{shortcode-a00fabf81e167e8f2454a3a73d04faf5a0212423}

A study co-authored by Harvard Kennedy School professor David J. Deming found that American adults have embraced generative artificial intelligence faster than they did the internet or personal computers.

According to the study, 39.4 percent of Americans between the ages 18 and 64 used generative AI two years after ChatGPT’s November 2022 launch. The number is nearly double the 20 percent of Americans who used the internet after two years of its launch and PCs after three years.

The study — published as a working paper in the National Bureau of Economics Research last month — is the first nationally representative U.S. survey of generative AI adoption at work and at home, according to the paper’s authors.

In an interview, Deming said the team of researchers wanted to compare the baseline use of generative AI to that of previous technological advances, like the internet or PCs.

“It just became important to me to get some baseline evidence of how often this is being used, mostly to try to compare it to other big waves of technological innovation in the past,” Deming said. “We wanted to know if generative AI was going to be on that scale.”

Researchers replicated questions about the internet and PCs from the Current Population Survey — a national survey that provides information on the U.S. labor force — in their study in order to compare the findings across the surveys, according to Deming.

The study estimated that generative AI supports between 0.5 and 3.5 percent of American work hours and that, at current levels of use, labor productivity could increase by between 0.125 and 0.875 percent.

The researchers also found differences in generative AI use among different demographics.

Workers who are “younger, more educated, and higher-income” use generative AI more frequently, according to the study, which also found that men are 9 percent more likely than women to use generative AI at work and 7 percent more likely at home.

Martin Wattenberg, a Computer Science professor who was not affiliated with the study, wrote in an emailed statement that the study’s “top-line numbers are dramatic, but the demographic details, such as usage differences related to gender, deserve attention too.”

“The fact that so many people are using this technology underlines the importance of understanding its effects on the world, and making sure we can control its consequences,” Wattenberg said.

The researchers plan to release new waves of the survey, with the goal of understanding how generative AI’s use changes over time.

Deming said the main takeaway from the study is that generative AI “is on the same scale in terms of early adoption as PCs and the internet.”

“Knowing now that generative AI is on the same scale as the other two things and thinking about how omnipresent they are in our day-to-day life — it doesn’t allow you to predict exactly what’s coming next, but you should not just think this is a fad that is going away,” he added.