Could AI revive the socialist dream?

In all the speculation about what future artificial intelligence is about to usher in for us, everyone seems to agree on one thing. It is generally assumed that if AI will transform our economic system, it will do so in the form of more of the same — a more intense capitalism. Depending on your perspective, that is for better (higher productivity) or worse (more winner-takes-all exploitation), yet the debate is mostly about what sort of remedies, redistribution or regulation are needed in response.

But might all this be a failure of imagination? When we view the technological leaps we are living through in the light of the great century-old disagreement known as “the socialist calculation debate”, a profoundly different possibility suggests itself. If AI fulfils even a portion of the promises made for it, it might herald not a more intense version of the capitalism of the present but a “back to the future” trajectory: it could serve as the handmaid of the socialist planned economy that so many dreamt of in the 20th century but that was never remotely realistic — until now.

1989 marked the political victory of capitalism over socialism. But in technical and intellectual terms, the battle had been fought and concluded long before. The socialist calculation debate, largely forgotten today, raged most fiercely in the interwar years between the thinkers who understood most deeply how the economy functions. And it asked not which social system was more just, egalitarian or free. Rather, it took the more fundamental question to be which was most efficient: whether unrestrained markets or government fiat could most rationally allocate society’s economic resources.

Among the economists lined up on the pro-capitalist side of this debate were luminaries of the Austrian school, such as Ludwig von Mises and Friedrich von Hayek. Against them stood, most famously, Oskar Lange, a Polish economist and diplomat, and Abba Lerner, a brilliant if eccentric British-American economic theorist. The latter pair both argued that state planning could allocate resources as well as (if not better than) markets, and showed how this could be so within the framework of classical economics.

Imagine a super-AI armed with access to all the data we let our devices record . . . such technology would surely bury Hayek

Imagine a super-AI armed with access to all the data we let our devices record . . . such technology would surely bury Hayek

From the late 1800s, theorists had elucidated how, through the mechanism of adjustable prices, free markets can converge on common prices and efficient, stable allocations of resources, pioneering the mathematical modelling of Adam Smith’s “invisible hand”. Pro-planning thinkers, however, used these theories to show how centralised agencies could mimic — and then improve on — the effects of markets. (One idea, for example, was to publish draft production plans and let managers of state-owned companies make bids for supplies and output amounts at centrally given prices — then adjust the plan until all the bids were compatible and matched the planners’ goals).

But to achieve this in practice, a central planner would need massive amounts of information from every little corner of the economy. That was the fundamental insight of Hayek’s 1945 article “The use of knowledge in society”, which set the tone for the half-century that followed. Hayek’s genius was to see market prices not merely as devices to balance supply and demand, but as carriers of information. The knowledge needed to allocate resources well — of who needs what and who can provide what on what conditions — is so local and dispersed that no central planner could ever hope to collect it. But, Hayek argued, the market price, in a single number, conveys just what is needed for local decisions that match the overall efficient market allocation. And this information, much of which just sits in people’s heads, simply does not exist in centralised, communicable form.

Until the end of the last century, this was a winning argument, underlined by the miserable failure of socialist regimes to provide for their populations’ material wellbeing. But if the free-market side won the socialist calculation debate in practice, the socialist side secured a toehold by showing how planning could succeed in theory. The Hayekian conclusion always remained contingent — and only true so long as governments lacked the necessary informational and computational capacity for central planning to outdo free markets.

The question is whether AI changes that. When I studied the socialist calculation debate in the 1990s, the joke was that communism had collapsed just as the computer revolution was making planned economies look less hopelessly unrealistic.

Consider the giant steps already taken in IT in recent decades. In Hayek’s time, information about local prices and quantities could only be collected by human workers armed with clipboards. Today, with so much commerce being done online, prices can be comprehensively “scraped” off the internet. One pioneer was the aptly named “billion prices project”, a Harvard-MIT research initiative launched in 2008 by economists Alberto Cavallo and Roberto Rigobon to collect real-time prices to improve on official statistics. The same goes for quantities: inventories, order books and sales volumes are all digitised and in principle accessible to a government planner.

Add the internet of things, and the amount of information gathered, processed and stored multiplies further. That might have, until recently, been unusable in raw form. But if there is one thing AI can clearly do, it is to organise unstructured information. The upshot is that the practical challenge of collecting, organising and processing data can no longer be seen as a remotely plausible constraint on the efficiency of a planned economy.

That is not all. AI offers the promise of transcending price information altogether by predicting who wants, needs or can produce what at what cost, better than market prices could ever summarise. Behavioural algorithms, notably in advertising, are already doing a decent job of this: we are familiar with the uncanny feeling when search engines and social media serve up an ad that seems just a little too on-the-nose. Judging by the advertising revenues of Big Tech, producers of all kinds of products think that ad-placing algorithms are worth the cost.

Price adjustments do not stabilise the macroeconomy; so we get business cycles, bubbles and busts

Price adjustments do not stabilise the macroeconomy; so we get business cycles, bubbles and busts

Imagine then a super-AI armed with access to all the data we let our devices record for every transaction we enter into, and much more besides. Such an AI would surely bury Hayek.

Indeed, a big motivation for the socialist calculation debate was that market prices get a lot of things wrong. People don’t always know enough to choose the best goods; they get addicted, defrauded or bankrupted. Consumption and production have spillover effects on third parties — “externalities” — that pricing decisions do not capture; so we get pollution, resource depletion and climate damage. Price adjustments do not stabilise the macroeconomy; so we get business cycles, bubbles and busts. These shortcomings of the price mechanism were easy to grasp in the wake of the Great Depression. The same realisation may be gaining strength again, given the disenchantment with capitalism expressed by so many young people in particular.

Against this background, it beggars belief that once a super-AI was asked to advise on the allocation of resources, it would merely follow the outcome of the highest bids in free markets, rather than calculate the most efficient allocations it could find by itself. Why on earth would AI agents abdicate from central planning?

This thought experiment raises more questions than it answers. It hardly proves that AI will bring about socialism. But it does show that AI removes one of the 20th century’s strongest — perhaps the strongest — argument against central planning. And it suggests that to the extent that both public and private sectors adopt AI to guide decision-making — how much to invest; in what; how best to tax for desired policy outcomes? — our economies will look increasingly planned and decreasingly market-shaped.

This raises some huge political questions. Would the current leaders of AI development — almost all hardcore libertarians — support capital allocations being decided by AI-powered planners rather than free financial markets, given how much the latter favour them? Conversely, might the general public, whose AI scepticism is currently rising, be won over by more “rational” economic decision-making that delivers more for them than our existing form of capitalism? Above all, would an economy where AI informs most decisions be able to overcome the conflicts of interest and co-ordination that have always held back productivity in our actually existing economy — all kinds of tragedies of the commons and prisoner’s dilemmas? (Consider, for example, whether a super-AI empowered to allocate resources would tolerate our very slow rate of decarbonisation.)

We don’t know, in other words, whether the superhuman capabilities of AI agents will break free of all-too-human constraints, incentives and interests. Readers will have to judge whether that is a source of disappointment — or relief.

Martin Sandbu is the FT’s European economics commentator

Find out about our latest stories first — follow FT Weekend on Instagram, Bluesky and X, and to receive the FT Weekend newsletter every Saturday morning