Research Highlights
The Time to Focus on AI’s Impact is Now
—
Slowing economic growth over the past decade underscores the importance of AI to deliver on its potential productivity benefits.
By Robert Seamans and Jason Furman
If the digital age morphs into the artificial intelligence (AI) age, labor economists and policymakers want to be ready to accommodate that transformation and its effects on labor and productivity. With that in mind, NYU Stern Professor Robert Seamans reviews the evidence to date and suggests strategies of dealing with the impact of AI.
Professor Seamans’ conclusions are laid out in “AI and the Economy,” a paper co-authored with Jason Furman of the Harvard Kennedy School. While Professor Seamans describes evidence of a large increase in AI-related activity over the past several years and indications that AI and robotics together have the potential to increase productivity growth, he also predicts mixed effects on labor, especially in the short run.
The authors assert that the well-documented slowdown in advanced economies’ productivity over the past decade, compared to the previous decade, underscores the need for policies that will support efficient AI development and use by both existing firms and startups. They also explore how AI might affect the trendline of male labor force participation, which has fallen from 98 percent in the 1950s to 89 percent in 2016 and ask whether men with less than a high school degree will become increasingly left out as AI use expands.
One pressing concern is that AI innovations will create new or transformed occupations so quickly that the ability of workers to acquire relevant new skills will lag, and thus large segments of the population may be unemployed for sustained periods of time. The share of jobs requiring AI skills, such as machine learning and deep learning, has already increased almost five times since 2013.
The authors use data on robotics to infer the impact AI will have on the economy. For a sample group of 17 advanced economies, robots added an estimated 0.4 percent of annual GDP growth between 1993 and 2007, accounting for a tenth of GDP growth during the period. Though it’s too early to measure AI’s impact on labor markets, “A reasonable inference… is that… AI will not be labor-displacing but could still pose significant downsides and raise other concerns,” they write.
While such effects may take decades to materialize, the authors emphasize the need to consider potential policy responses. The more AI produces sector-specific, rather than macroeconomic shocks, the more a response should be targeted and focused, building on successful past efforts rather than using an unprecedented approach such as universal basic income (UBI), wage supplements and guaranteed employment, they argue. An AI-specific agency might eventually be needed to sort out the benefits of labor supports, data portability and other strategies for addressing these issues.
Professor Seamans’ conclusions are laid out in “AI and the Economy,” a paper co-authored with Jason Furman of the Harvard Kennedy School. While Professor Seamans describes evidence of a large increase in AI-related activity over the past several years and indications that AI and robotics together have the potential to increase productivity growth, he also predicts mixed effects on labor, especially in the short run.
The authors assert that the well-documented slowdown in advanced economies’ productivity over the past decade, compared to the previous decade, underscores the need for policies that will support efficient AI development and use by both existing firms and startups. They also explore how AI might affect the trendline of male labor force participation, which has fallen from 98 percent in the 1950s to 89 percent in 2016 and ask whether men with less than a high school degree will become increasingly left out as AI use expands.
One pressing concern is that AI innovations will create new or transformed occupations so quickly that the ability of workers to acquire relevant new skills will lag, and thus large segments of the population may be unemployed for sustained periods of time. The share of jobs requiring AI skills, such as machine learning and deep learning, has already increased almost five times since 2013.
The authors use data on robotics to infer the impact AI will have on the economy. For a sample group of 17 advanced economies, robots added an estimated 0.4 percent of annual GDP growth between 1993 and 2007, accounting for a tenth of GDP growth during the period. Though it’s too early to measure AI’s impact on labor markets, “A reasonable inference… is that… AI will not be labor-displacing but could still pose significant downsides and raise other concerns,” they write.
While such effects may take decades to materialize, the authors emphasize the need to consider potential policy responses. The more AI produces sector-specific, rather than macroeconomic shocks, the more a response should be targeted and focused, building on successful past efforts rather than using an unprecedented approach such as universal basic income (UBI), wage supplements and guaranteed employment, they argue. An AI-specific agency might eventually be needed to sort out the benefits of labor supports, data portability and other strategies for addressing these issues.