When you hear predictions about future machines that can think and act at or above human levels, fears about mass technological unemployment seem understandable. But until such thinking-and-doing machines exist and are ready for market, such worry is a time waster. “AI-powered robots are going to take all the jobs, obviously” shouldn’t be today’s baseline way of thinking about technology and labor markets.

Here’s a better way to think about the issue, according to “Artificial Intelligence and the Future of Work,” a new report from the National Academies of Science, Engineering and Medicine (NASEM):

• First, jobs typically consist of multiple tasks, not all of which can be automated. 
• Second, automation often increases productivity—we do more with less—which lowers prices and increases demand, potentially creating more jobs. 
• Third, new technologies also typically create new types of work. Over 60 percent of current jobs didn’t exist in 1940. 
• Fourth, workers and institutions adapt, developing new skills and ways of working alongside technology. Public policy matters.
• Finally, current demographic trends point to labor scarcity, not surplus. Public policy matters here, too, such as new restrictions on immigration.

What’s really helpful and informative about the NASEM report are the real-world examples that support sound economic thinking about tech and jobs. What’s happened with accounting, for instance, shows how automation can complement rather than simply substitute for human workers. An influential 2017 academic study predicted that accounting, bookkeeping, payroll, and tax-preparation work would be automated with 97 percent probability, suggesting massive job loss thanks to widespread adoption of accounting and tax software. Rather than declining, however, employment in these occupations doubled between 1990 and 2024, growing at more than twice the rate of overall employment.

Or look at radiology as an example of a job that’s a bundle of tasks rather than just one thing. In 2016, AI pioneer Geoffrey Hinton famously declared that people should stop training radiologists, predicting AI would outperform them within five years. Despite the development of AI systems that match or exceed human performance in analyzing medical images, 2021 was actually a record year for radiologist job postings, with early 2022 postings more than double the 2019 rate. What that forecast missed is that radiologists perform 30 distinct tasks, of which reading images is just one. This example also shows that just because a technology is available doesn’t mean it will be widely used ASAP: “The American College of Radiology’s AI Central site lists more than 200 Food and Drug Administration–approved radiology AI algorithms. However, actual adoption of these algorithms has been slow.” Something to keep in mind when thinking about the impact of generative AI.

And I really love this bit about aircraft pilots, which shows how technology can boost demand:

When jet engines replaced propeller aircraft, pilot productivity (measured in passenger miles flown per hour) increased dramatically. Traditional economic logic might suggest this would reduce pilot employment. Instead, the total number of pilots increased because the improved productivity led to lower costs and increased demand for air travel. This illustrates how technological advances that boost productivity can expand rather than contract employment opportunities through their effects on market demand.

Of course, there are lots and lots of examples of these various economic forces playing out, such as ATMs failing to eliminate bank tellers or computer graphics failing to eliminate Hollywood special effect artists. Heck, after 250 years of industrialization—and 250 years of fears about machines taking all the jobs—the US is seeing some of the lowest unemployment rates in decades.

Again, from the report:

Although there is widespread concern about the impacts of AI on jobs, at the time of this writing U.S. unemployment rates are very low compared to historical levels; apart from a spike in unemployment owing to the COVID-19 pandemic, they have been extremely low for the past several years. In addition, population and labor force growth rates in the United States and across the industrialized world are expected to decelerate. Against this backdrop of structurally strong demand for labor and structural headwinds impeding increases in labor supply, it is difficult to predict whether adoption of new AI will result in a decline in aggregate labor demand, manifesting in either fewer jobs (relative to working-age population) or, more likely, lower pay for existing work. Additionally, adoption of recent AI advances in the workplace is still nascent and—despite recent improvements—measurement of AI’s impacts is still limited, precluding a definitive assessment of the current impacts of AI on the workforce.

Maybe don’t expect to get a universal basic income anytime soon.