A.I. and the Crisis of Capitalism: A Historical Approach - John Cassidy

ELI5 / TLDR

Every time capitalism gets a powerful new technology, the same fight breaks out: who gets the gains, who eats the losses, and can the political system hold it together? John Cassidy, the New Yorker’s economics writer, walks through 250 years of that pattern — from the Luddites smashing looms to Marx marveling at capitalism’s productive power to Keynes promising a three-hour workday — and lands on a simple, uncomfortable observation: AI is replaying all of it, except faster, and this time the professional middle class is on the firing line instead of factory workers.

The Full Story

The Question AI Reopens

Cassidy frames AI not as a standalone revolution but as the latest wave of a digital revolution that’s been running since the 1980s. What makes it different from earlier waves — internal corporate networks, the public internet — is the range of old capitalist questions it drags back into the open. Job displacement. The labor share of income. Monopoly versus competition. The relationship between capitalism and democracy. These aren’t new questions. They were first asked in the 1770s.

Adam Smith Was Not Who You Think

The talk begins at Cromford Mill, 1771, the first power-driven cotton mill in Derbyshire. Five years later, Adam Smith published The Wealth of Nations. Smith was optimistic about technology — machines that made labor “much facilitated and abridged” — but deeply skeptical of big business. He wrote the book partly as an attack on the East India Company, a royally sanctioned monopoly that had just needed a parliamentary bailout after managing to lose money despite its trade monopoly east of the Cape of Good Hope.

Smith thought joint stock companies shouldn’t be allowed to exist.

The parallel to today’s AI giants — massive corporations burning through capital and coming to governments for favors — is left for the audience to draw.

The Logic of the Luddites

The first industrial revolution displaced skilled artisans: hand-loom weavers, stocking knitters, wool croppers. These weren’t anti-modern idiots. They petitioned Parliament first, citing centuries-old statutes that had protected their trades since Charles II. When Parliament ignored them, they turned to violence. One letter to a factory owner read:

“Remember we’ve given you fair warning that if your factory burns it’s your own fault. It’s not our desire to do you the least injury, but we are fully determined to destroy both your machines and your steam lumps.”

The state response was repression — more British soldiers were deployed against the Luddites than were fighting Napoleon in the Peninsular War. Ringleaders were hanged. Others were shipped to penal colonies. But the core issue wasn’t irrationality. It was legitimacy. The Luddites saw laissez-faire capitalism as what E.P. Thompson called “a foul imposition” — the shredding of a centuries-old social contract.

Marx Saw the Productive Power First

Cassidy makes a point that often gets lost: Marx and Engels were in awe of capitalism’s output before they condemned its distribution. The Communist Manifesto’s famous passage about “Egyptian pyramids, Roman aqueducts, and Gothic cathedrals” is essentially a love letter to industrial productivity:

“What earlier century had even a presentiment that such productive forces slumbered in the lap of social labor?”

The problem was never the technology. It was that wages stayed flat while productivity soared — a period economic historians now call “Engels’ Pause,” roughly 1800 to 1850. Marx saw technology driving monopoly, intensifying work rather than shortening it, and reducing the individual laborer to insignificance beside what he called “a monstrous automaton.”

Why Capitalism Survived Anyway

The Marxist prediction of revolution didn’t materialize in industrial societies because, eventually, wages did start rising. The revolutionary impulse depended on absolute immiseration, and that condition lifted — partly because of the labor movement, partly because technology became more labor-biased. The debate about why is still live among economic historians.

What emerged instead was the crisis of imperialism. John Hobson, a British liberal journalist, argued in 1902 that capitalism’s enormous productivity created surpluses that couldn’t be absorbed domestically, driving imperial expansion to find new markets and investment opportunities:

“The modern policy of Great Britain and other great powers is primarily a struggle for profitable markets for investments.”

Lenin and Rosa Luxemburg picked up this framework. The first half of the twentieth century — two world wars, the Great Depression, the rise of fascism — can be read as these contradictions playing out at civilizational scale.

The Social Democratic Fix

The post-war settlement was essentially a three-legged stool: Keynesian demand management keeping unemployment low (which gave workers bargaining power), embedding capitalism in inclusive institutions (Karl Polanyi’s insight), and collective bargaining enshrined in law (the Wagner Act, the Treaty of Detroit in 1950). This system worked for roughly thirty years.

It broke down in the 1970s. Left Keynesians like Joan Robinson and Michael Kalecki had predicted this — not because of inflation, which was the proximate cause, but because the underlying social bargain would eventually eat into profits, and capitalists wouldn’t tolerate that forever. Profit rates were indeed declining on both sides of the Atlantic by the late 1960s.

Two Forces Since 1980

Cassidy sees the last forty-five years as two parallel forces: a political counterrevolution against social democracy (Thatcher, Reagan, the dismantling of labor power) and a second industrial revolution based on digitization. Three waves: corporate networks in the 1980s enabling globalization, the public internet from the mid-1990s, and now AI. The populist backlash from both left and right — Bernie Sanders and Donald Trump both ran as critics of global capitalism in 2016 — is the political expression of these combined disruptions.

What Economists Are Saying Now

Five years ago, most economists could fit AI into the standard framework: a new general-purpose technology, like steam or electricity, that causes short-term disruption but eventually creates new industries and raises wages. That optimism is fading.

David Autor at MIT, probably the leading labor economist of his generation, still sees opportunity alongside risk. But Daron Acemoglu, the Nobel-winning economist, has turned darker:

“None of the big companies are pouring even a small fraction of their investment into developing AI as a pro-human, pro-worker tool.”

The logic is straightforward. AI companies need to justify trillions in capital investment. The fastest path to profit is selling labor-replacement tools to corporations. The market is steering AI toward substitution, not complement.

The Reform Menu

Centrist economists have proposed several interventions. Autor and Acemoglu want to restructure the tax code — currently riddled with capital subsidies while labor gets almost none — to level the playing field. They also point out that healthcare and education are heading toward 30% of US GDP, with government as the biggest purchaser, giving it enormous leverage to shape what AI companies build.

Private-sector mechanisms exist too. The Anthropic settlement offering authors compensation for training data. The Writers Guild deal in Hollywood, which insists AI cannot legally be credited as a creator. Unemployment insurance and retraining for displaced workers. Aaron Dube, the labor economist, supports temporary wage subsidies.

The Bigger Problem

Cassidy finds these solutions reasonable but possibly insufficient. If the tech companies are right about the trajectory — and Dario Amodei of Anthropic is among those predicting full labor substitutability once robotics catches up — then several crises hit simultaneously. Mass job displacement. A fiscal crisis, because the tax base depends on labor income. A distributional crisis, as the capital-labor income split (historically 70-30, already shifted to 60-40) could tip further.

Amodei himself has written:

“At that point, our current economic setup will no longer make any sense. There will be a need for a broader societal conversation about how the economy should be organized.”

And then he adds: “I suspect that some new and stranger thing will be needed, and that it’s something no one today has done a good job of envisaging.”

The Promise That Keeps Not Arriving

Cassidy ends with the visions of abundance that keep recurring across two centuries. Marx imagined a world where people could “hunt in the morning, fish in the afternoon, rear cattle in the evening, and criticize after dinner.” Keynes, in 1930, predicted that within a hundred years people would work three hours a day and leave behind the “pathological propensities” needed for capital accumulation. AI might finally deliver the technological conditions for that abundance. But at the moment, we’re leaving everything to the market — to exactly the kind of joint-stock companies Adam Smith warned about 250 years ago.

Claude’s Take

This is a genuinely excellent lecture. Cassidy is doing something that sounds simple but almost nobody does well: he’s using 250 years of economic history to make the AI debate legible to people who aren’t steeped in either tech or economics. The Luddite section alone is worth the watch — the idea that they had a coherent theory of legitimacy, not just rage, is the kind of reframing that changes how you think about current protests against tech disruption.

The most valuable move here is the refusal to predict. Cassidy is honest that he doesn’t know whether AI is the real deal or partly a bubble (an audience member presses him on the trillion-dollar investment question with no clear returns). He explicitly says he’s “presupposing it’s a big deal” for the purpose of the talk. That intellectual honesty, combined with deep historical knowledge, makes the analysis far more trustworthy than the average hot take from either techno-optimists or doomers.

The weak spot is the solutions section, which Cassidy himself acknowledges. Tax reform and procurement leverage are sensible but feel like bringing a garden hose to a potential wildfire. If even Dario Amodei is saying the current economic setup “will no longer make any sense,” then the centrist reform menu may be a category error. But Cassidy’s point is that we’ve been here before — not with this exact technology, but with this exact pattern of productive power outrunning social institutions — and the solutions that worked (Keynesianism, collective bargaining, social democracy) were also unimaginable before they were constructed.

Score: 8/10. Rich historical sweep, honest about uncertainty, genuinely illuminating on the Luddites and Marx. Loses a point for the Q&A section running a bit thin and the solutions remaining necessarily vague. But this is exactly the kind of long-view thinking that’s missing from most AI coverage.

Further Reading

• John Cassidy, Capitalism and Its Critics (2025) — the book this lecture draws from, covering the full arc from the industrial revolution to AI
• E.P. Thompson, The Making of the English Working Class — the classic on Luddite logic and early working-class consciousness
• Karl Polanyi, The Great Transformation — the idea of “embedding” capitalism in social institutions, written at Bennington College during WWII
• John Hobson, Imperialism: A Study (1902) — the liberal case that capitalism’s surplus productivity drives imperial expansion
• Daron Acemoglu & David Autor — their recent work on AI, labor substitution, and tax code reform
• Dario Amodei, “Machines of Loving Grace” — Anthropic CEO’s essay on AI abundance and the limits of current economic arrangements
• Keynes, “Economic Possibilities for Our Grandchildren” (1930) — the three-hour workday prediction, still unfulfilled