Where the System Ends, Politics Begins

Can AI fix the economy — or will it repeat our old mistakes?

If you’ve been reading me for a while you’ll remember my early piece, “Who Paints the Dew on the Daisy?” It’s about theories of change; I list several increasingly sophisticated ideas about why and how things in the world change. Theory 3 in that article is “It’s all Clockwork,” and I present it as a thoroughly discredited idea of how the universe is put together.

So I was very interested to discover Abby Innes 2023 book, Late Soviet Britain, in which she lays out a detailed case for neoliberal economics being based on Theory 3.

If she’s right, this has huge consequences—and also shows a path forward for humanity. While I and others have been talking a lot lately about post-capitalism, Innes does not have to go that far to make her point. Capitalism might still be made to work; it’s just that this particular version of it is toxic.

The reason for Innes’ provocative title is that she concludes that while they might seem to be opposites, Soviet economics and neoliberalism share a core theory of how change happens—a core ontology or model of how the universe is put together. That theory is wrong, and its application, whether by Leninists or Thatcherites, is doomed to fail in exactly the same way for each.

The upshot is that both of these “materialist Utopian” visions of economics assumes that the world is essentially Newtonian, mechanistic and predictable—and so are the people who live in it. Thus “homo economicus” who is assumed to have perfect information about the state of the market and always acts with perfect rationality to maximize its utility function. The world contains perfect markets (or markets that would be perfect if we just found a way to rub the dirt off our lenses) that function frictionlessly. Or, contrasting this but identical in result, a centralized planning system that has perfect knowledge of the state of the country and the needs of its economic actors, and makes (or could be made to make) perfect decisions about the allocation of resources.

The world doesn’t work this way. Stephen Wolfram’s Rule 110 is classic example of why: it’s one of the simplest fully-deterministic cellular automata, yet its output is irreducibly complex. Wolfram’s contention is that nature is packed with trivially simple, deterministic systems whose behaviour cannot be predicted even in principle. Rule 110 proves it.

Then there’s “homo economicus.” Humans A) lack perfect information about their options, and B) don’t always choose what’s best for them.

If only these two facts were true, that would be enough to doom both Soviet planning and neoliberal market fundamentalism. There’s an even deeper problem that both system share, however: they start with conclusions, then look for facts to support them.

Both approaches are deductive: they begin with a set of axioms or principles and then work through their logical (or mathematical) implications. Only after this do they go to the outside world to see whether it works the way they’ve predicted. This approach could work, if the world only contained the units and measures that their model assumes. But if the world works in different ways, then when they test the model they may measure the wrong things; if GDP is not an accurate reflection of what’s actually going on in an economy, but it’s the metric your model assumes, then you get what computer programmers call GIGO: Garbage In, Garbage Out.

Another way to put it: if all you’ve got is a hammer, everything looks like a nail.

(It’s possible to study economics differently. Eleanor Ostrom won the Nobel Prize for actually going out into the field and observing how people organize commons arrangements. As she put it, “a system that works in practice will work in theory”—the reverse of neoliberal and Soviet approaches.)

The Alternative to Fantasy Economics

Innes’s framing of the alternative to these systems is both fascinating and highly relevant to us right now. She argues for an open-ended economics, one which is constantly evolving because the world itself is constantly changing. Attempts to find regularities and underlying rules that govern economic order are foolish and doomed to fail; they are attempts at finding timeless axioms for a system so complex that it defies prediction. The alternative is to invent and deploy models that describe the current situation and allow us to make choices now. Such models cannot be axiomatic; they can’t be deterministic, nor allow for Harry Seldon’s Psychohistory to predict our future.

Innes describes the “materialist Utopias” of central planning and market fundamentalism as being united by the vision of eliminating politics. For them, politics is a symptom of a flawed system; if politics is happening it is because economic conditions are not perfect for some subset of the public. The answer is to perfect the system, so that there are no longer political choices to be made. You just run the repaired system, and all boats right themselves. This vision is fully on display in the Soviet attempt to replace government with a strictly operational, managerial bureaucracy. It’s also clear in the neoliberal attempt to privatize everything, including functions that no sane state privatizes, such as health care.

Yet in a world where uncertainty is fundamental and no system can allocate resources perfectly, politics is not only necessary, it’s central. Our relationship with nature has to be one of constant negotiation, we are now entering a period when (whether we go post-capitalist or not) our relationship with our physical surroundings has become political. The climate is now a political actor; it doesn’t need human advocates to be one. But even within human economies (which for the Soviets and neoliberals are closed systemss, but in reality are subsets of global flows of energy and material) politics between humans can never be expunged. In fact, it is critically important that we learn better how to be political and act politically. Economics cannot substitute for that. For better or worse, economics is that.

But… The Singularity!

There’s a current mania, similar to the one surrounding nanotech in the 1990s, that suggests that humanity is either about to be replaced, or that at the very least, we will soon cede control of economic life to AI. Artificial Superintelligence will have effectively perfect knowledge of the world and superhuman decision-making powers. By combining the best algorithms of central planning and decentralized markets, it will finally polish our lenses, remove all friction from economics, and (here’s the crucial thing) by allocating resources perfectly, render politics obsolete.

Does this sound familiar? Maybe, a little like an updated version of Theory 3’s clockwork universe with its omniscient Planner or Invisible Hand?

If so, and if the AI version of post-political economic perfection fails for exactly the same reasons as the Soviet and neoliberal ideologies, then what actually will happen when we build that superintelligence?

…Well, I’ve been here before.