Something is coming apart in the rich democracies, and it is coming apart quietly. There are no tanks in the streets, no burning parliaments, only a slow seepage, the way damp climbs a wall. People wake up to find that the machines have changed again overnight, that the town they grew up in has become a place where things used to happen, and that the old promise (grow the economy and prosperity will trickle down to everyone, along with freedom and a little dignity) has turned out to be a story told to children at bedtime.
In a long conversation with Aaron Bastani on Downstream, the Nobel laureate Daron Acemoglu set out the argument of his book, *What Happened to Liberal Democracy: Remaking a Politics of Shared Prosperity*. His answer is not the fashionable one. It isn't only the algorithms, he says, and it isn't only the provocateurs of the right. The rot is structural. Technology was pointed the wrong way, whole regional worlds were dismantled, and the elites stopped doing the patient, unglamorous work of building coalitions from below.
Begin with the machines, because that is where the men who own the future would like us to begin.
Since the 1950s, artificial intelligence has been haunted by a single dream: to build a mind that can do what we do, and then, naturally, to do it instead of us. Today the dream has its prophets. Sam Altman, Elon Musk and Dario Amodei speak of artificial general intelligence in the cadence of scripture, warning of existential risk in one breath and raising billions in the next.
Acemoglu's point is simple and unsettling. Beneath the apocalyptic language lies an older ambition, the replacement of labour, and it pulls capital and brilliant minds away from tools that might make an ordinary nurse, teacher or electrician better at what they already do. "I think the AI field has been so influenced by science fiction from its inception," he says, "and by willing it we're making it closer to reality."
The fantasy writes itself into the world, and it travels a familiar road: from IBM's mainframes to Microsoft and Oracle, from Google and Facebook to the frontier labs of today, power narrowing at every bend, with little in the way of antitrust, unions or democratic oversight to slow it down. What that road delivers, in his account, is wage compression, deepening regional inequality, and productivity gains that turn out, for all the trumpeting, to be oddly thin. What we might have had, and might still have, is pro-worker AI: machines that sharpen human knowledge instead of hollowing it out.
The cost of pointing technology the wrong way is not only measured in wages. It is measured in places.
When manufacturing collapsed in the late twentieth century, it did not merely take the jobs. It took the rooms in which different kinds of people used to meet. A company like General Motors, with its sprawling web of regional suppliers, was a kind of accidental commons. The engineer and the assembly-line worker, the accountant and the woman in personnel, worked under one roof, lived on the same streets, sent their children to the same schools and argued in the same union halls. Wages rose with productivity. Money flowed into mid-sized towns. Trust, that most fragile of public goods, was built daily across the lines of class.
Then came offshoring, outsourcing and automation, and the corporation that could grow its revenues while shedding its people like old skin. Drawing on the work of David Autor, Acemoglu notes that cities like New York, Boston and Chicago once paid a premium to workers without degrees. They were places you went to rise. Now they punish you for arriving: the rent devours the wage, and the work on offer is the low-paid servicing of other people's comfort. Meanwhile the small towns, stripped of their industry, sink into a slow civic depression.
The result is not just inequality. It is a map. The country splits along its highways, and the split becomes political.
Into that wound, Acemoglu argues, the progressive elite poured salt, believing it was pouring medicine. He distinguishes the old, brutal social engineering of the totalitarian state from a softer variety that needs no secret police. This one is administered by universities, activist networks and managerial classes, and enforced not by prisons but by a diffuse, informal policing of speech he calls, rather wonderfully, "the wiki."
Its intentions were often decent; it set out to confront real and historical injustice. But it came from above, and things that come from above tend to be received as weather rather than as argument. "You have to start from a bottom-up process of building compromises rather than a top-down imposition of values," he says, "because otherwise you're going to completely contradict the commitment to self-government."
The working-class communities that had once been the backbone of the left felt lectured, corrected and quietly despised. Waiting in the wings were the right-wing showmen, the media entrepreneurs and the algorithms that feed on resentment the way fire feeds on dry grass. They harvested the anger. The elites, alarmed, policed harder. The anger grew. The spiral tightened, and both sides came out of it angrier and more entrenched than when they began.
While the West was absorbed in this quarrel with itself, it was misreading its rival. China, we are told, cannot truly innovate, because creative destruction, the anarchic churn of upstarts toppling incumbents, does not thrive under the gaze of the state. Perhaps. But Acemoglu suggests this is the wrong question. China's strength is not invention from the ground up. It is execution: engineering, scaling, optimising, building.
Meanwhile the West made one miscalculation after another. It opened its markets to subsidised Chinese manufacturing and took down its own defences. It let technical education, vocational training and physical infrastructure decay. It imported engineers rather than training its own, and so never had to look at the gaps in its schools. And it poured its capital into the cathedral of AGI rather than the workshop.
The United States may lead in large language models by somewhere between six and sixteen months. China leads where it counts in the physical world: robotics, logistics, factories in which AI is not a chatbot but a nervous system. Firms like BYD and Huawei can employ on the order of a hundred thousand people in research and development, a standing army of engineers that ought to make the West blush.
Acemoglu's most delicate argument goes deeper than industrial policy. It concerns the idea at the heart of liberalism itself. Universalism, the conviction that every person has equal moral worth and universal rights, is one of liberal thought's great achievements. But pushed far enough, he warns, it can dissolve the ground it stands on. When it hardens into an ideology that treats, say, open borders as the only permissible moral position, it forgets that self-government needs a self: a bounded *we* that can deliberate and decide.
Freedom of speech, of association and of worship become slogans on a banner without a functioning polity in which to exercise them. And nobody, he suggests, governs themselves at the scale of three hundred million. Democracy lives in smaller places: a neighbourhood, a union, a guild, a congregation, places built on faces and favours and obligations that cannot be unfollowed. Social media offers a counterfeit of this, connection without consequence. When a lofty, frictionless universalism sneers at local attachment as backward, it saws through the branch on which democracy sits.
Seen this way, his remedies are less a programme than a single principle: bend the incentives, deliberately, back toward the common good. Acemoglu does not believe the market will heal what the market helped to break, nor that some coming technological marvel will descend to save us. He wants existing antitrust law actually enforced, great fortunes taxed so that capital does not pool forever at the top like water in a sealed tank, and investment steered toward AI that raises human capability, and with it, wages. Above all, he wants the regional factories, technical colleges and humble local institutions rebuilt, because those are the rooms in which citizens learn to govern themselves.
None of this is glamorous. None of it fits on a keynote slide. It requires something our leaders, in politics, in the academy and in Silicon Valley, have been conspicuously short of: humility. The humility to stop issuing mandates from above, and to distrust their own technological hubris. And the patience to do the slow, unfashionable work of building coalitions across class and geography.
The damp is still climbing the wall. The question is whether anyone will start the repairs before the plaster gives way.