People who warn about a future of "technological unemployment," in which most human labor has been replaced by machines, are often described as if they were Chicken Little, frantically predicting what others have frantically—and falsely—predicted for at least a couple of centuries. What if the more pertinent folktale was the Little Boy Who Cried Wolf? Such predictions have been false so many times before that we have been lulled into supposing they will always be false. What if one day, when we have almost ceased to believe in wolves, one finally arrives?
Of course, in the case of mass technological unemployment, the arrival is not likely to be sudden. It could be that we are already seeing the first signs of it. That is the possibility that Eduardo Porter considers in his latest column for the New York Times. Porter references the work of Wassily Leontief, one of the few economists of his time who was seriously concerned about the obselescence of labor. Leontief responded to those who dismissed this concern by reminding them of what had already happened to another species that once played a major role in the economy: the horse. Porter writes:
Horses hung around in the labor force for quite some time after they were first challenged by “modern” communications technologies like the telegraph and the railroad, hauling stuff and people around farms and cities. But when the internal combustion engine came along, horses—as a critical component of the world economy—were history.
Cutting horses’ oat rations might have delayed their replacement by tractors, but it wouldn’t have stopped it. All that was left to do, for those who cared for 20 million newly unemployed horses, was to put them out to pasture.
In 1915 there were about twenty million horses in the United States. By 1959 there were only four and a half million.
At the time, most economists dismissed Leontief's argument, and it isn't hard to see why. Human laborers can be retrained to do all kinds of jobs. Horses can do only a few kinds of physical work: they can pull and they can carry—that's about it. But the force of the analogy doesn't depend on the very limited similarity between human labor and horsepower. It depends on the similarity between the disruptive force of the internal combustion engine and that of machine learning. Because of the engine, at a certain point machines could do more efficiently almost everything horses had done. Today, many economists who have reflected on the potential of machine learning worry that robots will one day be able to do most jobs that human beings do now, including many white-collar jobs that have long been considered immune to automation.
Why should economists worry about this? Whatever our feelings about horses, do we really regret the advent of tractors and automobiles? Didn't life become much better when work horses were put to pasture?
Better for human consumers, certainly. But it all depends on your point of view. The replacement of horsepower with machines was obviously not better for the work horses—even if you think, as some radical animal-rights activists do, that work animals are essentially slaves. For the redundant horses weren't all put to pasture; many of them ended up in glue factories.
Most economists believe that the productivity gains enabled by automation will greatly benefit consumers, even as they enrich those who own the machines. But one important premise of our current economic system is that most consumers are also workers, who buy what they consume with the money they make from their jobs. The question is: What will happen if there are no longer enough jobs to go around? Where will the income a consumer needs come from?
One common answer to this question is the Universal Basic Income, a kind of entitlement program whereby everyone would received a check from the government. Call it Social Security for all. Which makes sense, since in this scenario many, if not most, people will be effectively "retired" all their lives. They may still have careers of a sort, developing skills that allow them to do satisfying work of intrinsic value. Only, because that work will have little or no market value, they will no longer be able to sell their labor. This is actually a rather sanguine scenario. It could be that technological unemployment will end up spoiling us all with idleness rather than liberating us into a leisure devoted to intrinsically valuable activities. We need to take both possibilities seriously. Technological unemployment is not just an economic and political problem; it's also a cultural one.
But let's focus here on the economics and the politics. If Leontief was right, capitalism is now creating technologies that will upset the basic dynamics of capitalism by evacuating one of its basic social categories: labor. This may not be exactly what Marx foresaw, but it confirms one of his basic intuitions—that capitalism would be undone by internal contradictions. One day the owners of the new machines—a.k.a. capitalists—will be the only ones with a market income, and that income will depend on consumers with some kind of non-market income. If the prophets of Silicon Valley are right, that day may be sooner than you think. There are several ways this prospect could play out, some brutally ugly, some borderline utopian. But all of them would involve a departure from capitalism as we know it.
Just last week Porter wrote another column in which he poured scorn on the idea of a Universal Basic Income. He asked the obvious questions: How would we pay for it? What would it do to the incentive to work? And why should it be universal? Shouldn't we instead direct our limited public resources to welfare programs that help only those who really need help? In that column, Porter touched on machine learning only in passing. A new wave of automation might pose a problem in the remote future, he suggested, but for now there are still plenty of jobs. In his new column, he seems somewhat less confident that technological unemployment can be safely treated as a futuristic speculation. He ends by noting that Lawrence H. Summers, the ne-plus-ultra establishment economist, has lately changed his mind on the topic. Summers is no longer so sure that robots putting us all out of work is something only "stupid people" worry about.
[A]t some point Mr. Summers experienced an epiphany. “It sort of occurred to me,” he said. “Suppose the stupid people were right. What would it look like?” And what it looked like fits pretty well with what the world looks like today.
For large categories of workers, wages are inadequate. Many are withdrawing from the labor force altogether. In the 1960s, one in 20 men between 25 and 54 were not working. Today it’s three in 20. The population is generally healthier than it was in the 1960s; work is almost uniformly less demanding. Still, more workers are on disability.
“Maybe the stupid people weren’t quite as stupid as I thought they were,” Mr. Summers conceded. “This was at least a serious concern that had to be thought about.”
In a world in which many Americans do not work during large chunks of their lives, we might have to conceive of Social Security and disability much more broadly than we do today.
That, Mr. Summers said, “could start to look like a universal income.”
That world is not yet here. It will not be here tomorrow. And there are still good reasons to doubt it will ever arrive. Let the skeptics continue to insist on those reasons, but let them also approach the problem empirically, as Summers has begun to do. This is not a debate that can be resolved with axioms about the unmatchable versatility and ingenuity of human workers. If the anxiety of Luddites has sometimes been irrational, so has the complacent humanism of too many economists. For thousands of years, farmers would have told you there was nothing like a horse. They were right until they weren't.