How AI Could Destroy 4 Million Jobs Per Year—And Not Kill The Labor Market

Artificial intelligence software may reduce job growth by 1 million to 4 million jobs each year, but create more than that, economists at Goldman Sachs said.

Jobs whose jobs were threatened by past waves of automation, such as fitness instructors and real estate agents, have continued to grow.

AI could create work for people using AI, as well as boost demand for service workers by raising incomes.

AI may take your job, but it may ultimately create more positions than it eliminates.

At least, that’s what economists at Goldman Sachs are banking on. Researchers at the investment bank published a report Friday predicting AI would not cause a “job apocalypse.” Instead, they outlined an optimistic case for the effect of artificial intelligence software on the labor market in which the technology takes over millions of jobs, but millions more are created to make up for them, keeping unemployment low.

The paper was something of a rebuttal to a viral report last weekend from Citrini Research, which outlined a scenario in which AI software replaces so many human workers that it crashes the economy. Worry over AI-related job losses contributed to the stock market’s volatility last week, analysts said.

If Goldman Sachs economists are correct, fears about AI causing mass unemployment are overblown.

Goldman did not disagree that AI could be massively disruptive. The technology will displace anywhere from 1 million to just over 4 million jobs annually in the years ahead, Joseph Briggs, global economist for Goldman, wrote. Despite that, Goldman does not believe AI will significantly raise the unemployment rate.

“The U.S. economy creates more than 30 million gross new jobs per year, and technological change is the main driver of long-run employment growth,” Briggs wrote. “We expect that these dynamics will repeat and AI will create new jobs while it destroys others. We therefore do not anticipate a job apocalypse.”

AI could create jobs by improving efficiency, opening up new positions for people to use AI tools, and creating demand for service workers by boosting incomes, he wrote.

Indeed, Briggs dug through economic data and found no evidence that AI has so far caused significant job losses outside of a few specific occupations, such as software development. He also argued that just because AI can do a job doesn’t mean humans won’t still be needed, even for the very same role, if history is any guide.

“Employment of fitness instructors (who were replaceable in the 1980s by fitness tapes and even more so today with the proliferation of fitness apps) and real estate agents (whose role as an intermediary has largely been replaced by online platforms) has outperformed overall employment both recently and over the last 25 years,” he wrote.