Will artificial intelligence kill off money?

We have forgotten the true virtues of money. We commonly view it in purely instrumental terms — as a device that facilitates exchange and stores value over time. Compared to bartering, coins and paper currency are profoundly convenient. But money is more than just an instrument. As Fyodor Dostoevsky famously observed, “money is minted freedom.” It supports our existence as autonomous individuals in a decentralized economy.
The situation will be very different, however, if we someday live in a world governed by an artificial intelligence endowed with complete information and infinite processing capabilities. Under those conditions, there might no longer be any role for money.
To understand how money sets us free, consider any chain of transactions. When we receive money, we are in control. It is up to us to decide whether to hoard it or to spend it on whatever we choose. Only money gives us that capability. Money, moreover, is universal. It allows you to buy anything, anytime, from anyone — and that seller can increasingly be located anywhere. This specific freedom comes not from wealth, but from the possibility of choice.
This choice must not be taken for granted, because payment instruments have always been susceptible to paternalistic interference. In the 19th century, some firms would pay their employees with scrip that was only accepted in company-owned stores. Nowadays, technology makes it possible to issue “programmable money” with a special purpose, limited use and even a preemption date. Such digital tokens could be used to prohibit “non-virtuous” consumption (such as alcohol or tobacco) by recipients of public assistance.
We need money because we live in a market economy, not a preprogrammed world. Holding money protects us against uncertainty (it has an “option value,” in economic parlance). The demand for money increases sharply during crises because people need to be prepared for all contingencies.
But now fast forward to a future where machines organize, decide and execute all economic activities. They transact among themselves, credit and debit reciprocal accounts and automatically ensure discipline and the enforcement of contracts. There is no failure and no default. Do we still need money?

When we receive money, we are in control. It is up to us to decide whether to hoard it or to spend it.

Jean-Pierre Landau

Such a world is already partly in view. A decade ago, Hal Varian, chief economist at Google, noted that “there is a computer in the middle of each transaction” across our digitalized economy. Many of our daily acts are automated, algorithmic trading dominates in many securities markets and payments are increasingly programmed.
If cryptocurrency advocates’ vision of the future materializes, “smart contracts” will govern financial intermediation in a universe of decentralized finance. Some people even suggest that, in conducting monetary policy, sophisticated algorithms could replace central bankers.
For now, though, we merely use the machines. Humans still make decisions as free agents, expressing preferences and acting on them. Central banks do not blindly follow rules. They make judgments after considering difficult tradeoffs, especially in times of crisis or when facing negative supply shocks.
In fact, in an automated economy with human control, money is more necessary than ever. But it must adapt now that digitalization has collapsed distance and time. Money, too, must be digital, taking the form of tokens on our mobile phones — an e-cash that can be transferred instantly across the world without having to transit through a complex web of accounts and counterparties.
A threshold will be crossed with AI. Some scenarios project a universe where AIs do not simply process information and execute commands, but also make decisions and even determine their own objectives. They would be the “agents,” acting on preferences that may not necessarily align with those of humans.
With these capabilities, an AI could take over the allocation of resources and the distribution of income. It would assess the millions of possible economic equilibria and identify the one it considers preferable. This vision of “techno-socialism” is very far from the model of a decentralized, free society. It would be a high-tech version of old-fashioned communist central planning.
Debates about the future of AI often refer to the “singularity,” meaning the point when AIs will have the ability to improve and augment themselves, thus rapidly surpassing humans in all measures of intelligence. In this scenario, humans would no longer control their own destiny.
Will it ever happen? There are huge disagreements among AI experts, but money can serve as the ideal indicator. The best sign that AI has effectively taken over would be if money becomes irrelevant in economic life.
A moneyless world may be technically feasible. But whether it would be worth living in is another matter.

• Jean-Pierre Landau is Associate Professor of Economics at Sciences Po. ©Project Syndicate