Automation: Taking or Changing Jobs?

By Patrick Coate, PhD

On Friday, Treasury Secretary Steven Mnuchin made headlines by saying that the displacement of jobs by artificial intelligence was “not even on our radar screen … 50-100 more years” away. This drew a host of shocked responses from those who point out that AI and automation have already eliminated millions of American jobs and may replace many more in the next 5 or 10 years, let alone 50–100. In fairness to the secretary, his comment responded directly to a question about artificial intelligence. In the broader discussion, he spoke about low-paying jobs being replaced by automation, which many of the critiques consider together with artificial intelligence.

The impact of automation is of unquestionable importance: One report released on the same day as Secretary Mnuchin’s comments suggested that 38 percent of U.S. jobs are at “high risk” of automation in the next 20 years. The types of risk are different across occupations. A self-driving vehicle may displace a trucker entirely (even this highly skilled FedEx truck driver). We are not so close to creating robot lawyers, but some labor-intensive legal tasks can be automated. Lawyers would work in tandem with automation, but law firms would need fewer lawyers. Journalism may be similar.

As machines do jobs humans used to do less efficiently, we can either use less labor or get more production. Keynes famously predicted a 15-hour workweek by 2030, but the workweek has not changed much since he said that in 1930. Technology has served thus far to help us make and buy better goods and services, not work less. David Autor, a prominent researcher on the labor market effects of automation, wrote skeptically about the effect of automation in displacing rather than complementing the value of labor at large. In a 2015 article in the Journal of Economic Perspectives, he wrote, “I expect that a significant stratum of middle-skill jobs combining specific vocational skills with foundational middle-skills levels…will persist in coming decades.” Autor cited trades and occupations in repair and medical support as examples of the type of jobs where the routine—and thus more easily automated—functions of the job are difficult to unbundle from the analytical and interactive aspects.

He also argued that certain other jobs will not disappear but evolve. One example of how this could work is the changing role of bank tellers since 1980: the number of bank tellers in the United States has not changed much with the rise of the ATM, but the job has changed a lot, becoming relatively more about customer service and less about filing and handling cash. While there are some workers with job-specific skills who may be replaced by AI and through automation, other jobs will be created or changed.

In short, if Secretary Mnuchin was referring to the total effect of AI on the demand for labor, rather than the effect for specific skills, I think he is correct.

Sign up here to be notified of new articles from Patrick Coate, PhD and AIER.