Singularity economics.

The need for new economic models in an automated society.

Scene from Fritz Lang’s Metropolis (1927)

Henry Ford had already addressed the impact of automation on labor as early as 1908. In his book “My Life and Work”, Ford explains that automation can only improve the lot of the average worker; if a machine allows one worker to do the job of five, then it does not make sense for a factory to fire those four ‘surplus workers’.

Productivity is quintupled for the same cost of labor, and, simultaneously, unit costs of production are driven down, creating wider margins. Each of those five workers should be allocated to this wonderful new machine, and each of those workers can now afford more of the product that machine produces. The factory owner wins, the worker wins, and capitalism works properly.

This conclusion is consistent with classical economic theory of the firm. Capital (machinery, buildings, etc.) is a multiplier for labor — wages, which are a function of the marginal product of labor, are tied to the productive power of capital and how much capital is available. In return, the value of capital — how many machines to buy and what to pay for them, is tied to how effectively workers can make use of them.

But what happens when you have a machine that doesn't need any workers? One would say that you would need someone to manage, maintain or direct these machines. But what if the factory can do that on its own?

What if people aren’t needed any more to direct the inputs and outputs of the machines — machines that detect market demand and the supply situation automatically, then take raw material, turn it into goods, then market, sell and distribute those goods automatically?

It’s here that classical theory of the firm breaks down — there’s a discontinuity. This can be shown mathematically, but in plain language what happens when it doesn’t matter how many workers you have (or if you don’t need any workers at all) is unclear. On the one hand it looks like adding a single worker increases productivity infinitely much. On the other, it looks like the marginal product of labor is zero — no one is getting paid anything.

This is a singularity in the true sense, rather than the gee-whiz ‘exponential growth’ of magical A.I pixie dust. The model we use breaks down at the point where production is independent of labor inputs.

This sounds like an abstruse academic problem, but the implications are deep. Classical theories of the firm have not been exact, but have been ‘good enough’ for a long time. At the very least, they reflect some of what we believe about the relationship between capital, labor, production and markets.

The fact that automation is set to break these models so fundamentally implies that the way we think about these things is not adequate for that particular future. We’re going to need new models, and now is the time to get to work. By the time the robots arrive, it might be too late.