Artificial intelligence: opportunity or challenge for the US economy?

Jason Furman, chairman of the President’s Council of Economic Advisers gave a speech last week on the “opportunities and challenges” presented by artificial intelligence. The potential challenges of artificial intelligence are, to put it in the terms of some recent economic policy conversations, problems posed by the “rise of the robots.” The question at the heart of the debate is how concerned should we humans be about the impact of future technological change on our economy and society? Will we all be thrown out of work? Will new technologies make production so easy that economic growth absolutely booms? Or, perhaps more realistically, will something in between happen?

The answer lies, to an extent, on how technological growth affects demand for labor in the United States and around the world.

First, a nod to Furman’s point that a few more robots might be a good thing for the economy writ large at this point. Insofar as more artificial intelligence would boost productivity, the introduction of more “robots” would give the U.S. economy and those of other high-income countries a much needed boost in productivity growth. How much further “AI” innovation could deliver in an economy where the diffusion of innovation may be a major factor holding back productivity is up for debate.

Yet most of the concerns about the rise of the robots are centered on distributional concerns. The robots might boost productivity growth and therefore the growth of economic output, but will they displace large swaths of workers? Maybe permanently? This first fear—that robots will massively displace labor across manufacturing and services industries—is a concern because in the future additional capital investment in all manner of robots may well supplant labor rather than supplement it. If robots can essentially replace a large chunk of labor, then businesses will stop hiring workers and instead replace them with robots. Imagine an economy-wide version of robots on factory floors.

Past experience with technological change, Furman argues, shows that new technologies don’t reduce demand for all labor, but rather shift the composition of demand for workers. Massachusetts Institute of Technology economist David Autor agrees, arguing that commentators overstate the ability of robots to substitute for labor and forget how capital acts as a complement to labor. Other research shows that capital and labor complement each other, so that more capital accumulation or declining costs of investment actually raise the share of income accruing to labor.

Yet advances in artificial intelligence definitely change the kind of skills that labor needs to supplement the next generation of robots. This kind of deep technological change could reduce demand for, say, middle-skilled workers while boosting demand for less-skilled workers and more highly-skilled workers. The overall level of labor demand might not change, but the overall composition of the earnings distribution could well result in more income inequality. Of course, technology isn’t the only thing that affects the distribution of income—labor market institutions such as union membership also play a significant role.

The extent to which future technological change affects the distribution of income will probably rest on how it impacts the overall demand for labor as well as the kind of skills that become more valuable in the future. There is some evidence that overall demand for skilled labor is on the decline, but how long this recent trend continues and how much technology is responsible for it are open questions.