergstrom over on Mastodon:

We can explain the prevalence of misinformation and polarization as a consequence of (1) disatisfaction with status quo explanations, (2) psychological desire to find plausible rationalizations for pre-existing beliefs, and (3) preferential attachment dynamics such as arise from even non-individualized search engines, citation processes, and other such systems.

Meta-void

I’ve long been sceptical about claims for the Metaverse. But I did not think much about the technical hurdles it faces.

A long and detailed post by VC Matthew Ball is packed with technical reasons why VR/AR/MR has not lived up to the hype. Ah, I hear you say. The Metaverse is not limited to those platforms and will be accessed via smartphones and laptops instead.

In this thought-provoking presentation, games designer Raph Koster - and ex-Disney / Playdom vice president, explains that the Metaverse is neither new nor unique and that its main problems are not technical — but social.

Digital Privacy is complicated

Techcrunch has a story on how US police gathered evidence from digital service providers to prosecute several abortion cases.

Privacy fundamentalists - ie almost everybody on Twitter - consider people’s right to keep their data safe, secure, and out of prying hands to be sacrosanct. Among children’s rights organizations, the digital world’s nastiness requires more protection. It is impossible to move either of these groups’ positions. That is, until both sides clash.

And a judge upholds DOJ’s broad “geofence” warrant allowing prosecutors to use Google location data to identify people who entered the Capitol on Jan. 6.

Privacy Sandbox for the Web aims to phase out third-party cookies by using the new privacy techniques, like differential privacy, k-anonymity, and on-device processing.

Recommender algorithms optimised for engagement -- it's complicated (links: 1)

Mon, 23 Jan 2023 00:00:00 GMT

From this week on, instead of regularly posting links on Twitter as I have done in the last decade, I will post the most interesting ones on Substack once a week. So without further ado…

What’s wrong with optimising for engagement?

The pursuit of engagement by online services has been blamed for many social ills, including addiction, polarization, radicalization, eating disorders, and misinformation. Engagement can be many things, including time a user spends on a piece of content, a like, a share, a click.

This comprehensive essay by researchers Luke Thorburn, Jonathan Stray, Priyanjana Bengani discusses “engagement” in relation to recommender systems, which are algorithms that personalize content in products such as social media, news, streaming services, and online shopping.

The issue is made even more pertinent as Twitter this week defaulted all of its users to its algorithmic timeline, “For You.” If you want to see the more basic reverse chronological timeline, you must now keep selecting it.

In 2016, Twitter and Instagram switched to algorithmic feeds from chronological feeds. But because of user feedback, both re-introduced the chronological timeline as an option.

What’s Right and What’s Wrong with Optimizing for Engagement refutes some easy and lazy assumptions many have that we can simply do without recommender systems optimised for engagement. For example, it’s not just commercial imperatives that are a driver of the use of recommender systems for engagement. Even non-profits like the BBC, use them.

Though chronological feeds are often touted as a panacea, a feed of all new items would be near unusable on Spotify, or Amazon - and only make sense on systems designed for following a small set of sources. On social media, reverse chronological feeds also benefit content producers that post a lot without regard to quality. So it may encourage spammy behaviour.

Proof of Polarisation

At least two papers have been published recently that cast doubt on whether social media indeed causes large-scale polarisation. Now there is a new paper by Italian researchers that have found firm proof that it does.

But not on all social media. Personalised recommender systems like the Facebook and Twitter feeds show that the human propensity for homophily drives groups of similar users towards similar news that supports their bias.

BUT not on Reddit where the system is not personalised.

https://arxiv.org/abs/2004.09603

Proof that Facebook makes us sad

Right at the start, when Facebook launched, a time when its feed was still rather dumb and reverse chronological, it was already having a measurable and significant detrimental impact on the mental health of students.

So says a new study that appears to draw the first causal connection between the use of Facebook and poor mental health. We have had a strong sense that this may the case. There has been a significant increase in metal health issues for more than a decade in several countries.

Two of the authors Luca Braghieri, an Assistant Professor in the department of Decision Sciences at Bocconi University in Italy; and Alexey Makarin, an Assistant Professor in the Applied Economics group at the MIT Sloan School of Management spoke about their groundbreaking study in this Tech Policy Press podcast. Recommended.

Podcast: https://player.captivate.fm/episode/d0b394fc-d3db-4741-b3e2-24183a9e1e3a/

Paper: https://www.aeaweb.org/articles?id=10.1257/aer.20211218

Russian troll impact overstated

According to a study conducted at New York University, Russian troll Twitter accounts had a minimal direct impact on 2016 US election voters; 70% of the exposure was attributed to only 1% of users.

https://www.washingtonpost.com/politics/2023/01/09/russian-trolls-twitter-had-little-influence-2016-voters/

The problem is who does what kind of work

A book review of historian Michael Sonenscher’s, ‘Capitalism: The Story behind the Word’, by Angus Brown, sets out how the book makes an intriguing argument: that many of the problems we associate with capitalism are actually because of the inequalities generated by the differentiation of labour, and not about who owns capital.

The result of this division of labour, either between the skilled and unskilled, between town and country, or even simply between different occupations, includes the problems of social atomisation we think capitalism produces, as well as the destruction of traditional communities, ways of life, and professions. And I guess — though it’s not stated by Brown — differences in levels of status?

These consequences are not (only) made possible by the private ownership of capital in our societies but have been with us since before capitalism — since what Adam Smith dubbed “commercial society” came into being.

Smith never mentions Capitalism in The Wealth of Nations, although he and Hegel identified these problems. In fact, capital, if used right, like through a welfare state, might be one way to lessen the negative impact of the division of labour. Lots of food for thought here as the AI revolution dawns.

https://tocqueville21.com/books/michael-sonenscher-capitalism/

Why are men so ridiculous?

A provocative question was posed and answered in a Twitter thread by sociology professor Alice Evans. Is it true that men are funnier than women? If so, why is that?

Short, nasty, and brutish

Hobbes, who this newsletter has to thank for its name, was right. Yet another study shows that in the pre-state period, human societies were far more violent.

“Of the skeletal remains of more than 2300 early farmers from 180 sites dating from around 8000 – 4000 years ago to, more than one in ten displayed weapon injuries, bioarchaeologists found.”

https://www.ed.ac.uk/news/2023/violence-was-widespread-in-early-farming-society

Ungovernable Republicans

And here is an informative podcast on how the disruptive freedom caucus is causing dysfunction in the newly Republican Party-controlled House of Representatives and what that might mean for Ukraine, debt ceilings, and squabbling with Europe over Biden’s Inflation Reduction Act.

Has the Internet really lead to increased conflict?

Fri, 06 Jan 2023 00:00:00 GMT

The patron saint of Wired Magazine1, Nicolas Negroponte, wrote in his book Being Digital in 1995 that the Internet “was about to flatten organisations, globalize society, decentralize control and help harmonize people”.

While the first three claims are at least partially true, many people, myself included, think that the last is flat-out wrong. When I posted this on Mastodon this week I was asked by Dean Eckles what the evidence was that the Internet did not in fact help harmonize people.

…the Internet “was about to flatten organisations, globalize society, decentralize control and help harmonize people”.

You could point to the increased polarisation is most western countries, to networked authoritarianism in China, to the use of social media in undermine anti-corruption efforts in South Africa, to ethnic tensions stoked on Facebook in Mayanmar / Burma, to ethic violence organised on WhatApp in India, and that’s just off the top of my head.

But Eckels is right, it does not prove causality. His excellent written testimony to US Senate points out that there is no evidence that algorithmic ranking (other than the baseline of reverse chronological) on social media has contributed to more anger and bitterness. On the contrary, he points to WhatApp, which was used to mobilise violence in India, and it only does baseline ranking.

But could we ever prove direct causal mechanisms?

On this site I have agreed that algorithms primed for engagement (to increase advertising revenue) could not take all the blame. And I have expounded before on why social media might be leading to more antagonism in societies, and will list them - and new theories - in a follow-up post. And when you go through these possible explanations, it is clear they will be hard to prove causally.

Ezra Klein thinks we do not have to prove the causality of the mechanics to recognise we have a problem. He says social media has become a “civilisational fact”. And we don’t see the positive fruits of harmony, or anything else, but lots to the contrary. And, although he concedes other factors could be at play, that’s enough for him.

I think it is worth people really reflecting on this idea that in a matter of roughly two decades, social media has gone from being barely a thing at all to something used by billions and billions of people around the world. I mean, it has become a civilizational fact faster than almost anything in human history. And something operating that macro of a scale should show some civilizational effect.

If it is good, we should be able to say, well, this is what has gotten better. GDP is growing faster because we’re sharing so many more ideas, and so innovation is sped up. Or we’re more humane and gentle and compassionate towards one another because we’re able to see each other across boundary and faction and country and generation. We’re kinder because we’re sharing so much more. We’re happier because we’re so much more connected.

Something, something should have gotten better. And I would say — and I think the evidence is very clear — nothing has. You cannot point to one macro indicator that has gotten substantially better, faster, anything, in the time since social media came on the scene. And I’m not saying that is 100 percent the fault of social media, but I am saying that it implies, at least, there is not some gigantic value here, that before it was offered to us, we were really struggling.

Yet we seem to be on the cusp of the operationalising of several scientific breakthroughs, which has lead economist Tyler Cowen and The Economist to muse that the so-called great stagnation in productivity (that he coined), may be over.

But in the big picture, Klein’s a compelling point. Not much is much better, and a lot is worse in many places right now. The Internet could yet prove to be a boon for economic growth, but so far, we do not see that in an increase in peace and harmony around us.

In a follow-up post, I will set out possible reasons why this may be so.

Though not so influential now, Wired Magazine was founded by Stuart Brand and was the heir to Brand’s important Whole Earth set of publications. It was key in its own right in setting the tone for digital culture.

Countries can not die, they stumble on

Tue, 03 Jan 2023 00:00:00 GMT

Caption: Stable Diffusion: “The silhouette of a person stumbling through mud.”

On holiday in South Africa for the past two weeks, my sister’s daughter asked me what one could do to save the country. I have a ready-made answer. I have thought a lot about this exact question. It is part of the reason I am interested in states, societies and development — and started this newsletter.

But I struggled to say it.

“It’s hopeless”, I finally managed.

In 2006, writer André P Brink predicted an apocalypse for South Africa. That same year another South African writer, Rian Malan, disagreed about such dramatic talk. Even when he agreed with Brink on South Africa’s demise, Malan did not foresee a sudden denouement like a civil war but rather slow, sad decay.

So far Malan has been proven right. Yet even civil wars don’t end countries. Just look at the Congo or Syria. Bloodied and scared zombies, they stumble on.

Unless splintered or invaded by a conquering army, countries cannot die. When they default on loans, they can lose access to foreign credit and reserves, precipitating an economic crisis. But countries cannot even go bankrupt, as many would have you believe. In ever more profound misery, countries keep chugging on.

However, in such circumstances, a country’s citizens’ suffering escalates; they live in ever-escalating fear and lose the ability to flourish. Their lives eventually become brutish and short. This reality is not evenly distributed. First, it affects the poor. Eventually, it threatens everyone except the extremely wealthy.

South Africa is living through this process. So far, slowly, though some signs indicate the decay is accelerating.

The crisis in South Africa is cast as a battle over left versus right-wing ideologies. While South Africa certainly can benefit from some market liberalisation in specific sectors, the real danger is the lack of an effective and strong state which guarantees functioning markets — and much besides.

Yet mine is not a left diagnosis; it’s Weberian. The South African problem lies in the realm of culture more than policy. And although culture offers some hope, it is not as fixed a category as race, culture is tremendously hard to change. And rent-seeking parasitic elites have no interest in addressing the problem.

There may be one, but I can’t see a likely alternative pathway for the country to stop this tragedy from trudging on.

Do ads make for angry online spaces?

Thu, 01 Dec 2022 00:00:00 GMT

Caption: Midjourney: People in front of laptops shouting at the screen.

On Mastodon, I’ve found a fascinating blog post on why Mastodon moderation will indeed work. I want to focus on this particular claim:

What I missed about Mastodon was its very different culture. Ad-driven social media platforms are willing to tolerate monumental volumes of abusive users. They’ve discovered the same thing the Mainstream Media did: negative emotions grip people’s attention harder than positive ones. Hate and fear drives engagement, and engagement drives ad impressions.

Mastodon is not an ad-driven platform. There is absolutely zero incentives to let awful people run amok in the name of engagement. The goal of Mastodon is to build a friendly collection of communities, not an attention leeching hate mill. As a result, most Mastodon instance operators have come to a consensus that hate speech shouldn’t be allowed.

Yes. Research (see here and here) indicates that aggressive and abusive behaviour drives engagement and, therefore, could drive revenue for ad-driven social media.

However, my priors are that the idea that ad-free environments will be abuse free is partly wrong because the issues at stake are more complex.

Here are three reasons why - backed by research and theory - I think it’s not that simple.

1. Humans are gonna human

Aggressive, bold statements are hardwired into the human psyche and societies. This is where the ideas of Walter Ong and oral psychodynamics come in. Before writing technologies existed, Ong argued, societies needed a means of transmitting ideas. For ideas to spread, those conveying a message needed to leave an impression to be remembered. Due to this, speech tended to be repetitive, aggressive and lacking in nuance and depth.

Before writing technologies existed, Ong argued, societies needed a means of transmitting ideas. For ideas to spread, those conveying a message needed to leave an impression to be remembered. Due to this, speech tended to be repetitive, aggressive and lacking in nuance and depth.

Today several theorists argue that social media is more similar to oral psychodynamics than literate or written ones.

2. Scale, homophily & the tech aristocracy

When I joined Twitter back in 2009, I remember how friendly a space it was. How one often got polite replies. Not unlike Mastodon feels to me today. What could explain this?

Scale, I think. Online back then was dominated by a tech aristocracy. Early adopters broadly shared the same level of education and values. As Twitter became easier to use and online access broadened, it became more diverse. This is a breeding ground for problems like context collapse.

In that sense, UX changes to make Mastodon easier to use that increase its user base would also incentivise more friction.

3. Status

Status is another primary driver of behaviour on and offline and may point to why Mastodon feels like it has a different culture now, but this could change. Follower counts are the marker of status in social networks. This study argues that the actions for which actors receive recognition vary as they move up the status hierarchy.

…the actions for which actors receive recognition vary as they move up the status hierarchy.

When social media users have low status, they have to grow their following counts by sharing information that can be evaluated for quality (truth, usefulness). Once they have status, this restriction no longer counts, and that is where the dynamic of being more assertive and aggressive can help them grow their theory followers even quicker. Being aggressive is far less work than writing a well-argued blog post.

This is not out of kilter with what other studies have found with respect to the advantages of high status1.

UX & culture for peace

One way to disincentivise this is to hide follower counts (Mastodon’s is prominent) or even to limit the number of followers a user could have.

If Mastodon’s culture is influenced by something, I would argue it’s also because of the culture, laws and norms of the country its founder and benevolent dictator finds himself in — Germany. The German legal system and constitution do not recognise speech as this kind of uber right that trumps all others — as the US does. And that’s a good thing.

PS: Moderation on Mastodon is already going wrong. See here and here.

Those held in high esteem enjoy numerous advantages. High-status individuals attract more attention (Simcoe and Waguespack 2011, Bowers and Prato 2018, Reschke et al. 2018), receive outsized credit for their contributions (Kim and King 2014, Waguespack and Salomon 2015), and can more readily access a variety of resources (Merton 1968, Bol et al. 2018). High-status firms can negotiate better terms from buyers and suppliers (Benjamin and Podolny 1999, Hsu 2004, Nanda et al. 2020), receive favorable treatment from authorities (McDonnell and King 2018), and can hire more able employees without offering higher salaries (Bidwell et al. 2015, Tan and Rider 2017).

Mastodon will have to change if it were to replace Twitter

Tue, 22 Nov 2022 00:00:00 GMT

Caption: Stable Diffusion: Mammoth in a city

Three weeks ago, when thousands of Twitter users set up Mastodon accounts, mastodon.social, the instance1 of Eugen Rochco, the founder and benevolent dictator of Mastodon, slowed to a crawl. mastodon.social also happens to be the largest single instance of Mastodon in terms of user numbers.

Conceived by Eugen, Mastodon holds some of the promise of the early web. It is part of the “fediverse”, which is a network of federated (i.e. seperate but connected) instances that provide web publishing functionality and file hosting, but that also communicate with each other — while being independently hosted. So Mastodon is to a large extent2, like the early web, distributed, not centralised.

Fediverse services can be distributed because of protocols designed for this purpose, like ActivityPub, the W3C protocol on which Mastodon runs these days.3 Both ActivityPub and Mastodon are also designed so that thousands of users can be hosted on a single install of the Mastodon software. Hence the size of the mastodon.social instance.

The political economy of Mastodon

In the past, the cost and institutions required to publish and distribute media, was of massive political and social consequence. The arrival of the web, dropped the cost of publishing substantially. Still come the year 2000, only 7% of the global population had internet access, and only a fraction of these had published anything in public online.

Still come the year 2000, only 7% of the global population had internet access, and only a fraction of these had published anything in public online.

Then, then suddenly, the affordances of so-called Web 2.0 made it free and easy to publish from a user perspective. Pubishing on the Internet moved from being the preserve of a tech aristocracy to near everyone.

Caption: Only 7% of the global population had internet access in the year 2000.

But for the platforms that make this publishing possible the costs are not insubstantial. Hosting thousands of users requires considerable computing resources, engineering and administrator time and therefore — money. In the last few days, as more users signed up to Mastodon and servers slowed, the issues once again came to the fore.

How much does it cost to host a Mastodon instance? That really depends on who you host it with if you don’t have the wherewithal to do it yourself. You also need to know how fast you want your server to be and how many users you want to use your instance.

According to masto.host, a dedicated Mastodon host, hosting currently cost between $6 per month for an estimated 5 active users and $20 per month (plus tax) for an estimate of 100.

That’s also not the whole story. The cost also depends on what kind of users you host. On social media, some users are more equal than others. Two weeks ago, British broadcaster and celebrity, Stephen Fry joined Mastodon. This prompted developer Aral Balkan to blog about hosting costs when a user is very active and has a large following. Aral then had 23,000.

That’s not the whole story. It also depends on what kind of users you host. On social media, some users are more equal than others.

Aral hosts what amounts to a single user instance at masto.host. He shares his instance with no one. What prompted him to write about this topic was that it was his birthday, and as a result, he was very active on Mastodon. He posted an image of himself and received 230 replies from well-wishers. All of them were answered by Aral.

Hugo Gameiro, who runs masto.host, explained to him his instance’s performance:

You just get a lot of engagement and that requires a ton of Sidekiq power to process.

For example, let’s look at your birthday post … besides requiring thousands of Sidekiq jobs to spread your post through all their servers (you have 23k followers, let’s assume 3k different servers), as soon as you create the post 3k Sidekiq jobs are created. At your current plan you have 12 Sidekiq threads, so to process 3k jobs it will take a while because it can only deal with 12 at a time.

*Then, for each reply you receive to that post, 3K jobs are created, so your followers can see that reply without leaving their server or looking at your profile. Then you reply to the reply you got, another 3K jobs are created and so on. *

If you replied to the 100 replies you got on that post in 10 minutes (and assuming my 3K servers math is right). You created 300K jobs in Sidekiq. That’s why you get those queues.

Reading this, particularly the part in bold, implies that having users on the same server, most probably has efficiency advantages. But multi-user instances have also been struggling, partly no doubt because as they grow in users, the amount of ties to users on other services must grow exponentially.

As for his costs, Aral says:

Prior to the latest Twitter migration, I was paying around €280/year (or a little over €20/month) for my Mastodon instance on a custom plan I had with Hugo from the early days. This week, I upped that to a roughly €50/month plan. And that’s still not enough as my birthday post just showed so Hugo, kindly, has suggested he might have to come up with a custom plan for me.

Single user Mastodon and paying your own way

Aral points out that Stephen Fry, who joined the instance Mastodonapp.uk, and already has more followers than him, will bring substantial costs to that server and potentially do what amounts to a denial of service attack - overwhelm its capacity so its stops working if they don’t add resources.

He argues eloquently that Mastodon and ActivityPub should be redesigned so that all installations are single user ones. Then, everyone pays their own way. Aral believes that Mastodon could make single hosting cheaper, too, if it and ActivityPub were designed to encourage true decentralisation (i.e. single installs), it could be simpler, and “having more instances in the network wouldn’t be a problem”.

To him, ActivityPub and Mastodon are designed the same way Big Tech is: to encourage services to host as many “users” as possible. Says Aral:

“Today, we equate the size of mastodon.social (the instance run by Eugen) with how successful Mastodon (the software created by Eugen) is. This is very dangerous. The larger mastodon.social gets, the more it will become like Twitter.”

On this I am in partial agreement with Aral. On Mastodon on the 31st of October, I wrote:

A social network (like Twitter) can not be run well, by using old-school email-like distributed technology (in email’s case, the SMTP protocol).

It will be slow, updates, and messages will arrive asynchronously, and algorithms will necessarily be dumber.

In the end, everybody will gravitate to the best-run and governed Mastodon server. An economy of scale. But it will be expensive.

To finance it, it will require ads.

Where I differ from Aral, who regularly rails against “surveillance capitalism”, a concern which I think is largely overemphasized by many4, is that a Mastodon instance becoming more Twitter-like, is neccesary — if we want something Twitter-like.

Why?

What’s the matter with Mastodon?

Firstly on cost: These (even $6 per month) are not insignificant costs, especially for Twitter users in a place like South Africa. Advertising in wealthy countries subsidise service and even internet access in developing countries. Because Facebook was earning enough through advertising, it could promote and pay for Facebook 0 (zero-rated access on mobile phone networks) in many developing countries. Unlike Facebook, in India Twitter caters only for the most wealthy layer of society, partly for this reason.

Techichal decisions have performance and affordance implications. Centralisation vs distributed is one such case. One reason the distributed Bitcoin protocol could never be widely used as a electronic medium of exchange is because it is too inefficient and slow. Gmail certainly would not be as good at recognising and catching spam if it was distributed. Shopify shops experience very few chargebacks and fraud, because the platform can monitor behaviour across thousands of shops on a global platform level and spot patterns of behavior. It is hard to imagine how something as powerful and useful as Google Search could be architectured in a distributive way.

Similarly, you simply can’t use the power of machine learning, data and other algorithms to surface the kind of personalised and serendipitous content, nevermind quickly, as you could on Twitter — with a distributed design. The other features that would make Mastodon more Twitter like, like searching for a user by name, or toots that link to a URL, is best done in a centralised way.

Of course all these things that make Twitter so powerful, is also what can make it toxic - or a place where misunderstandings, harassment and disinformation are rife. But rather than enschew this kind of technology, we would be better served by regulating it.

Of course all these things that make Twitter so powerful, is also what can make it toxic - or a place where misunderstandings, harassment and disinformation are rife. Rather than enschew this kind of technology, we would be better served by regulating it.

Mastodon’s onboarding and discovery user experience (UX) is neccesarily poor because of its distributed nature. This will mean one of two things. It remains the domain for more technically literate users, or centralised services will be built on top of it to compliment it.

It remains to be seen how Mastodon will cope with spam and moderation once it grows in size. Early signs (here and here) are that relying on distributed intances with their own rules, volunteer moderators, and server admins with god complexes, will make for even worse and more arbitrary decisions than on Twitter.

The power of weak ties

Distributed Mastodon instances are often organised around a profession, a topic or an identity. And as several studies show, this homogenuity will probably make Mastodon less good at sparking new ideas and creativity for the average Mastodon user than Twitter. There is strength in so-called “weak ties.”

By highlighting all this, I don’t mean Mastodon is a failure, but that it is different. Mastodon is probably better at facilitating smaller scale and stronger communities (so-called strong ties) and probably will enable more in-depth exploration of a subject.

Because of the lack of algorithms selecting content on Mastodon, who you follow matters a lot more. As does your time zone and when you are online. You could still get a lot of novel information if the follows of your follows are very diverse and smart, but you will have to look harder or miss stuff that’s important to you. So Mastodon has its advantages and value over Twitter - like for institution building and mobilising groups to get things done.

Embrace and Extend

There is also a flaw in Aral’s argument. As he says, email and its protocol SMTP were also designed for us to communicate in a distributed way with our contacts. Then Gmail came along, which now hosts millions of email accounts. Aral acknowledges the consequence of this:

Do you know what happens to your email if Google says (rightly or wrongly) that you’re spam? No one sees your email.

If, however, in a counterfactual world, each email user had to manage their own SMTP install on a server, Gmail’s appeal would have been so much more potent when it came along. Do you have enough disk space for email? Do you manage your own spam list? Are hackers targetting your server to hijack your SMTP instance to send spam? Good luck with expecting the public to be concerned with all that. In this single-user counterfactual world Gmail would have dominated much faster and to a greater extent.

As long as Aral’s single user instance is built to communicate with others via a shared protocol, it is possible that somebody builds a client, probably in the cloud, that is so superior in terms of UX, ease of use, not to mention — cost, that people will gravitate to it, and comes to dominate the system.

This strategy for dominance is’nt new, it’s called Embrace and Extend, and the most celebrated case was countered, not via a techical solution, but in a court of law. I’m taking about the anti-trust case against Microsoft, who embraced HTML, but added it’s own features to its browser and operating system in an attempt to make its software indispensable when using the web. Until the law put a stop to it.

If, however, in a counterfactual world, each email user had to manage their own SMTP install on a server, Gmail’s appeal would have been so much more potent when it came along.

And just look at Web3 today. To keep things simpler, the Ethereum protocol is necessarily limited in the functionality it provides, especially to a user that interacts with the protocol using a client and not a server. So innovation needed to happen on top of the protocal to make it accessible to the general public. And these solutions are centralised services.

As Signal founder Moxie Marlinspike blogged earlier this year: Almost all Ethereum’s distributed apps use either the centralised Infura or Alchemy services in order to interact with the blockchain. That’s because those things that are truely distributed move at a glacial pace. But centralised services can rapidly iterate. It’s no contest. I agree with Moxie when he says:

People don’t want to run their own servers, and never will.

Markets, technology or opting out of politics

All kinds of issues can be solved through market solutions, according to the neoliberals. For libertarians, technology can often achieve the same ends, especially if without any government intervention. Left-wing anarchists meanwhile, want to opt out and use technology to roll their own, for “freedom”. It leads them all to a similar place in one key respect. If tech and the markets solves all our problems, we would not have to do messy politics, and the institution building and compromises it often implies.

I’m not for opting out, or simply relying on the market. I’m for shaping the technology we want. In his wonderful essay, Do arfitacts have politics, Langdon Winer writes:

For that reason, the same careful attention one would give to the rules, roles, and relationships of politics must also be given to such things as the building of high ways, the creation of television networks, and the tailoring of seemingly insignificant features on new machines. The issues that divide or unite people in society are settled not only in the institutions and practices of politics proper, but also, and less obviously, in tangible arrangements of steel and concrete, wires and transistors, nuts and bolts.

We do not have to accept that technology, the “tangible arrangements of steel and concrete, wires and transistors, nuts and bolts” are deterministically just the way it is. And the best long term way to shape our technologies for the best of society, is with “the institutions and practices of politics proper”.

Or as Yochai Benkler recently summarised:

technology develops as a function of institutional choices; that it is the subject of politics and the site of politics; and that it makes a difference.

It does not help that all the big social media companies are under the legal jurisdiction of the United States, a place where politics is so divisive, the political system is so sclerotic, and the constitution so simplistic and dated. But that Mastodon and many of its largest instances are based out of Germany means we should be able to rely on politics and law much more.

If Mastodon is going to replace Twitter, it will most likely have to become more Twitter-like. Regardless, what we need is properly regulated social media.

Principles for regulation of something like Twitter

What could sensible regulation of this Twitter-like site look like? Not everyone would agree on all of these, but off the top of my head, here are some principles that could guide regulators above and beyond existing laws on defamation and hate speech.

Users should be able to use pseudonyms and have multiple accounts, but from the platform’s perspective, all must be KYC’d (know-your-customer) like in banking;
User identities should only be revealed to law enforcement upon the issue of a warrant and only in democracies with the rule of law.
Bot accounts must be registered by a KYC’d human.
The larger your follower count, the greater an account’s responsibility for the content they create or share.
Sharing another user’s content makes an account as liable for damages or criminal prosecution, suspension or banning as writing it.
Users must be able to export all their content, including a list of their connections, in a non-propriety format which makes it readily available to be parsed by common digital tools.
Users should have the right to hide their content and their profile from public search.
Worth considering: In light of this research, and my speculating here, limiting the maximum amount of followers an account can have (say 10,000), or requiring social media users to get a “driver’s license”5 and make a Hippocratic oath before they can have more than a certain amount of followers.

An instance is an install of the Mastodon software on a server. Typically you’d have one per server.

One way Mastodon is not distributed is its governance and production, which is gatekepted by Eugen.

Mastondon, launched in 2016, preceding ActivityPub, which was published in 2018. Mastodon first ran on the OStatus protocall.

I have dealt with some objections to the idea of surveillance capitalism as it is often expressed with respect to data here and here, and advertising here and here.

Where they are tested on their knowledge of issues such as identifying misinformation, the impact of toxic content and how information spreads.

Kill all ads

Wed, 16 Nov 2022 00:00:00 GMT

Caption: Stable Diffusion: “American psycho on a billboard”

It’s de rigueur to say that if you don’t pay for a product, then you are the product. Shoshana Zuboff’s The Age of Surveillance Capitalism, a thick tome, has been an unlikely bestseller.

Together with former PayPal founders Peter Thiel and David Sachs, Elon Musk is part of an influential group of libertarians that has united conservatives and former leftists in Silicon Valley. If not using quite the same rhetoric, Musk is angling for war with the advertising establishment under the banner of free speech.

Musk’s thinking here as it relates to Twitter isn’t difficult to picture. For him, Twitter is saddled with hundreds of unnecessary moderators, editors and ethicists, catering to easily offended liberal sensibilities. Musk has put the platform in debt, so getting rid of this expensive lot can help fill the hole in its books. Some advertising could be partly replaced by subscription fees.

Before making an offer for Twitter, Musk may not have understood advertising problem he’d face executing his vision. Or more specifically, “Brand Safety”. For those that are unfamiliar, “Brand Safety” is the idea, widely held in marketing and advertising, that brands should not appear in “unsafe” environments.

Even though it is said that we live in an age of surveillance capitalism, it is curious that on social media or the media in general, there is comparatively little conversation and debate about the ads business (compare it with crypto).

Ben Evans (below) is right. Advertising, or online ads, in particular, is not well understood. Not in tech, not anywhere. How does the online ads business even work? And what are the finer points of how they operate on Twitter?

Ads is not a big industry, and it is becoming smaller

There is a lot of talk that “data is the new oil” lubricating the wheels of the advertising industry. In truth, if we make a crude comparison, advertising (including the sizeable part of it that does not use data much) is a much smaller sector than oil and gas, even in years when the latter did not do very well, and advertising boomed, like 2021 (see graph).

It is also a little remarked upon fact that since the year 2000, at least in the United States, advertising has been in decline as a share of the overall economy. From 1.5% of GDP to below 1%. Hardly a sign of raging surveillance capitalism. At the same time, advertising funds many of the services we rely on every day, and much of journalism, a vital public service.

Caption: Source: Ben Evans

Brand vs direct response advertising

A basic understanding of online advertising requires an overview of two forms of advertising that preceded online advertising but found their own unique expression in this interactive medium: brand and direct response ads.

In other mediums like print and TV, the kind of message but also the standards of accountability and brand safety, vary from one type of advertising to the other. In digital, the technology infrastructure and data that are required to deliver successful brand or direct response ads also differ.

Brand advertisers aim to get the public to recognise a product or, even better, have positive associations with it. Then when consumers walk into a store, they reflexively reach for the product they recognise.

Brand advertising dovetails well with the legacy mass media of the past and still typically serves mass-market products that do not require much by way of audience segmentation. Large audiences provide economies of scale. Big budgets are required, but audiences are cheap on a cost-per-thousand basis.

Direct response advertisers, in turn, want immediate results. That’s hard to achieve cost effectively if your product is not quite differentiated and your ads targeted, so this form of advertising tends to cater for more niche products, small audiences and budgets. The more targeted the advertising, the fewer people they need to see their ads, and the more efficient and “cheaper” the campaign.

Brands are for normies?

Brand advertisers expect environments that are not violent or overtly sexual (it’s what killed Tumblr and forced a clean-up of up-skirt sub-Reddits), even “positive”. Think of the squeaky-clean corporate cliché. It’s not far from the truth. Direct response advertising has fewer pretences. See, for example, how classifieds, even in respectable newspapers, used to allow advertisements for “massage” and other risqué services in many countries.

Ironically, in the past, criticism of such stuffy and vanilla mainstream corporate culture has been the domain of left counter-culture (See bell hooks’s “Teaching to transgress”, or De Sade, Bataille, Foucault) and the Bay area’s own digirati (like Stuart Brand, Kevin Kelly) and their anti-corporate hippie ethos. Angela Nagel in “Kill all Normies” is excellent on this history.

The rise of direct response on the Internet

Traditionally brand (TV, Cinema, Billboards, Magazines) have been much larger in money terms and prestige and glamour. Brand celebrated higher aspirations like creativity (The Cannes Lions is a brand event). Its advertising’s poetry and speaks the language of emotion. Direct response was a place for bean counters, dour, mechanical and unimaginative in comparison. It is personified by its workhorses: classifieds, telemarketing, and direct mail.

But then the Internet happened. Search is clearly a form of direct response and not brand advertising, yet soon it became the biggest and most successful category in the nascent online ads industry.

Caption: In the UK the world’s most developed ad market come 2011 Search was rising rapidly, online Display was barely 5%, TV (mostly brand) by far the largest.

In the early Dotcom boom, the rest of the web industry soothed themselves, “never-mind that we can’t compete with Google”. “Like direct response, elsewhere, search is for small businesses and niche products”, was the argument. Getting brand advertising budgets to move online was where it was argued the real money was.

Yet the new Online “Display” ads were clickable, which meant they held the possibility of both brand and direct response. But very few people clicked. Thus, web portals like Yahoo! tried to sell banners in a similar way to how TV ads were sold (as a reach product), underselling clickability. The industry struggled.

The quest for targeted “Open” Display Ads

To be sure, a few did focus on transforming Online Display advertising into a data-driven direct response medium. After all, digital is an interactive medium. A cornucopia of data brokers, ad exchanges, and ad networks based on cookies sprung up. Some were cowboys. Data was handled fast and loose.

Now sometimes referred to as “Open Display” ads, they are the bread and butter for 95% of ad businesses (including most former newspapers online) that depend on them.

Caption: The Total Digital Ad Market here is Google and Facebook, neither of which is Open display Advertising.

Why “Open”? Most websites (or apps) are only occasionally visited, let alone once a day. They don’t have much data on who their users are. Thus, the purveyors of Open Display Ads banded together, using data brokers to share data via cookies. And these brokers use machine learning in an attempt to make their predictions of who users are better.

But how effective are Open Display Ads? We now have very thorough research on that. Professor Christine Tucker of MIT researched and tested the data held the world’s largest data brokers in 2019. After all the hype about data and oil and Open Display ads being sold as direct response, professor Tucker expected “finely tuned and intrusive data about what she was thinking that day”.

Below you can see some of the most common categories of data sold. It’s pretty basic. The top two are age and gender. Yet Data brokers averaged about 50% accuracy for gender data, not much better than a random guess! And there was no correlation between the amount of data a broker held and the accuracy of their predictions about users.

Tucker’s research also shows that the poorer you are, the worse the prediction - because poorer people’s data is so fragmented. The lack of data on the less privileged she calls “data deserts”. “Surveillance capitalism” as we imagine it with respect to ads, appears to be the worry of the relatively well-to-do.

The arrival of “On Platform” Display Ads

A different category of ads, sometimes now referred to as “On Platform Display”, has emerged in recent years. For a while, only one company personified it. When Facebook came along, Display ads performed noticeably better, but without relying on data brokers and operating in a closed system. Why did they perform better?

Facebook’s ads appeared in users’ feeds, exactly where their attention was focused, and in the same format as the content they consumed. Ads could have an element of social proof: people (even your friends) could like, comment, and even share them. This adds a potential for organic distribution and also potentially makes these ads cheaper, encouraging advertiser creativity. The line between direct response and brand advertising got even more blurry.

But Facebook also had massive reach (millions of users) who spent a lot of time on the platform and as a result — far more data. Not data Facebook specifically asked for. Behavioural data like who people followed and what they liked.

Machine learning comes to Facebook

But like Twitter’s ads, who also appeared in the feed, Facebook ads did not perform that much better than Open Display Ads. 2013 marked another step change. Facebook introduced a new ad feature called “Lookalike Audiences” that employs machine learning. Using it, you can target audiences in a sophisticated and deep manner.

Say a business had a Facebook page that had 10,000 Likes. They could tell Facebook’s algorithm to find 300,000 people to target with ads on Facebook that are most similar to those users that Liked their page. The advertiser would not know who was included in this audience, or indeed their salient demographics or characteristics.

In 2015 they enhanced their advertising capability further with the ability to track behaviour on third-party sites and apps. Combined with Lookalike audiences, it was really powerful and drove a lot of Direct response advertising.

Not because it added much to the volume or breath of data Facebook held. Often it only added one data field. But rather like an effective Stable Diffusion image generation prompt, it allowed Facebook to hone in statistically on the most vivid picture of which users in its database was most similar to those that have done something off its platform.

It allowed a retailer to tell Facebook which of its customers were buying a particular product, and Facebook could link these customers with actual Facebook users. Then, when they had enough of these Facebook could use Lookalike audiences to create an audience of Facebook users with similar traits, which the retailer could target with advertising.

This kind of advertising proved spectacularly effective. Having Direct response advertising that is this targeted has been extremely profitable for Meta, allowing it to start competing with Google. Advertising is not that big of a sector, but Alphabet and Meta dominated it to such an extent that for a while, both companies were in the top 5 most valuable ones in the world. Today Meta has over 10 million advertisers, many of whom are small to medium-sized enterprises.

Twitter, actually does not do too badly as an ad platform when you compare them against anybody but Meta in social. This graph (below) by digital ad expert Eric Seuifert shows it comfortably outperforms Pinterest and Snap. And Twitter’s revenue growth is not to be scoffed at, particularly from the vantage point of legacy media.

Caption: Source: Eric Seufert

Yet, in 2020 Twitter revealed that 85% of their advertisers were pure brand advertisers, and more recently that they only have around 3,000 advertisers overall per month. It has not been able to match Meta in terms of targeting. If you ever looked inside your Twitter dashboard (when you still could), you would have seen how little Twitter knows about you.

Does Twitter not have enough or the right kind of data to be able to do direct response well? Or have they just been slow to build the required tech? Recent Regardless, Twitter, to this day, is primarily known as a place for brand ads.

Caption: Source: Ben Evans. Outside of On Platform Alphabet, Meta and Amazon, online ads is small.

While doing better than most with its brand advertising, Twitter has been a bit stuck, never making the leap to direct response, while Twitter also does not have the reach that brand advertisers typically desire.

It’s an Open question whether having thousands of primarily direct-response advertisers will necessarily save Elon’s bacon. The lines between brand and other forms of advertising have blurred somewhat online. If Twitter descended into a free for all, it would probably have even fewer female users than it has even today. But many smaller businesses are less concerned with their public reputation.

It won’t be easy. But to be fair to Twitter, only Amazon has managed to join the eye-watering profitable club of highly targeted direct response that includes Alphabet (Google) and Meta (Facebook, Instagram) so far. The irony is that instead of selling data, these companies guard it jealously.

With its push into ads coupled with the hobbling of its competitors via App Tracking Transparency (ATT), it will make Elon’s life even harder. And perhaps Apple itself will join this exclusive club of three, cynically remaking the whole ads landscape and, indeed, the Internet economy in the name of “privacy”.

Source of some of the slides used, Ben Evan’s presentations.

The end of free speech

Sat, 12 Nov 2022 00:00:00 GMT

Caption: Image: Stable Diffusion: Bird in chains in the style of Picasso

Suddenly we are considering Twitter’s mortality after Elon Musk piled debt onto the platform and erratically “managed” it for a fornight. He ~~now wants~~ wanted to charge allcomers $8 per month for more visibility in search and replies, as well as for these users to see half the amount of ads.

There were even reported discussions of putting all of Twitter behind a paywall. With visions of 4chan, advertisers have taken fright at his talk of ’freeing the bird’. Some users appear to be abandoning Twitter for a mostly ad-free, distributed Mastodon.

Can Musk’s erstwhile plans succeed? It depends on how you define success. However, it raises an interesting set of questions about the business model of social media. Does something like Twitter have the potential to be a paid service? To what extent are ads a good or bad way to support a global public sphere?

Caption: Image: Stable Diffusion: Caged Bird in the style of Picasso

A paid for social network

Free is an unbeatable price. But if Twitter was a paid-for-to-post (but one could still read it for free) service what would happen? Some fraction of users would decide not to pay. Since these non paying users would no longer be part of the network, the network would be of less value to some new fraction of users, who would also not pay. That’s basic network economics.

The impact would not be uniform. Some networks of interest might be more vulnerable than others (EconTwitter vs BlackTwitter). If you were to raise prices to cover this loss in revenue this could cause another round of user departures.

It is difficult to predict where the stable equilibrium will lie. But it will be a smaller network than when it was free and ad-supported. The price charged would have a dramatic (non linear) effect on what the equilibrium size will be.

All these considerations are complicated by the fact that less than 20% of Twitter’s regular daily users are based in the US. Prices elsewhere must be different to reflect purchase parity (often dramatically so), otherwise one would see a larger drop in users in those locations’ networks. (Meta’s US and UK users have subsidised users elsewhere, because their advertising markets are so developed.)

If there is differential pricing depending on the country you find yourself in, some users will use VPNs to mask their location to get cheaper access. To counter this, will some content be location specific?

And then there is Apple’s payments tax. In the 1st year 30% of revenues of those users that use an iPhone would go to Apple, regardless where these users sign up to premium Twitter.

The bottom line is that Musk’s vision of Twitter as a global town square would be affected if it were paid for.

Caption: Image: Stable Diffusion: Caged Bird in the style of Picasso

Benefits?

What could the upside be? If ads are not very relevant (and quite frankly Twitter has been bad at targetted advertising) users will not experience this nuisance. In theory at least it will please the vocal if-you-are-not-paying-you-are-the-product-brigade.

Musk will not have to worry about “brand safety”. The non agressive and less risky environments brand advertisers demand. This could allow him to spend less on moderation. Though it’s an open question if a free for all would also see some users leave the platform.

It’s arguable that Twitter’s userbase has skewed male precisely because in the past it was seen as an aggressive and antagonistic space where harassment was rife.

Back to romance?

Wed, 06 Jul 2022 00:00:00 GMT

*“It is never wise to neglect the heart’s reasons, which reason knows nothing of.” * Eric Hobsbawm - The Age of Revolutions

When I was a young boy 6 or 7, growing up in the port city of Durban South Africa, I distinctly remember waking up one morning with a deep sense of dread. I ran to my parents’ bedroom deeply distressed. My mom, alarmed, asked me what was wrong? “I have, I have… …no adventure in my life!” I cried.

Little did I know I should not have worried. We lived in a world under the threat of nuclear annihilation. My own country was under the already teneous grip of the apartheid state. We lived underneath the radar, we lived underneath the bomb.

Today we also live in tumultuous times. Soon I am celebrating my birthday with friends. Since I organised it, who would have guessed a war of aggression would break out? Who would have guessed a world leader would threaten to use nuclear weapons? So, when looking for a theme for my party, I cast my mind back to an analogous messy and significant period in history.

It was easy to settle on one.

Caption: A recent reinterpretation of new-Romantic fashion.

What the late 18th and early 19th-century Romantic movement were for is hard to put your finger on, but not what it was against. It was a reaction to the world-changing impact of the ideas and way of life that spread because of the French Revolution, and its cousin, the Industrial Revolution. It was clear what the Romantics detested - the middle. Says Eric Hobsbawm in The Age of Revolutions:

“The contrast between a world theoretically wide open to talent and in practise, with cosmic injustice, monopolised by soulless bureaucrats and pot-bellied philistines, cried to the heavens. The shades of the prison-house –– marriage, respectable career, absorption into philistinism - surrounded them.”

George Sand spoke for the rebellious reaction of many a Romantic when she proclaimed:

“Once my heart was captured, reason was shown the door, deliberately and with a sort of frantic joy.”

Byron was clear-headed enough to foresee that only an early death was likely to save him from the respectable old-age this new world made not only possible, but likely. He did that by joining the Greeks in their independence struggle against Turkey — and paid with his life.

But the fundamental problem the Romantics had with the new bourgeois world created by the twin revolutions was not its lack of adventure, but that it was profoundly asocial. And the Romantics set themselves against the alienation that resulted from that. This was a new world that:

“…left no other nexus between people than naked self-interest. It has drowned the heavenly ecstasies of religious fervour, of chivalrous enthusiasm, of sentimentalism, in the icy water of egotistical calculation. *It has resolved personal worth into exchange value, and in place of numberless indefeasible freedoms, set up a single unconscionable one, free trade.” *

The voice speaking here was that of the Communist Manifesto - but says Hobsbawm, it speaks for the Romantics raging against alienation also.

The end of history was a long time coming

Underworld’s first hit, We Live Underneath the Radar, was released in 1988. Only a year later, it was dated. Because in 1989, when I was 17, the Berlin Wall fell. Although I sensed that the world had changed dramatically and that somehow, I would be affected by it, I had no idea how.

But somebody had some ideas. Months before the surprise fall of the Berlin Wall, a precocious young man called Frances Fukuyama wrote an essay which predicted the fall of the Soviet Union and much besides. In The End of History Fukuyama wrote:

“The triumph of the West, of the Western idea, is evident first of all in the total exhaustion of viable systematic alternatives to Western liberalism… this phenomenon extends beyond high politics and it can be seen also in the ineluctable spread of consumerist Western culture in such diverse contexts as the peasants’ markets and color television sets now omnipresent throughout China, the cooperative restaurants and clothing stores opened in the past year in Moscow…”

Caption: The iconic launch of Macdonald’s in Moscow in January 1990.

Fukuyama argued that a powerful (and outrageous) assertion made by Hegel way back in 1806 was making itself evident in 1989. Hegel is known (among other things) for his view of history, that despite struggles and setbacks, the future is always brighter than the past. History, shorthand for human freedom, is always on the march forward.

In that view, Hegel was not unique. But Hegel was the first to proclaim history to be at an end. He did so in 1806 after witnessing Napoleon’s spectacular victory over the mighty Prussian army at the battle of Jena (today a 2-hour drive from Berlin).

To be sure it was a physical victory of blood and steel, but more importantly for Hegel was what the victory said about the world of ideas. He put Jena down as a victory of the ideas and ideology of the French Revolution.

Caption: Napoleon at Jena

Continuing his essay, Fukuyama argues that the greatest modern student and interpreter of Hegel was a Russian émigré to France named Kojève. Despite the horrors of WWI and II, Kojève kept to Hegel’s idea of The End of History. For Kojève said that*:*

“…while particular regimes in the real world might not implement these ideals (of the French revolution) fully, their theoretical truth is absolute and could not be improved upon…”

And if ideological development had in fact ended, “the universal homogenous state” would eventually become victorious throughout the whole of the physical world.

At The End of History, this universalising state will be liberal in the sense that it will recognize and protect humankind’s inherent freedom through the law, and democratic in the sense that it will exist only with the consent of the governed. (And since it would be everywhere there would be no need for borders. That’s great right?)

Many of the friends I invited to my birthday party I met as this universal homogenous state spread. In cool Britannia, in Berlin. Would we have known each other if it had not?

Yet, an often-unremarked part of the essay is that Fukuyama saw an ambiguity in the state of affairs the spreading universal homogenous state would usher in.

“For Kojève, this so-called “universal homogenous state” found real-life embodiment in the countries of postwar Western Europe - precisely those flabby, prosperous, self-satisfied, inward-looking, weak-willed states whose grandest project was nothing more heroic than the creation of the Common Market”

Moreover, wrote Fukuyama, wistfully (and in a very male-centred way), in this future with no history – there will be no passions, no big adventures. In this future there…

“…is no struggle or conflict over “large” issues, and consequently no need for generals or statesmen; what remains is primarily economic activity.

As a matter of fact, Kojève’s life aligned with his teachings Fukuyama tells us. Believing that philosophers like himself had no work to do since Hegel had already achieved absolute knowledge, he left teaching after the war and worked as a bureaucrat for the European Economic Community until his death in 1968.

Caption: Kojève the bureaucrat.

This kind of thinking by Kojève is what Timothy Snyder, Yale historian and an expert on Ukraine, calls “the politics of inevitability”. A politics that has been dominant in the US and Europe for decades. The basic notion on which this politics is built is that Fukuyama was right. History was over. Markets bring capitalism. The West is exceptional, “but will no longer be because everyone will soon become like us.”

Thinking like that brings with it a predictable set of problems says Snyder. Like, for example, wealth inequality, because all sorts of things that you expect to be progressive don’t turn out to be. So, we thought it doesn’t matter if oligarchs emerge. Because all that matters is creating a market economy. And that market economy will then magically resolve all the political and administrative, institutional, moral and cultural problems for us. We can buy energy from who we please. We do not need to regulate banks, as the market is self-correcting. This while technology and innovation will save us from environmental disasters. And oh, the internet is going to be liberating, democratising and a force for good.

And you never have to think about what values you may want, and how they may differ from the values of others. Because that’s redundant. And neither do your institutions, your technology, your art or architecture attempt to embody the values that you think would make for a better world. “The market is like mom, it will take care of you.”

Snyder’s insight is not unique. It is why Fukuyama’s essay soon became notorious for a kind of gormless liberal nihilism. But Fukuyama is perhaps unfairly criticised for being an evangelist for The End of History. The last paragraph of his essay which is rarely noted reads:

“The end of history will be a very sad time. The struggle for recognition, the willingness to risk one’s life for a purely abstract goal, the worldwide ideological struggle that called forth daring, courage, imagination, and idealism, will be replaced by economic calculation, the endless solving of technical problems,.. and the satisfaction of sophisticated consumer demands. In the post-historical period there will be neither art nor philosophy, just the perpetual caretaking of the museum of human history.”

There is a crack in everything, it is how the light gets in?

Yet many intellectuals see in a series of events, culminating in Ukraine a turning point, that asks serious questions about The End of History. Asked about the West’s failure to plan for the food crisis brought about by the Ukraine War, my favourite public intellectual, economic historian Adam Tooze, had this remarkable answer:

“We did not expect the war to go on. We expected the war to be short. We expected Russia to be in full control. So it’s the resistance of the Ukrainians, which is the wild card here. It’s exposed the inadequacy of the calculus of the west, as much as the inadequacy of the calculus of the Russians.

You know, that folks like me, who, who deal in technocracy and its logic and power, are often confronted by critics who say what you, you discount the role of resistance. You don’t focus enough on the opposition. You don’t understand the grassroots – and it’s true. And I genuinely don’t because I don’t think it matters. And relatively speaking, I think it’s actually powerless.

But if you ever wanted a counterexample of the way in which genuinely popular, organized resistance has totally disrupted the calculations of power, this would be it.”

And Timothy Snyder also thinks the war has larger implications because:

“…imagine that Zelensky had fled and that the state had collapsed, right? What kind of suffocating, intellectually and morally narrow, hollow world would we be in right now? What would we be talking about right now? Thanks to the fact that they’re doing something, they’ve given us this chance to think bigger.”

But let’s not kid ourselves. At my party, we will dance underneath the radar. But relatively safely. Where we will be, an hour from Berlin, is not the same as being in Kherson or Kharkiv.

Habermas captures some of this when he recently wrote about a gap that exists between Ukraine and Germany. Ukraine is struggling to become a nation-state. Germany is well beyond that. Germans, as well as the rest of us in the West, should think about this gap when considering our spontaneous reactions of enthusiasm and solidarity with Ukraine:

“We thrill to the heroism of the Ukrainians, which puts into stark relief the deflated state of our own politics. But our post-heroic culture cannot simply be cast-off in disgust”

Unlike Byron, I’m not joining the Ukrainian army. I am well into soft middle age. I myself quite like being comfortable and…

Not drinking too much Regular exercise at the gym (3 days a week)

Fitter Happier - Radiohead

So it’s in this strange in-between world, where History’s End still lives, but where we are now all too aware of its mortality (and I of my own), where my friends and I will see each other again.

At this party, like when I was 17, we will half dance underneath the radar. We are again kind of living underneath the bomb. Perhaps that weekend we can briefly get back to the future. A test run in casting off our post-heroic culture and alienation in disgust, dream of utopias — and have a little adventure.

“Art for art’s sake is an empty phrase. Art for the sake of truth, art for the sake of the good and the beautiful, that is the faith I am searching for.” - George Sand

The Open Web isn't dead... yet

Thu, 10 Jun 2021 00:00:00 GMT

This is Part 4, Part 1, Part 2 and Part 3.

In the halcyon digital days, Jeff Jarvis gave advice to digital business in his book, “What would Google do?”. One of his many Google inspired rules was, “do what you do best, link to the rest”.

Caption: Captain Pickard, assimilated by the Borg

Then, he wasn’t wrong. Initially, platforms focused on their main business concern and were happy to hand off users to other companies to provide other services. In many cases, this handoff was a feature. As the gateway to the open, linked web, Google became emblematic with its search engine.

Apple’s new App Tracking Transparency (ATT) is a dramatic departure from this earlier web. Consequently, the internet is likely to become less open and more consolidated. This trend, however, is part of a broader one that has gained momentum for some time, as I will explain.

Caption: Wired - The Web is Dead - Sources: Cisco estimates based on CAIDA publications ANDREW ODLYZKO

Already in 2010, Wired Magazine declared the Web dead. Chris Anderson named the chief culprits. Native apps and us, the users. Because “as much as we love the open, unfettered Web, we’re abandoning it for simpler, sleeker services that just work.”

That was hyperbole. The Web is still very much alive more than a decade later. For example, for various reasons, most e-commerce still occurs via the web, not apps (for now). Meanwhile, HTML 5, launched in 2010, and programming paradigms like Node-JS have made web-based apps more native app-like: powerful, responsive and user-friendly.

It has never mattered so little what users want and do in the digital sphere. In today’s world, what Big Tech does is more important. And driven by the centralising logic of data and network effects, the big platforms have been consolidating: building economies of scale and scope.

As a result, platforms have been moving things they used to hand off to others internally for over a decade. “Do what you do, and also the rest.” is the new motto.

So Google search has been competing with an increasing number of websites it used to link to by providing answers to some queries directly in its interface. And as I explained here, this strategy of Google is reaching new levels of sophistication as it muscles in on other services like banking. Amazon, in turn, had been moving into advertising. Its “On Platform” ads came out of beta way back in 2012.

And Apple has tried its hand at advertising a few times — and often failed. Its new ATT measures, even in the highly unlikely case they are not designed to do so, will benefit its own On Platform ads as its competition suffers.

Resistance is futile?

When Facebook launched its e-commerce offering in the US in August of last year, it mentioned that it was a response to Apple’s ATT tracking measures. In truth, Facebook and Instagram had been moving in the direction of hosting shops themselves for some time before then; Apple just vindicated and provided urgency to the strategy.

Be that as it may, Apple just gave retailers another good reason to rather open a Facebook or Instagram shop if they don’t have an Amazon one already rather than start their own on the open web.

The reason is simple. As with retailers on Amazon, Apple’s new tracking rules won’t impact the targeting ability as much of Facebook or Instagram shops that also use Facebook or Instagram advertising. That’s because they are end-to-end ‘On Platform’. In other words, the audience, the ads and the destination service are all inside the same service.

To be sure going “On Platform” will have a downside for these retailers. Like when having an Amazon shop, Facebook retailers will have little control over the user experience. But more importantly, they will receive minimal data about who is buying their products - they will not, for example, have the email addresses of their buyers.

It is the retailer who will make the sale, while Facebook will collect the data and know who your customers are. However, the alternative may be more detrimental. In digital, one of the biggest pain points is being discovered and seen. Effective advertising is essential. What is the point of resisting?

This trend of hoarding more data by big platforms is not new either. In the name of privacy, platforms have been building moats around the data they have been generating for some time. Remember the Facebook API? Around 2012, award-winning digital experiences were built interfacing with it. Intel’s Museum of Me, which created a personal gallery from a user’s Facebook experience, was one such effort.

It did not last. First, Facebook selectively cut competitors Twitter and Instagram out of parts of their API, making onboarding more cumbersome. Then Cambridge Analytica’s misuse of the Facebook API was the reason Facebook gave when depreciating much of its API functionality for everybody.

A new era of consolidation is beginning with Apple’s ATT.

ATT, consolidation and the rise of “Content Fortresses”

Apple’s ATT rules expressly permit companies with multiple digital properties to share data for tracking purposes. Why did Apple leave this huge loophole in its rules?

Apple no doubt wants to grow its own On Platform advertising (I discuss that here), and it has many services that sit across several applications. Aware it is being watched by regulators, it is keen to show that it treats other services like it does itself (it doesn’t - a subject for another post). Hence this exception.

Yet this single rule is likely to have massive unintended consequences and further online consolidation. And it seems to have already started. The Financial Times reports that:

“The belief that “bigger is better” is already driving consolidation. Glu Mobile, a California maker of games such as Kim Kardashian: Hollywood, realised in July 2020 that it needed to go on an acquisition spree, or be acquired, to “combine scale to address the business model challenges posed by the pending release of an Apple iOS update”, according to details in its most recent annual filing. Within a few months it was acquired by Electronic Arts for $2.1bn.

Christian Selchau-Hansen, chief executive of Formation, a software specialist for personalised marketing, says “gaming M&A has spiked” as companies have realised that advertising across a portfolio of games will deliver “an incredible advantage over someone that has a single gaming app.”

Others point at the acquisition by Applovin, the mobile ad platform and games publisher, of Adjust, the Berlin-based mobile advertising attribution company. It is putting all the pieces into place to have its own advertising ecosystem.

Meanwhile, in a post in response to Apple’s App Tracking Transparency, digital ad expert Eric Seufert warns about the rise of “Content Fortresses”. He points to Zynga, who “in its Q4 earnings call last week, mentioned two new initiatives that signal its desire to construct walls around its first-party properties in an attempt to build an autonomous ecosystem: cross-platform play, and building an advertising network.”

Some game publishers now consider ditching advertising as a business model.

A change of business models?

Other services will look at new revenue streams that do not rely on advertising. It is in this context that Twitter’s new premium paid-for features should be viewed.

How Shopify, the huge platform that powers many independent shops respond, will be interesting to watch. To maintain ad targeting capabilities, shops on its platform may consider moving to Amazon or Facebook’s ecosystems.

Shopify shops are already on one giant platform, albeit with a highly flexible and customisable interface. Shopify may be tempted to offer a platform-wide search engine, and if so, Shopify On Platform Ads and a Shopping Portal iOS app may not be far behind?

Another sector to watch is publishers. Expect a mixture of consolidation and closures. Some might try and output higher quality content (a good thing) to be able to charge directly for their content. But I suspect only a few publishers, especially generalist news providers, will be able to emulate the New York Times and successfully launch paid-for models.

So we can expect more bloodletting.

This is Part 4, Part 1, Part 2 and Part 3.

What is Apple really up to?

Wed, 09 Jun 2021 00:00:00 GMT

This is Part 3, Part 1, Part 2 and Part 4.

Caption: Caravaggio

Digital Advertising guru Eric Seufert is rightly worried it “will take regulators years to discover what Apple has accomplished with ATT: wresting advertising market share away from Facebook while using proprietary — because it owns the App Store!—install/revenue data to target ads for its ad network.”

Let’s take a quick step back and recap. Apple launched its Search Ads inside the App Store in 2016, soon after shutting its failed iAd advertising product. Apple reports this revenue under the new category of “Services” (It does not break out the advertising component that includes subscription revenues for things like News and iCloud.)

And this Services revenue matters. Hardware sales growth was slowing (no surprises here, it is massive already). But this new category kept growing at double-digit rates. In Q1 2020 for example, hardware sales slightly dipped, but Services grew at 22%. More importantly, making up 38% of Apple’s gross margin.

Then in April of 2020, Apple updated its Apple Search Ads campaign management API to add new parameters —supplySource and adChannelType. Back then Eric Seufert correctly guessed that these parameters point to the fact that Apple would extend the search ads into some of the other parts of its services. And soon enough Search Ads were spotted in Apple News.

And yes, if by now you are thinking this all sounds hypocritical coming from a company preaching privacy, then you are right. It is profoundly so. Seufert again: “Apple’s proprietary ad targeting—which the company insists isn’t “tracking,” because all data used is first-party—segments users based on the apps they download and the IAPs they make in those apps. It just so happens that Apple’s 1p data is pretty helpful for segmentation.”

And what is this Apple first-party data Eric is referring to? It is this

Notice, it includes location. And…

It is this talk of “first party” data that also inoculates the other big platforms from Apple’s ATT measures inside their own “Content Fortresses” - the “On platform” play that means smaller publishers, and Open advertising will suffer most.

This combined with Apple’s ATT exception that data can be shared between Apps that belong to the same entity is what is causing consolidation in the digital industry, as I will explain in my next post Part 4.

Apple has muddied the water. Regulators would be loath to intervene before the mud settles. It has made regulatory interventions harder —yet never has it been more necessary.

This is Part 3, Part 1, Part 2 and Part 4.

Are "Open Display" ads bad?

Mon, 07 Jun 2021 00:00:00 GMT

Caption: A vintage Nestle ad

This is Part 2, see Part 1, Part 3 and Part 4.

Introducing its new App Tracking Transparency (ATT) measures against “tracking”, Apple has been scant on the detail, but vocal as to why it is doing it. Speaking virtually at the Computers, Privacy & Data Protection Conference, Tim Cook, the Apple chief executive said:

“The fact is that an interconnected ecosystem of companies and data brokers, of purveyors of fake news and peddlers of division, of trackers and hucksters just looking to make a quick buck, is more present in our lives than it has ever been.

…it has never been so clear how it degrades our fundamental right to privacy first, and our social fabric by consequence.”

Cook draws a direct line connecting the “tracking” that Apple now seeks to block, to disinformation online and the fractious politics in democracies and societies in general.

But the likely impact of what Apple is doing belies the reason Cook has given. Curtailing tracking will have varying degrees of impact on parts of the internet advertising ecosystem. And it will likely significantly decrease the revenue of commercial institutions that support journalism, along with many other services, good and bad, that rely on it.

The decline of “Open Display” advertising

When we say “On Platform” ads, we mean ads served by the same entity that provides the audience and - increasingly and more significantly - also hosts the destination service or shop.

As I explained in detail in the previous newsletter, bar Amazon’s On-Platform ads, there will be some negative impact in terms of the efficacy of digital advertising almost across the board when advertising to iOS users going forward.

But the damage will be particularly great for so-called “Open Display Advertising”. This is advertising where the ads are often served by a different entity than that that provides the audience, and where the service advertised runs elsewhere.

And it is the kind of advertising on which everybody, except a few large digital platforms and a few niche operators of digital classifieds, relies. Of course, the platforms, especially Google, have a big hand in many aspects of the “Open Display” market too.

Google and Facebook (and recently Amazon), have been progressively increasing their share of the On Platform digital advertising market over the years. In 2019, the UK’s Competition and Market Authority already estimated the yawning gap for Britain: The “Open Display” advertising market was less than half the size (£1.8 billion) of those of the “On Platform Display” markets, (£3.8 billion). Initial evidence is that the pandemic has greatly accelerated this trend where On Platform advertising is growing, also on Amazon, while Open Display stagnates.

But whether Open or On Platform, Display advertising itself is not the biggest segment of the digital advertising pie. That honour goes to Search based advertising, which in 2019 was still significantly bigger than the other two combined (£7.3 billion), and in which Google has about a 90% share in the UK. Search will also be largely unaffected by the new Apple measures.

The smallest digital advertising segment in 2019 was classifieds (£1.4 billion), think services like Autotrader, Yelp, Gumtree, Trip Advisor. Classifieds tend to be “On Platform”, so should not be impacted by Apple’s new measures much either. Classifieds used to be a very big part of newspaper income so this disintermediation is yet another much less commented on reason why the news media industry is struggling in the digital era.

It is worth recapping why “Open Display” advertising is struggling in the first place and why they are likely to struggle more now. Ironically Jane Horvath, Apple’s privacy chief, put their finger on it. She according to the Financial Times argued in January that the “accumulation of large troves of data primarily benefits big businesses with big data sets”. Apple is now likely to accentuate this advantage by harming smaller businesses ability to have comprehensive data sets.

The hundreds of media properties that rely on “Open Display” ads simply do not have the behavioural data that the big platforms have, making their targeting ability weaker, and their ad prices lower. To augment this ability, this vast network of digital properties use ad networks that aggregate data to create profiles of the users of devices to target.

Apple’s ATT will mean that Apps running on iOS will not be able to access these ad network profiles, and neither will these apps be able to feedback data to these networks, making these networks less effective even outside of iOS apps.

While on the one hand, because it is On Platform, Amazon’s burgeoning advertising business will feel no direct impact and may well benefit, smaller businesses reliant on advertising (think publishers, games companies, even a business like Twitter) will most likely experience the greatest loss of revenue. Somewhere in the middle are the large platforms like Google and Facebook, which will feel some pain, particularly in the ad networks they run. But they may benefit in the long run.

If advertisers lose faith in the efficacy of Open Display advertising they will spend more On Platform and Search ads. And here Facebook and Google will remain big players, waiting for the Open Display refugees.

The fraying social fabric - what role do ads play?

If you follow this newsletter, you will note that the degradation of the public sphere by misinformation has also been a major concern. Let’s put aside factors like the shock of the global financial crisis and its unsettling of the liberal post-1990 consensus and the erosion of trust in experts and government. I have a hypothesis why this age of polarised politics may be happening: namely the erosion of the share of voice (reach) and authority of mainstream media gatekeepers.

A further shift of revenue away from “Open Display” advertising towards “On Platform” due to Apple’s measures will on the face of it exacerbate this problem, as it will further erode the revenue on which journalism relies. In other words, Cook seems to be making worse the problem he claims he wants to address.

All this is happening as the apparent duopoly that Google and Facebook hold over digital advertising is coming under increased scrutiny from regulators across the world. The figures above come from a 2020 report from the UK’s competition authority (CMA) into exactly that. It occasioned a recommendation to the UK government to create a new Digital Markets Unit (DMU) that will have increased powers with respect to measures against technology platforms. And just the last week the CMA announced further investigations into “whether Facebook might be abusing a dominant position in the social media or digital advertising markets through its collection and use of advertising data”.

Key questions regulators need to ask include, what is the social and economic value of advertising? Are there more socially desirable business models for media businesses? Does advertising help or hinder the start-up ecosystem? Are all kinds of advertising the same? should advertising systems be open or closed? Is consolidation really happening and to what extent is this a bad thing? All issues I will explore in the next newsletters.

This is Part 2, see Part 1, Part 3 and Part 4.

Apple upsets the app advertising cart

Wed, 05 May 2021 00:00:00 GMT

Apple’s new iOS update has a new feature called App Tracking Transparency, which it expects all apps running on their platform to implement, and is set to significantly impact digital advertising.

This is Part 1, Part 2, Part 3 and Part 4.

There is a lot of complexity in how the wider digital advertising ecosystem works, and with this new measure, quite a lot of uncertainty. Apple for one has not revealed a lot of detail, although it has been quite vocal as to why (more on that in the next post).

Apple has technical levers it can pull, and a lot can be simply enforced through technology. But some of what they claim they will want will require them to keep an eye on the wider ecosystem. To what extent will other players abide by Apple’s rules? Neither can we be sure how users will respond.

What is at stake? Advertising inside apps is a $90 billion market and growing fast. Both Google’s (32% y-o-y) and Facebook’s (46% y-o-y) digital advertising growth just shattered expectations for Q1 2021. And Apple had a stellar quarter itself. Yet a year ago, in Q1 2020, Tim Cook revealed that there are already a total of 1.4 billion active iOS devices worldwide.

In the US, 47% of all smartphones are iOS, over 110 million devices. In the UK, the most advanced e-commerce and digital advertising market in the west, 51% of smartphones are iOS. In Europe the figure sits at 30%. In the world’s largest smartphone market, China, “only” 21% of smartphones run iOS.

This post will look at some of the technical detail and speculate as far as we know what the impact might be on a granular level. If you are not a technologist, a digital ad person or a policy nerd, you might find this post a bit tedious. But I had to do this close-up look before I could start speculating about the macro.

The next post(s) will unpack some of the wider and more interesting implications, for the platforms, for publishers, for retailers, for privacy, for the future business model of the internet, for market concentration, for society and ask - should government intervene?

What is Apple about to do?

The App Tracking Transparency framework will ask iOS users of an app a simple question. Can the app they use track them? Yes or no. Considering the simple binary option, it is expected that most users will opt not to be tracked on most apps.

The first thing to note is what Apple means by “tracking”. Apple has not released a lot of detail, and much is still unclear.

“Tracking”

Apps and websites, by virtue of having to function, need to “track” what users do on them. So when Apple talks about tracking it means a specific kind of tracking that’s now prohibited. Apple explicitly prohibits apps and web sites, from disparate companies, from sharing data for the purpose of advertising.

It is the… “linking user or device data collected from your app with user or device data collected from other companies’ apps, websites, or offline properties for targeted advertising or advertising measurement purposes” …which will be prohibited.

So what will not be impacted?

Let’s start with a simple example. Did you leave items in your iOS Ebay app cart without finishing checkout? When you return, Ebay can “remember” that you had not completed the buying and the items can still be there.

Can you favourite a Tweet using the Twitter iOS app, and using the IFTTT service, have it sent to a Google Spreadsheet? Here we have an instance of the sharing of data between your iOS app and a different company’s service. But it’s not for advertising purposes. So it should be ok.

What about saving an Audible book in the Amazon app to a list and it shows up in your Audible app? Here we have another instance of the sharing of data between apps. And it’s also not for advertising purposes. And more crucially, Amazon owns Audible. It’s not another company. Even when if it was for advertising purposes it would be ok. (More on this huge loophole, and its possible effects in the next post.)

Let’s say you are an app owner or developer. Can your app send data to an analytics service (say Mixpanel), to see how users use your service? As long as it’s not used for advertising measurement purposes, then yes you can. Apple does not really need to police this much either. They simply won’t give the analytics service the IDFA (the Apple device identifier) that makes more granular device tracking on other digital properties - and therefore for ad purposes - possible.

Apple explicitly makes two more exceptions. “When the data broker with whom you share data uses the data solely for fraud detection, fraud prevention, or security purposes. For example, using a data broker solely to prevent credit card fraud.”

A sensible exception considering the amount of fraud out there. Pooling fraud data (a card number that’s been used fraudulently) across the internet makes it much harder for fraudsters.

The next exception may raise the eyebrows of those that hoped Apple was limiting tracking (or surveillance) in the name of stopping unbridled commerce (capitalism) on the internet.

Apple also does not want to limit the ability to track users for the purposes of building and checking credit profiles, even if users ask not to be tracked. So Apple says: “When the data broker is a consumer reporting agency and the data is shared with them for purposes of (1) reporting on a consumer’s creditworthiness, or (2) obtaining information on a consumer’s creditworthiness for the specific purpose of making a credit determination.”

A lot of in-game and other purchases are made via apps. And at times credit makes the wheels of commerce turn. Apple takes a cut from some of the revenues, plus some retailers won’t want to have shops on iOS if they can not offer credit.

What is no longer allowed?

But what if you are a retailer (not on the Amazon or Facebook platforms) and rely on advertising? Then it’s likely your ads are also running on an ad network embedded inside apps and websites. With this change, you can still run your ads inside the apps, but you will not be able to target them nearly to the extent you did before, for probably most iOS users.

Why is it that retailers not on platforms will lose some of their ability to target ads? Well, ad networks sit across many websites and apps. They provide both distribution for ads and “share data” to build profiles of users to target. I use the scare quotes, because although ad networks track and report which website or web page a device has visited, this information is usually centralised, and not exposed in its raw form to a member of the network. It is abstracted using statistical guesswork as targeting parameters like “male”, “London”, and “Arsenal supporter”.

Why use an ad network? Very many websites and apps do not tend to have large enough audience sizes (called reach in the industry) on their own to justify an advertiser spending time to advertise on them alone. A problem that’s more acute if your data is not granular enough to allow for narrow targeting. Even when they do have reach, many websites and apps simply do not have the troves of data (nevermind machine learning) that Facebook or Google has to enable detailed targetting.

How do ad networks build a profile of users to target? By knowing a device’s approximate location, and which websites and apps a device visits, a guess can be made as to the identity and interests of the device owner - so-called behavioural targetting. This data is often enriched by combining it with other datasets.

Currently, all this matching activity of disparate digital properties is based on device identifiers, IDFA on the Apple platform. IDFA can link a journey from a link in an app, to a website, to a device. Without knowing which apps iOS users use and which websites they visit, how often and from where they do so, the picture ad networks are able to form becomes vaguer.

What if you are an “ordinary” publisher, say the Daily Mail, the Guardian, the Mirror or a vertical like Planet Rugby? The relevancy of the ads you will be able to run on your website or app will be affected since the ad networks you use won’t have much data on many iOS users as before. Many users that land on your site and app, will just be that, an undifferentiated user. You will have to show many more ads to get a click. Expect ad prices to fall.

Now Facebook and Google do run ad networks themselves called Facebook Audience Network, Google Ad Sense, Google Ad Manager and AdMob (which incidentally Apple lost out on to Google in a bidding war in 2009). With them, publishers can offer the tech behemoths’ far richer targeting parameters but on their own properties. (Facebook, in general, seems to have best profiles due to the richness of meaning they can extract from their social network)

This part of Facebook and Google’s ad business will be significantly impacted, as iOS apps for users that refused permission will not be able to run the targeted form of these network’s ads. Facebook does not break out the revenue it makes from its network. But Jounce Media estimated, that pre-covid, that Facebook’s Audience Network would generate about $3.4 billion in gross revenue in 2020. (Since then Facebook has reported bumper increases in ad revenue.)

Facebook’s overall revenue for 2020 ended up being $86 billion. So while Facebook’s Audience Network makes a lot of money (in the region of Twitter’s total take of $3.72 billion for 2020), it’s probably less than 5% of Facebook’s overall revenue.

For Google, the revenue from its ad networks is several times this number. In 2020 that was $23 billion. But its overall revenue was over $180 billion, of which Search Ad revenue was $104 billion, and YouTube almost $20 billion. And bear in mind Google relies far less on ad revenue from iOS than from Android users. For Facebook, it’s the other way around. It is more reliant on iOS (Instagram was only available on iOS initially).

“On Platform” app ads

When you are advertising inside the Facebook app on iOS, inside Google Search and the YouTube iOS app, or for that matter using Amazon’s ads on iOS it is a very different story.

Inside the big platforms, ads are a part and parcel of the service. Facebook, for example, knows a lot about its users from its social network itself. (Studies have shown, that only based on what users like on Facebook, their race and sexuality can be determined with an over 80% accuracy.)

And in terms of Apple’s rules, when it comes to sharing data for advertising between Facebook and Instagram, or Google Search, Shopping and YouTube, and of course between Amazon and the streaming site Twitch - they are considered one company. So advertisers will have more options than publishers and retailers to still finely target their advertising to a large batch of iOS users.

Not that advertising On Platform with Facebook and Google will have no problems from this new iOS tracking regime. It would be naive to think that neither Facebook and Google did not further complement their existing rich profiles of users with data they got from their app use. That supplementary data for many iOS users will now be gone.

But also, attribution to the advertisers of some of their iOS users in ad campaigns will become harder. Let’s say the retailer above uses Facebook Analytics to track their Facebook ads that run inside the Facebook iOS app. They can no longer use the IDFA device identifier to track the attribution of campaigns on a granular level to the user actually buying.

This means that if 50 do not track iOS Facebook app users click on the Facebook ad to go to your online shop, and 5 ends up buying, Facebook will not have data on which 5 did so. In fact, unlike before it would not know some of the 5 had bought at all.

That’s because Apple now recommends apps use the Apple replacement API (SKAdNetwork) for conversion tracking. It will allow for some conversion data to be passed back “at the campaign level”. At most, it will mean you could measure when a conversion took place right away. In other words: somebody saw the ad clicked and bought right there. Apple will let Facebook know which ad caused a conversion.

But up to now, Facebook, because it could track the device, would also count a conversion, when a person clicked, and then went back to buy up to a week later (A behaviour advertisers consider normal, especially for big-ticket items that require “consideration”.)

While some conversions now won’t show, for those that do, you will no longer see the gender, age range of the buyer, or their approximate location (Facebook gives this information up to the region or city level at present) in Facebook Analytics. The same type of attribution problem will afflict Google’s ad analytics. Facebook says that to compensate it will run some statistical models to try and estimate the true conversion numbers. Be that as it may, one might expect this degraded attribution might put downward pressure on ad prices.

It’s also worth noting that because Amazon’s ads run entirely in their system and link to products also inside their system, their advertising seems to be not impacted at all. And as Benedict Evans points out, Amazon ads could be raking in as much as $20 billion per year (2020), and it is growing really fast (this Q1 results points at as much as a 73% leap! from a year ago) and already be more profitable than their hugely profitable AWS.

“Retargeting”

It would also seem that (the hugely effective/ useful/ annoying) retargeting (where you visit a product online and then see an ad for it inside an iOS app), using IDFA, would no longer be possible either. Remember it’s actually more unusual for people to see a product in an ad and to buy right there and then.

Some commentators think you will still be able to use Facebook for On Facebook retargeting since it can still track users via their Facebook IDs, (Provided of course the product or page the user interacted with originally was not another iOS app with tracking disabled for that user).

But if you read the wording of Apple’s terms for their service, it would seem that if Facebook allowed its IDs to get around the retargeting issue inside their app for the users that declined tracking they would be in contravention of the spirit of what Apple is now trying to achieve. Something Techcrunch confirmed from an Apple spokesperson. Yes, Apple controls only the IDFA system, but they expect other services not to use other means such as hashed emails to track users.

Machine learning “audiences”

Lastly, it is worth considering how this change will impact Facebook and Google’s machine learning-driven audience tools.

Facebook’s Lookalike Audiences have existed since 2013 and it’s incredibly powerful. Give (via the Facebook pixel on the checkout) Facebook a set of users, say people that bought your product in the UK the last 90 days, and ask it to give you 200,000 most similar Facebook and Instagram users in the UK to target with ads. In tests Lookalikes usually blow even a finely tuned manual selection of demographics and interests out of the water. The only downside for advertisers is they don’t know how Facebook’s algorithm selected the audience to target.

Speak to experts in machine learning and they will tell you that the amount of data points a machine learning algorithm has really matters. This is why Facebook’s Lookalikes have worked so well. Since Facebook will no longer be able to link many of its users to their browsing, shopping and app use, one should expect these Lookalike audiences to be less predictive for some users in the future.

This is Part 1, Part 2, Part 3 and Part 4.

Google takes a bite of banking

Thu, 01 Apr 2021 00:00:00 GMT

It long ago had media, retail is now a main course, and the automotive and health sectors are the new entrées. But Marc Andreessen’s dictum that “software will eat the world” has until recently not yet come to banking in the West.

And one of the main reasons for that, ironically, is that banking regulation and onerous national-based capital requirements have made platform companies very wary of entering fractured country banking markets.

However, during Q4 2020, Google announced that it had relaunched its Google Pay app in the US, “making it easier to organise finances and save money.” It will for example feature a powerful - and no doubt lightning-fast - search function, allowing users to see and group transactions in new ways. It will also allow users to crawl their Gmail account, find receipts and read them, to be automatically included in their Pay financial tracking.

Its Pay Tab will emulate features of Venmo, like put multiple people in a chat and let them send and request money from each other. The Insights tab will allow you to connect other bank accounts to get an overview of your finances.

Better UX, but also a bank account

Yet, this is not only a series of UX interventions to make things more convenient, as powerful as this has proven to be. The Explore tab, for example, will now allow users to see personalised deals and discounts (no doubt machine learning-driven).

But more importantly, Google will launch a digital bank account (including savings), called Plex, next year, with a range of US banking partners, including City Bank.

Unlike most current accounts, these will feature no charges, no minimum amounts, but also, for now - no overdrafts. Although Google’s banking partners will provide the current accounts, it will feature minimal partner bank branding. Google will provide the user interface, and the current accounts will all, no matter the bank, carry the same name: Plex.

Aggregate and facade

This is a classic Aggregate and Facade Strategy. In the last quarter of 2020, Google announced no less than 3 initiatives that use this strategy. News Showcase is one such example, an initiative where Google hopes to ease pressure from beleaguered publishers and irritate regulators.

With the Aggregate and Facade strategy, Google uses lighter touch product innovation to leverage Google’s position as a gatekeeper to many points on the Internet. The aim is not to recreate a product or service end-to-end, but to mediate aspects of it, like service discovery. It hands-off core functionality, to not one but several partners, but still gains access to data. With this partnership approach, Alphabet, Google’s parent, also hopes to evade regulatory intervention.

What’s in it for Google?

By partnering with banks, which provide the financial plumbing, Google can hand off local regulatory responsibilities in each jurisdiction.

Google gets access to more valuable payments and current account data, just what an “AI conglomerate” needs. Google already offers a machine-learning-infrastructure-to-enable-lending product to banks.
The Pay App’s improved convenience and usefulness increase users’ likelihood of using it for payments, which are lucrative (the payments market size matches that of advertising).
Google becomes a gateway to banking, a market several times the size of advertising and hosting.

The payment and current account data it obtains will enhance Google’s ability to transform it via machine learning to profile users. This data will conceivably enable Google to:

Improve user experience and convenience across its services;
Increase their advertising products efficacy;
Help partners to do personalised offers inside Plex;
Create risk profiles and provide credit in conjunction with their banking partners.

What’s in it for Banks?

You might wonder. The very real danger here for banks is that they become commoditised back-office providers of financial plumbing and regulatory licences. But Google will probably prove to be an irresistible force. There are several reasons for this.

Current and savings accounts provide banks with the kind of capital regulators now prefer banks to hold in large amounts after the financial crisis.

And Google does not have to take the biggest banks head-on. Although there has been much consolidation, there are still hundreds of smaller local banks worldwide. Many of them struggle with the technical and design know-how to build user-friendly applications and offer the efficiency that machine learning-based data profiling makes possible. Some, like City Bank, are huge but lag the other big national US banks for current accounts.

Remember Google is not partnering with banks for their brands or the distribution that the banks’ remaining branch footprint offers. They are doing it for their licenses to bank in particular territories. Google will take care of service discovery and distribution.

What’s in it for customers?

Free current accounts, with great user experience and convenience.
Machine learning-driven deals and coupons.

Google’s data, coupled with its profiling algorithms, will make it easier and more efficient for merchants to offer deals (coupons) inside Plex, as Google can help them identify who to target. Later, because of superior machine learning risk profiles, it could also make it possible to provide users with cheaper credit, together with their partner banks, than banks outside of Plex can provide on their own.

What about other tech competitors?

Apple has already announced a Credit Card in conjunction with Goldman Sachs.
Amazon is already providing credit to merchants.

Less constrained by regulation at first, Chinese digital platform companies like Alibaba have gone down this route already, even providing credit themselves through Ant Financial. This happened to such an extent that regulators feared for the future of the Chinese state banks, and the stability of their financial system as money was sucked from it to the platforms. The Chinese authorities subsequently mandated that platforms provide loans through banking partners.

The regulatory conundrum

If successful it would be the most significant “Aggregate and Facade” strategies Google has done and allow it to gain a substantial foothold into yet another industry, financial services.

Regulators choice weapon for targetting companies for regulation is through the lens of market power. Economic theory says that market power should not be equated with big is bad. It is only problematic if market prices lead to more expensive prices or less innovative services - as economic theory suggests it would.

Yet the theory of market power would not be the last victim of digital technology. It remains an open question if this development would reduce consumer welfare. It might do the exact opposite, making banking free and credit cheaper for consumers.

Bear in mind, that in the US, a significant portion of the population is still unbanked (5%) and underbanked (17%). In some developing countries, the number of unbanked and underbanked is much higher. One of the main reasons for this is minimum balance fees. On top of that The Bank for International Settlements (BIS) has published research that showed platform companies have access to data that allows them to extend credit, where banks are running blind.

However, if Google also moves into banking, levels of overall concentration, already high, would go much higher. The risk may be considerable for the stability of society as a whole. We need another rubric other than market power to look at this scenario. One not dissimilar to that used when determining media plurality, ie. some bespoke limit on the concentration of power. Yet such a major regulatory intervention needs a good theory as a foundation to provide it with legitimacy.

We do not have one in our toolbox—yet.

Nothing compares to Data

Wed, 31 Mar 2021 00:00:00 GMT

How should we understand data? In a quality, The Information essay, technologist and publisher Tim O’Reilly was the latest to grapple with the question. The metaphors we use for data is wrong he says. And that’s problematic because metaphors are like maps. They can provide answers or lead us astray.

Data is the new oil

O’Reilly argues that the reason we don’t have a good handle on what data is, is illustrated by the misinterpretation of British mathematician Clive Humby’s now popular 2006 metaphor that “Data is the new oil”. We have come to see the quest for data as a “There Will be Blood”-kind of gold rush. We all should claim our share before an information age equivalent of Big Oil monopolises everything.

Taking Facebook profits as a proxy for data generated wealth, O’Reilly illustrates that data is actually not that valuable when the money it generates is redistributed among Facebook’s users. Worldwide he estimates a profit of $7.50 per Facebook user per year. This rises to $11.48 for European users and $36 for North American ones. That’s $3 a month, far less than the meagre cost of a Spotify subscription per month.

“Data is the new sand”

But says O’Reilly, Humly was trying to make a more nuanced point that’s gotten lost. Namely, that data, like oil, is only valuable after it’s been transformed by industrial refining. So now, to get rid of the association with riches, O’Reilly offers us a new metaphor instead of oil: “Data is the new sand”.

Silicon is sand - and the second most common substance on planet earth. It is of little use until enormous techno-industrial processes transform it into something useful that sits at our modern economies’ heart—silicon chips. Without our advanced knowledge and cutting-edge manufacturing techniques, made possible by gigantic amounts of capital, it’s just well, sand. The economic value gets added by layers of hardware, operating systems, technologies, and applications that rely on these chips. Yet nobody asks, is sand valuable?

So is it with data he argues:

“Take Google: It must crawl the entire web—collecting and indexing trillions of webpages, tweets and other data, much of it in real time—and then search the results using complex algorithms and artificial intelligence to answer user questions 3.8 million times per minute. It must route its Google Maps customers to hundreds of millions, if not billions, of destinations, in real time, every day. Each day, it must store and process 4 billion photos and videos on Google Photos and stream 5 billion videos on YouTube. Facebook routes billions of posts to billions of users in personalized news feeds. These services require massive investments in data mining, refining, operations and management, as well as a constant investment in R&D.”

Yet, even if he makes a good case, O’Reilly knows this is not the whole story. Unsatisfied, he immediately goes reaching for another metaphor.

“Data is the new Oxycontin”

There’s are things that neither the metaphor of data as oil or sand captures. And for O’Reilly, it is this: it has a quality where it can both harm or greatly benefit tech companies’ users. He, therefore, likens it to the powerful opioid Oxycontin (also because as an aside, he says data is addictive). If Google knows what you want, then smart targetted advertising can help you. At the same time, pushing disinformation, no matter how engaging, in an election is wrong. What users want is not always good for them. Similarly, the problem with Big Tech companies hoovering up medical information is not one about privacy. This medical data can be to a user’s benefit. Users know this and give it gladly. It can help with treatments and cures. Unless of course it is used to discriminate (say for insurance purposes). So fundamentally, the question O’Reilly poses as the primary one is this: what is the benefits and risks of having control over so much data?

I would argue his metaphor, Oxycontin, as something powerful, that could be used for both good or bad, also does not provide a comprehensive account of the nature of data. Just like data as sand only elides part of the picture: the industrial processes needed to make data useful, AND that it is not of much value to individuals.

Without wishing to be facetious, we are still missing several other significant peculiarities of data in these metaphors: That, not all data is of equal value or even potential value; That aggregate data is more valuable than the sum of its parts is another - it can be an economy of scale and scope; Linked to this is something that’s now becoming apparent. Massive amounts of data is what deep learning algorithms need to perform better; Data is also a non-rival and intangible good, yet unlike many non-rival goods, it is excludable; On top of that data is personal, lots of it is the byproduct of how ordinary people live their information-age lives, and this data can make them transparent and legible to the controller of this data; and, as I have written before, data often is social and spacial, the result of an interaction between people, or represent relationship(s), place(s).

But the thrust of O’Reilly’s concern about data as a question about power is broadly in the right direction. It is, in fact not unlike the question legal scholars have started to ask when grappling with data: Is a novel formulation about power and who gets to exercise it needed? This is an area of inquiry on which legal theorists have just embarked, and so, unsurprisingly they also lack a metaphor for data. But it is quite possible, that data is so unique, so abstract and malleable, yet its reach and impact so large that no single metaphor will do it justice.

No "free" markets in networks

Sat, 13 Feb 2021 00:00:00 GMT

Recently, author Cory Doctorow, weighed in on the debate around free speech on the internet after Facebook and Twitter banned Donald Trump’s accounts, and Apple, Google banned the Parler social network app from their app stores.

Doctorow’s solution to excessive big tech power - a point I agree with - is competition (anti-trust) law.

I begged to differ (screenshot below).

An exchange followed, where I made the case against regulation to introduce competition (breakups), arguing that the US was world leaders in telecom when it had a regulated monopoly — but was as a result of this obsession with competition, falling behind Asia.

The US was at one point in time far ahead of the rest of the world in teledensity - and the cost of telecoms to its consumers was very low. Today it is one of the relatively more expensive and slow places in the world to be connected to the internet. But Doctorow disagreed with this solution.

However, two subsequent pronouncements by prominent tech and telecoms leaders show growing support for my point of view.

Writing in the FT, ex-Google head Eric Schmidt warned that the west would fall behind Asia in 5G roll out, and argued for the suspension of 5G auctions, for government subsidies, and for infrastructure rollout obligations.

“The massive sums winners paid for the spectrum will reduce their financial capacity to actually use it. Instead, it will probably result in disinvestment and downsizing.

At the same time, $81bn is merely a trifle on a government scale, equivalent to less than a month of US debt issuance and the money is unlikely to be spent on the 5G network the country needs. The outcomes are predictable: Americans will face higher prices and weaker digital services — yesterday’s internet tomorrow.”

What Schmidt does not expressly say (it’s still not fashionable in the west), but what is implied, is that this will entail less competition. That’s part of the profitability and investment equation.

And a few days ago, Alvarez-Palette, the head of Telefonica said much the same, namely that “regulators should stop treating spectrum auctions as an opportunity to raise money and renew competition policies to allow consolidation and innovation in the market.”

“This is a structural error and a new industrial approach is needed,” he demanded that, “the EU must better align competition policy and industrial policy if the EU’s industrial strategy is to succeed”. If not, Álvarez-Pallete believes it will become increasingly difficult for the sector to make the necessary investments.

For Schmidt, the urgency does come from competition, but not between western companies, but rather between the west and particularly China:

“China will soon have a national network with speeds of 1 gigabit a second. With China’s head start, the next generation of technology giants — and the products and services they build — are not going to be European or American but Chinese.

About a year ago Schmidt made a startling point with respect to the race in Artificial Intelligence when he argued that the US business can not compete with China if the US government does not assist them:

“Success matters for more than our companies’ bottom lines and our military’s battlefield edge. We must show that these new technologies can advance individual liberty and strengthen free societies. For the American model to win, the American government must lead.”

What if here-comes-everybody is the problem?

Sat, 30 Jan 2021 00:00:00 GMT

“Give me the liberty to know, to utter, and to argue freely according to conscience, above all liberties.

Though all the winds of doctrine were let loose to play upon the earth, so Truth be in the field, we do injuriously, by licensing and prohibiting, to misdoubt her strength. Let her and Falsehood grapple; who ever knew Truth put to the worse, in a free and open encounter?”

John Milton - Areopagitica.

“True’ and ‘false’ are attributes of speech, not of things. And where speech is not, there is neither ‘truth’ nor ‘falsehood.”

“Hell is truth seen too late.”

― Thomas Hobbes**, Leviathan**

There’s been lots of handwringing and head-scratching about Trump — and the devoted following he has, which seems impervious to reason. No amount of pussy-grabbing, tax evasion or courtroom defeat of election allegations stick to Teflon Trump. The populist crusade to “drain the swamp” remains intact.

And now we have Reddit’s WallStreetBets subreddit, a forum that started the pump of GameStock shares, in the face of market fundamentals. Again this is being cast as a revolt against elites. Where in the former, Trump’s populism seems to have sullied our marketplace of ideas, WallStreetBets, is messing with the efficient market hypothesis*.

It is a moment impervious to reason, good taste and sense.

I’ve seen several explanations for it:

Social media algorithms, primed to maximise ad revenue, incentivise engagement - that is to say aggressive, sensational, simplistic content, and creates “filter bubbles” of opposing tribes that receive very little contradictory information in their media diet.
Related to this, the moderating influence of mainstream media is waning. They are struggling to survive - because social media is eating all their ad revenue.
There’s the widely held opinion that the problem is of a society where many people at the bottom have been left behind. The 99%, whose income has stagnated for decades, while the elite has flourished, are craving recognition and dignity. See Dignity - Seeking respect in backrow America.
There’s the theory that the problem is partly with the people at the top, namely Peter Turchin’s theory of “elite overproduction”: “a society which is producing too many potential elite-members relative to its ability to absorb them into the power structure. This, is a cause for social instability as those left out of power feel aggrieved by their low status.” Think Rudi Guliani.
That state actors with bad intent is sullying our public sphere. I even heard the editor of TechCrunch Europe, Mike Butcher, say that state actors might have been involved in the Gamestop bull run.

My own theory is not novel. In 2014 Martin Gurri posited that: “Technology has categorically reversed the information balance of power between the public and the elites who manage the great hierarchical institutions of the industrial age—government, political parties, the media.”

Gurri wonders whether we will figure out new forms of institutions to do what our old ones did in liberal democracies. And so restore the balance of power. But what if our dreams for liberal democracy - and the current reality that new technologies create - are incompatible? When the internet arrived it was explicitly hailed as democratising. That it will bring us closer to ideals of freedom, personal expression, freer information and a freer market - an idealisation of elements of liberal democracy. And this would all be an unqualified good.

There’s a much earlier book by Noam Chomsky and Edward Herman, which - also makes an appeal to our sense that democratising our media sphere will make our societies better, more progressive, juster. They argue, mainstream media, helped by technological constraints and the economies it created, inhibit complete democratisation of media in favour of elite and commercial interests. I’m referring to the book Manufacturing Consent.

Chomsky and Herman argue, not unlike Gurri later, that mainstream media played a specific - if unspoken - role in that they: “are effective and powerful ideological institutions that carry out a system-supportive propaganda function, by reliance on market forces, internalized assumptions, and self-censorship, and without overt coercion.”

But what if Chomsky, Herman are right about mainstream media protecting elite interests and the system, but wrong about the beneficial effects of democratisation, also for the downtrodden? What if reaching our democratic dream and giving everybody the possibility to publish has unintended consequences? Consequences like opening the Overton window so wide, that conflict and chaos ensue? What if the good society requires that not everybody has a voice?

Rebecca Mckinnon posited in 2011 that China practises what she calls Networked Authoritarianism. Instead of the Internet threatening the regime, it is used by it, to monitor opinion, something key to successful governing and that it would struggle to do otherwise. It also uses it to shape opinion. They do so through a mixture of surveillance, filtering, censorship, but also through armies of agents participating in and steering the conversation. Counterintuitively it has increased the regime’s legitimacy.

Like mainstream media’s unspoken “system supportive propaganda function” in pre-internet democracies, the Chinese have figured out how to use social media to support the system. Could democracies do that and still call themselves that?

* -“stocks always trade at their fair value on exchanges, making it impossible for investors to purchase undervalued stocks or sell stocks for inflated prices.”

The arbitrary overlords

Tue, 12 Jan 2021 00:00:00 GMT

Although action against Trump’s incendiary, but particularly his false language - and Parler, Gab’s distribution of it - was long overdue, we should all be concerned about the ousting’s significance.

It underlines where media power now lies. It will emasculate Trump’s immense impact on the public sphere. Couched by Big Tech in the language of terms and conditions of use broken - rules long exceeded by Trump - the action was up to the final say-so of a few Big Tech companies’ chief executives.

Clearly, Facebook in particular, only took action when power had already half exited the room. These executives are chiefly accountable to the market. In some cases, they came under pressure from employees. You have to wonder about the variable treatment other world leaders will get if they did similarly? My guess would turn on how lucrative the market, their employees’ level of interest in the country, and the leader’s prospects imminently losing power. In that sense, the decisions were arbitrary, if not cowardly and self-serving.

In the tech world, long beholden to The California Ideology (a mix of socially progressive and economically conservative thinking) - the banishing is a particular anathema to ideas held dear. Although there have been many instances to prove the contrary now holds, for many in tech, information still “wants to be free”, and ideas, just like the original architecture of the Internet, are supposed to “detect censorship as damage and route around it.”

The problem with these technological determinist out-of-control arguments is that it turns out that technology is only out of control if you let it be so. China has a thriving “Splinternet” behind a censorious firewall, some cryptocurrencies exist, increasingly regulated - but only within those jurisdictions in which they are tolerated. When rightwing dilettante Milo Yiannopoulos was banned from Twitter and YouTube, he disappeared from view and lost all power.

But there is also a normative passion to these libertarian arguments that go beyond what tech allows us to do — that it is morally wrong to censor, which is why some not so long ago still defended the existence “upskirt” subreddits on Reddit. Yet child pornography was the one hill free speech absolutists never chose to die on, and this small concession turns out to be fundamental — it sinks their whole argument.

No right is absolute then, even free speech, it ends where it crashes into others’ rights. These rights’ boundaries are not fixed, and continually needs to be renegotiated, and disputes settled. The pressing question is then not the utopian one whether we need censorship — but rather who is best to draw these lines, enforce them and who is best to settle disputes?

It's not (only) your data

Thu, 17 Sep 2020 00:00:00 GMT

“Why do you think people say they highly value data privacy, yet freely offer up their data to companies that request it? Do you think people would value their data differently if they were given ownership of it?”

Are we all cynics, simply impulsive, short-sighted or dupes!? And would an ownership right, where we can decide what to do with our data - even sell it - fix the problem?

I’ve been doing a course at the LSE, Data: Law, Policy and Regulation, where this question was asked.

I want to argue that most of us can see the benefits of using services like Facebook have for us, while the risks to our privacy and dignity seem small or hard to quantify - and for good reason.

While many claim a visceral fear of a creeping loss of privacy, so far actual events have not supported this angst. Take the totemic legal case concerning online behavioural advertising (OBA), and Real Time Bidding (RTB) in online advertising auctions currently before European (Irish) and UK data protection authorities.

The complainant, Johny Ryan of Brave wrote about the seriousness of the complaints:

“RTB auctions happen hundreds of billions of times every day, and leak the online habits of billions of Internet users into the data broker ecosystem. We believe is the largest data breach ever recorded.”

Yet to date no person or institution has come forward with specific claims that this “largest data breach ever” has caused them suffering or harm. Of course, the lack of known victims does not mean that the data is not valuable. It is - for starters - the basis from which advertisers can bid for our attention and smaller media companies and bloggers, who sell our attention, make money.

But currently, in terms of potential harm, in most* scenarios to most actors (bar Google and states) the format this RTB data comes in is not of much use besides the ability of disparate advertisers to use it to bid for advertising. Yes, advertising can be annoying. It can also be harmful for the individual, but particularly nefarious on aggregate for society, as the Cambridge Analytica case showed.

That’s one of the strange things about data. Not all data is the same. An email address is useful in ways that a first name is not. One piece of data, even very useful ones like email address is not of much use even if you have it for everybody (without context that makes its use valuable - it’s still what marketers call “unqualified”).

And even if you have a range of data about a single person it is rarely valuable to persons or institutions that do not have a lot of other similar data. Machine learning, a technology that finds statistical correlations in data, amplifies this effect where the more relatable data you have the more valuable it becomes.

So I agree with David Priest, “data is slippery”. Data’s value and power depend on the context. And “it has different values for individuals, companies and society writ large.”

The so-far victimless OAB case above illustrates just one example, and it makes me wonder over the appropriateness of the often-heard call to giving people a property right over their personal data. So for example, who should own the data that machine learning creates as statistically significant, after evaluating personal data, together with that of other individuals’ and publicly available data?

I also agree with Professor Mark MacCarthy that data has other characteristics that make it an inappropriate candidate for ownership rights, the most important of which is that humans are social, and data embodies our societies.

So says MacCarthy, that if we had data property rights a lot of it would have to be co-owned, say for example a recorded conversation. I’d go one step further, your data’s value is socially constructed. It is dependent on your position in society and who you interact with. In other words your location in the social graph (network). It would be worthless without you being embedded in a particular place in a particular society.

Personal Data is collective data simultaneously. As sociologist Zeynep Tufekci recently argued when it became clear that Strava revealed the location of American military bases:

“Data privacy is not like a consumer good, where you click “I accept” and all is well. Data privacy is more like air quality or safe drinking water, a public good that cannot be effectively regulated by trusting in the wisdom of millions of individual choices. A more collective response is needed.”

And then there is imperative of having strong successful economies and societies. In 2019, economist Jason Furman did a report for the UK on how to create more competition in the digital economy. He recommended that data of the large tech platforms should be shared, to encourage smaller competitors, and to enhance how the state delivers its services.

We do need a hybrid model when it comes to the protection of data, that combines the right to privacy with something else. But what that something else should be, looks more like data as a public good than data that is personally owned. Yes, you may have some claim to the data you generate, but so does society, for which it is of much more use.

* Addendum: Since I published this post something happened that made me revise my opinion somewhat. The New York Times received a data leak from a commercial database of the January 6 2021 riot in Washington. It showed that many of the rioters’ routes to the event could be traced in this database, and in numerous cases, they could be identified. These databases are created mainly by using device identifiers used for advertising targeting.

Why has this not happened in Europe? Probably because unlike the US, where there is little legal constraint on doing this kind of data gathering, the GDPR makes commercial trackers more cautious.

Does this change the main trust of my argument that we need a hybrid model when it comes to the protection of data, that combines the right to privacy with something else? No it does not.

Society actually exists, says Boris Johnson

Fri, 10 Apr 2020 00:00:00 GMT

Even in a deeply divided post-Brexit Britain, news that prime minister Boris Johnson had been taken to intensive care was greeted with near-universal disquiet, if not alarm. Much like his hair, Johnson cuts a jovial and bewildering political figure. A popular mayor of liberal cosmopolitan London, he famously wrote two newspaper articles, one pro and one anti-Brexit, to help him decide. An opportunist then, not an ideologue. But one with a penchant for cracking deeply offensive jokes.

On March 16 this year, as the seriousness of the Covid-19 crisis started to dawn on his government, Johnson hastily had a conference call with CEO’s of large British manufacturers to see if they could manufacture life-saving ventilators. The UK had only a quarter of the number of ventilators Germany had, and less than Italy, whose health system was already overwhelmed. Executives were stunned when he quipped that “Operation Last Gasp” was the purpose of the call.

That particular day was a red-hot news day. Professor in epidemiology Neil Ferguson of Imperial College, who sat on three scientific committees that informed the UK government emergency response, had published a paper. It predicted 510,000 deaths in the UK if nothing was done to mitigate the spread of the virus. More alarmingly, the paper indicated that the current government policy would probably mean approximately 250,000 deaths — and recommend far more stringent social distancing measures.

Only four days before, on March 12, the government’s chief medical officer, Chris Whitty, himself an epidemiologist, cheerily announced that the government would no longer attempt to trace - and therefore test - every suspected Covid-19 infection and their contacts. It was no longer necessary, Whitty said because the government was moving from a strategy of “containment” to one of slowing the virus’s inevitable spread. He claimed several weeks would still pass before the infection peaked in the population.

A day later, the government’s chief scientist, sir Patrick Vallance calmly elaborated in a TV interview. Talking down “eye-catching” measures, like banning large public events, the aim was now to stall the spread and flatten the peak of the infection. Most people need not fear the virus he assured. And when a large proportion of the population eventually becomes infected, they will have built “herd immunity”. What had already been the unspoken government policy — not to attempt to stop the virus’s spread but slow its progress had become public.

But by now, a large part of the UK’s media class had been following events in China, Italy and Spain for some time via social media. Stories like that by the New York Times on how China threw the kitchen sink at the virus and papers by the Lancet with clinical analysis of the viruses’ grim impact in Chinese hospital wards had already been shared widely. A British blogger in Spain translated frightening Madrid hospital statistics. Tragic accounts of “super-spreaders” at Italian football matches and Basque funerals escalated concern.

Amateurs quickly learned the basics of epidemic curves. Rough growth estimates were swiftly plotted and posted on Twitter. Using the 100th reported case as a benchmark, they concluded that, unlike Whitty’s assertion, the United Kingdom was only days behind Italy when it locked down. Venues started to close of their own accord. Alarmed at the lack of urgency, journalists began to ask more pointed questions of Johnson and his officials at press conferences.

In the aftermath of Ferguson’s bombshell paper, the government caved in. Regarding the number of deaths, the Ferguson paper released on the 16th did not differ significantly from those already used by government models since the beginning of February. Yet the airing of these numbers and the public demand for greater social distancing led to a swift, but still unacknowledged, change in government policy under Johnson.

Johnson has dodged questions about his government’s initial plans for “herd immunity” and the apparent acceptance of a vast number of deaths ever since. Strict social distancing measures and, eventually, a lockdown were always the plan, they now say. These would indeed be introduced as soon as necessary. But the problem with this line of argument was that once it was policy to keep deaths to a minimum, it was not entirely clear whether there was any good reason not to order a lockdown immediately. Even from the vantage point of cold economic calculus, an early lockdown looked preferable.

During the Brexit tumult, Johnson and fellow Brexiteers regularly pooh-poohed the advice of experts. But now Johnstone felt justified in saying he was following the science and experts. A subsequent investigation by Reuters found that Johnson failed to query the assumptions his scientists had made. But also that his scientific advisors had failed to clearly articulate the number of potential deaths and the full range of options available to him, especially early on.

John Edmunds, a professor of infectious disease modelling from the London School of Hygiene and Tropical Medicine, sat on the same government advisory panels as Ferguson. In the Reuters investigation Edmunds alleged that advisors never suggested more stringent social distancing measures before Ferguson released his paper on the 16th. And this was because they second-guessed the government. They never thought locking down would be politically acceptable.

Edmunds claimed it was only when Italy implemented a national shutdown on March 10, that the “policy space” had been “opened” for similar measures to be discussed in the United Kingdom. Why it was thought to be politically impossible up to that stage remains a subject of intense speculation. Was it because scientists thought Britons prized their freedom, as Edmunds told Reuters? Or was it because they thought government could not countenance the economic cost of a shutdown?

Only three months before, in December 2019, Johnson won his first election with a historic majority on the simple promise to “get Brexit done”. This was despite dire predictions from most economists and the Bank of England of the damage it would do to the British economy. The opposition to Brexit was fierce, from a broad and influential coalition of London-based businesses, the legal fraternity, academia and many on the left as well as some Conservatives.

Johnson, advised by the enigmatic but cunning strategist Dominic Cummings presented Brexit as a struggle between the ordinary people against an out-of-touch elite. As a fallback position, a hitherto unthinkable argument was raised. Even if the experts were right about the economy, why put it above British national sovereignty? Homo economicus was living on borrowed time in Britain.

This Brexit struggle was not just a gentlemanly battle of ideas. Johnson showed he had a ruthless side. To marshal the full support of the fragmented Tory parliamentary party, even venerable senior members of Margaret Thatcher’s cabinet, like the former Chancellor of the Exchequer Kenneth Clark, and a grandson of Winston Churchill, Nicholas Soames, along with twenty-one other members of parliament, were cast aside and expelled from the party.

Thatcher, who had famously led Britain into the European single market, is still a deeply polarising figure in British politics. The speech opening her campaign to join hailed “A single market without barriers—visible or invisible—giving you direct and unhindered access to the purchasing power of over 300 million of the world’s wealthiest and most prosperous people.”

When in 1987, she told an interviewer that “there is no such thing as society, only individuals and their families”, it was seen as a shocking but fitting description of her and the Tory Parties’ political philosophy that greed and selfishness are good. Yet Thatcher’s embrace of the single market was ideologically coherent. If we are all individuals, and society a mirage, then borders are a pointless hindrance.

To break the opposition to Brexit in the legislative chamber, Cummings honed the campaign message to the populist “the people versus parliament”, which Johnson embraced with enthusiasm. He further garnished it with words like “betrayal” and “surrender”, escalating the rhetoric. In an act of staggering callousness, he casually claimed that to “get Brexit done” would bring honour to the murdered Labour MP Jo Cox. Cox opposed Brexit in a constituency that supported it narrowly. One week before the Brexit vote, a lone attacker brutally shot and stabbed her to death on a public street. At his trial, Cox’s attacker, when asked to give his name, answered: “Death to traitors. Freedom for Britain.”

Before Johnson’s triumphant 2019 election, conservative think tanks pinpointed the voters that mattered most to get a clear majority. Several London seats fell in 2017 to Jeremy Corbyn’s Labour Party, and polls showed that more was likely to do so in 2019. To win, the Conservatives had to break the so-called red wall and win the votes of former working-class Labour voters in the North of the country. These were voters who had voted for Brexit. Johnson’s message to them was simple but effective. Cosseted London had ignored them for too long. His government would turn away from Europe and towards them, “levelling up” their regions with London and showering the neglected but beloved NHS with investment.

Not many people, including the Labour Party, took the Conservative Party’s open assault on Labour’s Northern red wall seriously. Did Johnson, a posh Eton-educated Tory believe what he was saying, and more importantly, would these voters? He had, after all, spent a decade in the senior echelons of the Conservative government, a government which had cut spending on welfare and the NHS and had neglected the North of the country.

On January 13 2020, the day Thailand confirmed the first Covid case outside of China, one of the government advisory committees Ferguson and Edmunds sit on met for the first time to discuss the threat of a new coronavirus in China. The committee did not advise raising the threat level, claiming the data coming out of China was confusing. The previous day, to speed the global development of diagnostic tests, the Chinese had made the genetic sequence of the virus available worldwide via the World Health Organisation. On the 15th and the 20th of January, Japan and South Korea, respectively, identified their first cases. By the end of January, it was clear to Johnson’s advisors that the virus had “pandemic potential”, yet the threat level was advised to be kept on “low”.

On January 31 this year, when Hubei had already been under lockdown for more than a week, Italy reported its first case of the virus, declared a state of emergency and suspended all flights from Wuhan. That night Johnson gave a televised address. “This is the moment when the dawn breaks and the curtain goes up on a new act in our great national drama.” he intoned. It was the start of a conciliatory and optimistic speech about Brexit, that did not mention the virus or indeed Brexit once, the word having been banned from government communications. Instead, it focussed on the possibilities for Britain now that it had “recaptured” sovereignty from Brussels.

Only seven weeks later on March 23, more or less on the timeframe the amateur social media epidemiologists had predicted, Johnson placed the United Kingdom under lockdown. It was only seven days after Whitty claimed the peak was still weeks away. Three days later, having been on the back foot because of his government’s shambolic pandemic policies for weeks, Johnson faced the further embarrassment of coming down with Covid-19 symptoms himself. As deaths and criticism continued to mount, he was hospitalised a week later on April 5. Still, people were shocked when the following afternoon, he was shifted to an Intensive Care Unit himself.

On the day he discovered his symptoms, Johnson had joined a nationwide clapping session to thank the NHS’s workers. But perhaps the most abiding image of Johnson in the short period of his premiership so far, was a remark uttered in a video broadcast, self-made on his phone in Covid isolation. Its purpose was to thank the 20,000 ex-NHS workers who had answered his call to return to the service to help, as well as the 750,000 people who had volunteered to help do things like distribute food packages to the vulnerable. This proves, Johnson said, that “there really is such a thing as society.”

This essay first appeared in Afrikaans in the Vrye Weekblad.

Musings on the role of the state in a digital world

Wed, 17 Jul 2019 00:00:00 GMT

Welcome to Leviathan Newsletter by me, Wessel van Rensburg.

In the mean time, tell your friends!