Artificial Intelligence, Automation, and Work

Innovation and Economic Growth

Article Snapshot

Author(s)

Daron Acemoglu and Pascual Restrepo

Source

in The Economics of Artificial Intelligence: An Agenda, Ajay K. Agrawal, Joshua Gans, and Avi Goldfarb, eds., University of Chicago Press, 2019 (forthcoming)

Summary

Automation tends to displace human workers, reducing wages by reducing the demand for labor. But automation also increases productivity and creates new-labor intensive tasks. Several factors constrain the labor market’s capacity to adjust, especially if automation proceeds too quickly.

Policy Relevance

Automation will lead to painful job losses for individuals, even if new jobs are created overall. Today’s education system might not give workers the skills they need to use new technologies.

Main Points

  • Alarmists predict that advances in artificial intelligence (AI) spell the end of human labor, an overly simplistic view; however, some wrongly deny that AI presents labor policy challenges, merely because past technological advances did not cause overall job losses.
     
  • Historical examples such as the development of agricultural machinery show that automation does not simply augment the abilities of human workers; rather, automation transforms the production process entirely so that more tasks can be performed by machines.
     
  • AI and robotics have a displacement effect, replacing humans in jobs they once performed.
     
    • The common assumption that technologies always increase demand for labor is wrong.
       
    • Automation will simultaneously reduce employment and wages.
       
  • AI and robotics have a countervailing productivity effect, as the cost of performing automated tasks declines, the economy grows and demand for non-automated jobs increases.
     
    • Relatedly, demand for capital increases, which also increases demand for labor.
       
    • The productivity of machines will increase in sectors that are already automated, increasing the demand for labor without displacing any workers.
       
    • Taken together, these effects will not be enough to offset job losses from automation.
       
  • AI will also lead to the creation of new tasks and jobs, a reinstatement effect; this factor will be the most powerful force counterbalancing job losses from automation.
     
    • If automation begins to reduce wages, firms will have an incentive to employ more workers instead of automating.
       
    • AI is already leading to the creation of jobs in some sectors, such as education, where AI is making it possible to provide customized services to small numbers of students.
       
  • Several constraints will hinder the labor market’s adjustment to automation, reducing productivity gains from automation.
     
  • One constraint is a skill mismatch, as the skills required for obsolete tasks are different from those required to use new technologies; observers should be concerned about the ability of the education system in the United States to provide workers with the skills they need.
     
  • Other constraints include factors that encourage overinvestment in AI at the expense of other technologies that enhance productivity.
     
    • The United States’ tax code encourages investment in capital, meaning that firms will prefer machines to other productivity-enhancing investments.
       
    • Labor market imperfections tend to raise wages, encouraging automation.
       
    • Focus on developing AI-related technologies will tend to displace other investments.
       
  • If public policy does not address the tendency of AI to displace jobs, the public reaction could stop the deployment of new AI-related technologies.
     

 

Get The Article

Find the full article online

Search for Full Article

Share