View this post on the web at [link removed]
Valentin Boboc is a Senior Economist at the IEA
One of the things that makes new technologies very exciting is the feeling that they could perhaps do anything we previously believed impossible. The issue with viewing technology – or AI in particular – as a black box with god-like abilities is that, occasionally, boring people like me have to point out that certain things are just unlikely.
The Financial Times ran a piece [ [link removed] ] concerning the possibility of Artificial Intelligence systems re-opening the economic calculation debate. The argument goes like this. If AI can handle gigantic data sets, and economic calculation is all about having too many variables, then perhaps there is hope that a sufficiently ‘intelligent’ or powerful AI could produce optimal central plans for the economy as a whole.
Economic Affairs is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Proponents of central planning are quite enthusiastic in their assumptions about the capacity of new technology to make their dream come true. Even through my short period of time traversing academia, I have come across multiple attempts to revive support for tech-oriented central planning or ‘cyber-socialism’ even before the era of advanced AI kicked off (see for example here [ [link removed] ], here [ [link removed] ], here [ [link removed] ] or here [ [link removed] ]).
The arguments for central planning usually revolve around one or both of the following lines: (a) New technology is powerful enough to handle the immense number of inputs needed for economic calculation; or (b) The number of inputs required for economic calculation is secretly much smaller than we previously believed and so calculation must be feasible.
Point (b) is quite weak, and it’s not hard to see that the number of variables involved in economic calculation is exceptionally high. If we remove money from the equation and we assume that all the data about people’s preferences, the goods and services available for purchase, their relative scarcities and so on are all known and readily available in a spreadsheet, then an economy with 10 goods contains 45 possible pairwise markets. For 1 million goods (which is a low number by today’s standards), there are close to 500 billion possible pairings.
The number of inputs is very high even in overly simplified models of the world. For a model to perfectly imitate the world economy, we’d need more variables than there are grains of sand on Earth.
If we accept that the number of inputs is large, then perhaps that doesn’t matter and all we need is vast computational power.
Unfortunately, even heroic assumptions about computational power do not make the problem disappear. Lucas Engelhardt estimated [ [link removed] ] that allocating 80,000 goods among six billion consumers (approximately the population of planet Earth) involves solving 480 trillion equations, a process that would take the world’s leading supercomputers around 10.5 quintillion years (NB: a quintillion has 18 zeroes).
Even if computer processors became ten billion times more powerful than they are today (which is highly unlikely), such a calculation would still take around a billion years. Even then, at the end of the billion-year calculation, we would have produced an optimal economic plan only for a single snapshot in time. But human societies are dynamic. New goods and services are invented and become available every day, making it impossible to account for every relevant input in a single static calculation.
Perhaps the most straightforward objection is Hayek’s original one: markets are not merely aggregators of calculation inputs but rather the medium through which the inputs are created. Preferences for different goods and services, optimal prices and optimal allocations do not simply exist in the void, waiting to be collected. Rather, they are discovered and revealed through market interactions. A planner cannot calculate without markets because replacing them also destroys the process that generates the necessary inputs.
For me, the easiest critique of technologised central planning is that we simply don’t have time for it. Markets get the job done better and faster. So why waste time when we can already have it all?
Thanks for reading Economic Affairs! This post is public so feel free to share it.
Unsubscribe [link removed]?
