The Keystone Research Center (KRC), a member of EPI’s Economic Analysis and Research Network, released the first report of a project funded by The Heinz Endowments on the future of work—looking at the impact of artificial intelligence (AI) and policy options for America’s fast-changing economy. The report concludes optimistically: AI need not increase economic inequality in the United States and provides an opportunity to revisit the orientation of US economic policy to ensure the opposite result.

“We may not have complete knowledge of what AI means for the future of work, but this paper lays out key principles necessary for a fair and prosperous future for workers,” said EPI President Thea Lee, who sits on the project’s advisory committee. “The work being done by KRC in this arena is an important step towards ensuring that the economy works for everyone, no matter what the future holds.”

Towards an AI Economy That Works for All analyzes AI’s likely impacts by examining past impacts of technology, including robotics and information technology, on the economy and jobs. It also considers how AI does—and does not—go beyond previous technologies and substitute for human capacities and intelligence. The report goes on to explore the roots of inequality in the economy within which AI technologies will spread, and the policies needed to promote equity.

“To date, the Future of Work debate has produced much speculation about job impacts but little attention to policies that can ensure broad distribution of the benefits,” said Dr. Stephen Herzenberg, economist and Executive Director of KRC. “Because AI will hit an economy and society already hobbled by gaping inequality, we must implement new technology in ways that make things better not worse.”

Key findings include:

• Fears of massive job displacement as a result of AI and continuing advances in robotics are overblown in the short- to medium-term. Industry-disrupting innovations that draw on multiple technologies—which for AI means not just software but a great many hardware technologies (e.g., sensors for automated vehicles)—almost invariably perform poorly at first and evolve more slowly than anticipated.
• AI will penetrate more broadly than previous IT systems in part because of machine learning, processes through which computers gain mastery in specific, narrow domains by absorbing massive amounts of data. (Computers that “learn” from over 100,000 X-rays, for example, outperform most radiologists.) Even so, AI-enabled automation will remain well behind human capabilities for the foreseeable future in tasks requiring judgment, tacit skills, and common sense—tasks that infuse many parts of most jobs.
• While AI will ultimately have widespread impacts, predicting when and in what occupations is notoriously difficult. We can be sure only that AI will cause significant disruptions—and as a society we should prepare for that.
• The disruptions caused by AI will occur in an economy with levels of inequality as great as at any time in our nation’s history. Some economists attribute much of today’s inequality to previous generations of technology that favor more highly educated over less educated workers. The new report finds that a far stronger case can be made that public policy (e.g., deregulation, including of labor markets) more fully explains the increase. AI offers an opportunity to reimagine the nation’s policies and to tip the scales back toward the interests of the 99 percent rather than the 1 percent.

The report also aims to spur richer debate about policies to ensure that AI generates broad benefits. Such debates have so far been bracketed by the view that we shouldn’t or couldn’t do anything (a view favored by proponents of deregulation and “free markets”) and by recent enthusiasm for a guaranteed income, premised on the idea that machines will do so much work that people must have means to support themselves divorced from paid employment.

KRC identifies eight broad policy categories which demonstrate that policymakers have a great many tools between doing nothing and implementing a guaranteed income. These policies include a reinvention of U.S. education and training so that workers can learn—and learn again if necessary—the skills needed to acquire a family supporting job. But education alone, while a good in itself, does not pack the punch needed to generate more equitable growth. Other policies are needed, such as rebalancing the scales between employers and workers, and generating more jobs through shorter work time and public employment creation.