2019-05-07

The Automation Question

I’ve been pondering the automation question. Everyone keeps howling about how the robots are going to take our jobs and usher in a future of fully automated luxury space Yangism, or maybe it will all be controlled by the rich, and the rest of us will starve and get liquidated by Jeff Bezos’ robot army as the price of labor falls below subsistence levels. But I’m not seeing it.

If the scariest kinds of automation were real, Elon would have been able to fully automate the gigafactory instead of having to hire 10000 workers, our clothes would be made by robots instead of Bangladeshi women and children, and there would be robots threatening to fix the potholes so that my bike doesn’t get a flat tire on the ride to work. No doubt there’s some automation occuring, but it sure looks to me like there’s way more useful work to do, that only humans can do, than we are actually able to do.

Let’s think about this for a few minutes.

First of all, what is automation? At first pass, it’s human labor being replaced by machine labor.

What about hand tool innovation and stuff like the flying shuttle in weaving? An advance in technology makes it suddenly much easier for fewer human workers to do the job of many. Productivity increases, and a bunch of people potentially lose their jobs as the relatively inelastic demand is now met by fewer workers. But price of labor and goods falls, and eventually the displaced people find new jobs doing things that didn’t used to be worth it, but now are. There’s temporary technological unemployment, which can last for some time and be politically significant, as in the early 19th century “Engels’ Pause”, but no one dreams of calling this automation.

Let’s get a bit closer. What about electrically powered autolooms programmed by punched cards, robots that drive battery parts around at Elon’s gigafactory, high precision pick-and-place robots, and self driving subways and cars? This is definitely automation, but the analysis would seem to be exactly the same as above. It’s just labor saving tools, and tools that can do new jobs that didn’t used to be able to be done at all. People get displaced, the price of labor falls, but the price of goods falls too.

It would be a mistake to just skip over the potential problem here, though: the price of labor may fall faster than the price of goods, and the price of goods may not fall evenly. We’re all familiar with that graph of inflation in various goods categories; TVs are cheaper than ever, and falling fast, while housing, food, and education are all rapidly becoming more expensive.

Is it possible for the price of labor measured against a basket of essential goods to fall permanently below subsistence? Can this be brought on by technological innovation?

The malthusian answer is obvious: bulk population outstrips bulk resources and processes available to produce essentials like food. That’s a nasty condition. But the automation worry isn’t the malthusian worry, rather it is the worry that even if there’s enough essentials, people can’t afford it because there’s nothing economically productive they can do for it, because machines took the jobs.

We can imagine a few possible levels of automation, which have different economic implications:

Stage 0: all labor is done by humans, using only dumb hand tools.

Stage 1: Labor is done by humans, using advanced high-investment specialized tools (or draft animals) that do most of the labor automatically, such that many fewer people are needed for a given level of productivity. Innovation here can cause an “engels’ pause” as technological capacity outstrips capital accumulation and thus surplus temporarily goes to capital investment instead of labor, but no permanent technological unemployment.

Stage 2: Capital can fully replace some formerly human jobs, but there are still enough jobs only humans can do that any given human can do. There is no one faced by permanent technological unemployment, only temporary technological unemployment.

Stage 3: Capital can fully replace the lowest-skill tier of labor, such that there is no job that some people can do, that a machine can’t do better. This is where we have permanent technological unemployment, but it’s arguable whether this unemployment is even technological, as there have always been people with basically negative productivity who are supported by the community (or not supported) for non-economic reasons. Many factors can cause this fraction to grow or shrink, technology being only one.

Stage 4: Capital can fully replace most people at most essential jobs, with most people becoming permanently unproductive in the primary economic and military stack, such that a small capital-owning class could in principle forcefully successfully secede from the rest of human society. This is the “hyperinequality” scenario. But even here, we have some small number of non-replacable humans in charge of the whole thing. This is the scenario imagined by some Silicon Valley elites who wring their hands about how their tech will take billions of jobs.

Stage 5: Machine capital can replace humans at all roles, even political leadership and the organization of power. As such, humans can be fully phased out, or liquidated, by the new machine overlords. This has to be understood as what is almost an alternate race of machines that displace us demographically.

Malthusian worries, where humans can’t do enough work to justify their existence in a world of elevated resource pressure from population and industry, seems possible. Maybe it will become a concern at some point. But it’s more about overpopulation and economic productivity inequality than automation.

It looks to me like we don’t get real, serious technological unemployment until stage 4. Whether even that situation could possibly be stable is questionable. What are these capital-owner elites doing that neither machines nor the rest of humanity can possibly do? Most economic flows are much more dependent on general intelligence than the automation hype would have you believe. And once you have general intelligence, what holds you back from replacing the top, too? Rather, this seems like a transitional form where you’re actually in stage 5, but either you haven’t realized it yet, the transition hasn’t completed, or you still think you can control it.

In stage 5, machines are a political and existential problem, not an economic problem. Political in the sense that they directly compete with us for sovereignty. This is a much more serious concern than technological unemployment, and only war or diplomacy, not UBI, would save humans at that point.

So I think automation is not an economic problem at all, until it becomes a very serious existential problem. I don’t think the people talking about it are mistaken. I think they’re lying. There is something else motivating the discussion, that they are simplifying as “automation”. I wonder what it is.