Why we should not count on algorithms or silicon brains to run the economy

The control room is hexagonal, containing a circle of white, fibreglass swivel chairs with red-brown cushions and push-button panels. The room is reminiscent of Star Trek but it is no film set.

Project Cybersyn was a bid in the 1970s to algorithmically manage the Chilean economy in accordance with democratic socialist principles under president Salvador Allende.

The idea was not new. Between the wars, economists debated the “socialist calculation” issue: could a benevolent central planner co-ordinate all the production and consumption necessary to run a modern economy, bypassing the greed and waste of the market with a more rational system?

The answer was not obvious to economists, at least not then. The uncompromising Ludwig von Mises argued that it was logically impossible; others that it was merely impractical. But Oskar Lange suggested it could be done: if an economy could be described as a series of simultaneous equations for supply and demand, then the central planner could solve those equations, if only by trial and error.

This proved easier said than done. Nobel laureate Leonid Kantorovich spent six years gathering the data and performing the calculations necessary to optimise Soviet steel production in the 1960s, far too slowly to be useful in a changing economy.

Computers promised more. In an essay published after his death in 1965, Lange wrote: “Let us put the simultaneous equations on an electronic computer and we shall obtain the solution in less than a second. The market process appears old-fashioned.”

That was typical of the awe with which we continue to view these silicon brains. But it was still premature: the computers of five decades ago weren’t fast enough. One credible estimate is that the Soviet Union produced 12-million types of product at its zenith, a mathematical knot that would have taken decades for a vintage computer to unpick. A modern economy produces perhaps 10-billion.

Chile’s Project Cybersyn never had much chance to prove its worth: like Allende himself, it died when the murderous Augusto Pinochet seized power in Chile in 1973.

It was probably doomed from the start. As described in Eden Medina’s book Cybernetic Revolutionaries, the sleek control room masked the fact that Allende’s government only owned four computers.

But the project’s ambitions no longer seem quite so unfeasible. We shouldn’t underestimate the task: Chile’s gross domestic product in 1970 was about $50bn — perhaps $300bn at current prices. Even today, Amazon’s revenue — at $135bn in 2016 — is less than that.

But the power of computers is growing far more quickly than economic output. Could we build an app to run an economy, not only to replace Steven Mnuchin and Janet Yellen, Jeff Bezos and Tim Cook, but to oversee the fine details of production and consumption everywhere, eliminating waste, recessions and inequality?

The idea has resurfaced in the work of Chinese economists Binbin Wang and Xiaoyan Li. They argue that modern computers and cheap sensors make it possible to optimise production in real time, personalised to the needs of citizens.

In some ways this has already happened. Advertisements on Google and Facebook are handled by vast algorithmic markets. If you work for Uber or Deliveroo, your boss is an algorithm. But companies have always been islands of planning in a sea of market forces; an economy in which the government controls all the platforms is something quite different.

One obstacle is tacit knowledge. A textbook economy of supply and demand curves is, in principle, the kind of system that can be understood mathematically. But as Friedrich Hayek argued in 1945, there is a great deal going on in any economy that cannot be counted or even described.

Decisions to produce, to consume and to take a risk trying to create something new are all taken with the knowledge of “particular circumstances of time and place”. Wang and Li believe that big data make this once-tacit knowledge explicit; I am not convinced.

Then there is the issue of power. Facebook and Google already have too much. What would Stalin have done with such information? Or Pinochet? China is already using the data exhaust collected by Alibaba and Tencent to exert social control.

But Hayek was right about the power of market prices to co-ordinate a complex economy steeped in tacit knowledge. Market forces remain a more powerful computer than anything made of silicon. We can shape its inputs and outputs with taxes that penalise pollution, redistribute income or encourage social goods.

But replacing the market with state-run algorithms is an idea that should stay in the realms of science fiction. /©

MARKET FORCES REMAIN A MORE POWERFUL COMPUTER THAN ANYTHING MADE OF SILICON