Modeling Possible Impacts of Large Language Models on the Geography of Work, Education, and Political Polarization
Much of the discussion around artificial intelligence, in particular the adoption of large language models (LLMs), has focused on the displacement of jobs as generative software tools become more adept at traditionally human fulfilled tasks. The speed and scope of the adoption of LLMs is uncertain but given the potential impacts of widespread and rapid adoption of the technology, it behooves policymakers to prepare for a shock to the major national production ecosystem. Much of the literature surrounding the shock to the structure of the economy focuses on the initial disruption to employment: which occupation and to what extent these occupations will be enhanced or replaced through LLMs. But following this initial shock, there are likely to be linked impacts of the geography of where work occurs, what type of education is required, who benefits the most, and the population's response to these shocks, represented in their political behavior.
A new paper by Scott Abrahams, an assistant professor at Louisiana State University, and Frank Levy, a professor emeritus at the Massachusetts Institute of Technology examines past major economic shocks to predict what one might expect from an intense LLM shock and the likely changes to geography of the labor in the United States. “We utilized the post-1980 manufacturing shock because it illustrated the three strategies people used (or did not use) to adjust: acquiring more education, migrating to areas of more opportunity, and changing the political affiliation,” says Levy. “Applying those strategies to today’s circumstance, we identify continuing trends, including the declining belief in the requirement of a college degree for a well paying job and subsequent potential financial difficulties of four-year colleges, new migration patterns and the potential for changing political orientation. Our research highlights the need for policymakers to be proactive so they can dampen the impacts of a too-rapid LLM shock to the country.”
Read the full paper, “Could Savannah be the next San Jose? The Downstream Effects of Large Language Models.”
Read Steve Lohr’s New York Times coverage of the paper, “How A.I. Could Reshape the Economic Geography of America.”