Artificial intelligence has revived an old question in economics: if computers can process enormous quantities of information, could they eventually overcome the informational difficulties that make central economic planning impossible?
Modern AI systems can analyze data at a scale no individual human being could approach. They can process prices, production statistics, consumer behavior, inventories, transportation data, and countless other variables. It is therefore tempting to argue that the traditional Austrian criticism of central planning may eventually become obsolete. Perhaps the problem was never centralization itself, but simply insufficient computing power. This argument, however, misunderstands the nature of the knowledge problem.
In his famous essay “The Use of Knowledge in Society,” Friedrich Hayek argued that the knowledge required to coordinate an economy is dispersed among millions of individuals. Much of this knowledge is local, contextual, temporary, and difficult or impossible to articulate. It is not simply a database waiting to be collected by a sufficiently powerful computer.
AI can process information. It cannot turn centrally-collected information into the same economic process that generates market prices and entrepreneurial decisions.
The Knowledge Problem Is Not Merely a Data Problem
Consider a farmer deciding whether to plant wheat or corn. His decision depends on factors that may include his particular soil, expectations about weather, knowledge of local suppliers, relationships with workers, machinery conditions, expected prices, and his own judgment about future demand.
Some of this information can be recorded. Much of it cannot be meaningfully separated from the individual who possesses it. More importantly, the relevant information is constantly changing.
Hayek’s point was not simply that planners lack enough information today. It was that the economic problem exists precisely because knowledge is dispersed and continually changing. The planner cannot simply collect all the relevant facts once and then calculate the correct allocation. By the time the information has been collected, evaluated, transmitted, and incorporated into a plan, the underlying circumstances may already have changed.
This is why Hayek described the price system as a mechanism through which individuals can coordinate their actions without possessing knowledge of the entire economy.
A producer does not need to know why the price of copper has increased. The higher price itself may be enough to induce him to economize on copper and search for substitutes. The price communicates something economically relevant without requiring the producer to understand the entire chain of events behind it.
AI can make this process faster and more sophisticated. It cannot eliminate the underlying problem that the information being coordinated is generated by decentralized human action.
But What If AI Knows Everything?
A stronger objection is possible. Suppose an AI system eventually becomes capable of processing virtually every economically-relevant piece of information. Suppose it knows consumer preferences, available resources, production technologies, inventories, transportation networks, and even the plans of individual producers.
Wouldn’t the knowledge problem disappear?
Even this scenario does not solve the Austrian problem. Ludwig von Mises’s argument against socialism was not ultimately dependent on planners being unintelligent or poorly informed. His economic calculation argument asks a different question: how can an economic system determine the relative value of alternative uses of factors of production when those factors are not privately owned and exchanged on markets?
The issue is not merely whether a computer can calculate faster. It is what the computer is supposed to calculate.
Consider a government deciding whether a particular quantity of steel should be used to construct a railway, a factory, or machinery for producing consumer goods. Knowing the physical quantities involved is not sufficient. The planner needs to compare alternative uses of scarce factors economically.
Under capitalism, entrepreneurs make these comparisons through monetary calculation based on market prices. Prices for capital goods emerge through exchange and private ownership. Entrepreneurs can compare expected revenues with costs, calculate profits and losses, and revise their plans accordingly.
If the state abolishes private ownership of the means of production, it cannot simply manufacture genuine market prices by programming them into a computer.
Mises made precisely this distinction in his discussion of economic calculation. Even a hypothetical planner possessing extensive technological knowledge would still face the problem of rationally evaluating alternative uses of capital without genuine market prices.
AI Cannot Create the Institution That Generates the Information
This is where the distinction between Hayek’s knowledge problem and Mises’s calculation problem becomes especially important.
AI can aggregate information. It can identify patterns that humans overlook. It can forecast demand, optimize logistics, and help businesses make decisions.
But the market does something more fundamental: it creates an institutional environment in which individuals generate and discover information through ownership, exchange, competition, profit, and loss.
A price is not simply a piece of information that exists independently of the market. It is the result of exchanges between people who value goods differently. The distinction matters.
Suppose an AI system observes that the market price of a particular machine is $50,000. It can use that number in an optimization model. But if we ask the AI to replace the market itself, we face a circular problem: where do the economically meaningful prices come from?
A central planner could instruct an AI to estimate what the price of a machine should be. But an estimated price is not equivalent to a market price generated through actual ownership and exchange. The difference is between observing the outcome of decentralized economic decisions and replacing the institutional process that produces those outcomes.
AI May Strengthen Markets Rather than Replace Them
There is therefore an important irony. The development of AI may make markets more powerful rather than make central planning viable.
Businesses can use AI to process information that would previously have been too costly to analyze. Entrepreneurs can discover opportunities more quickly. Consumers can compare products more easily. Firms can improve forecasting and coordinate complex supply chains. In this sense, AI can reduce the cost of using knowledge that is already available to market participants. But reducing the cost of information processing is not the same thing as abolishing the need for decentralized decision-making.
Hayek’s central insight remains relevant: the economic problem is not simply that society possesses too much information for one computer to process. It is that relevant knowledge is dispersed among people, embedded in particular circumstances, and continually discovered through interaction.
AI may become extraordinarily good at processing explicit information. It may become better at forecasting, optimization, and pattern recognition. None of this means that it can become the owner, entrepreneur, consumer, and market participant whose actions generate the economic process it is supposed to coordinate.
The computer can calculate. The market determines what is worth calculating in the first place. That is why artificial intelligence does not invalidate the Austrian critique of central planning. If anything, AI makes the distinction between computation and economic calculation more important.
The central question is not whether a sufficiently powerful machine can process enough data. The question is whether economic calculation can exist without private property, market exchange, sound money, and the price system that emerges from them. AI can process the signals; it cannot replace the market that generates them.
