This essay reviews seven books from the past dozen years by social scientists examining the economic impact of artificial intelligence (AI). These works offer valuable insights—AI as cheap prediction, architectural barriers to adoption, data as an economic asset, implementation challenges. However, they offer little guidance when it comes to the transformative scenarios considered plausible by many AI researchers. Economists have made great progress in explaining how to use AI within existing production functions, who benefits, and why; what remains needed is rigorous advice to policymakers concerned about rapid increases in labor churn, scientific development, labor–capital shifts, or existential risk. (JEL C45, C80, D83, O31, O36)

Abstract We track the movement of high-potential startups using cross-state business registrations and estimate the utility of cities to moving startups using a revealed preference approach. 6.6% of these startups move across state borders during their first five years. Startup hubs like Silicon Valley and Boston tend to lose startups to other cities. Our findings show that startups prefer traditional hubs when they move soon after being founded, but later prefer cities with lower taxes. This pattern is not due to vertical sorting or industrial specialization.

Abstract How do barriers to the diffusion of academic research affect innovation? In 2008, the National Institutes of Health (NIH) mandated free online availability of funded research. This policy caused a 50 percentage point increase in free access to funded articles. We introduce a novel measure, in-text patent citations, to study how this mandate affected industry use of academic science. After 2008, patents cite NIH-funded research 12% to 27% more often. Nonfunded research, funded research in journals unaffected by the mandate, and academic citations see no change. These estimates are consistent with a model of search for useful knowledge. Inefficiency caused by academic publishing may be substantial.

Is coachability a usable concept for startups? Using a dataset on high-growth startups in a prominent university incubator, we show that firm characteristics do predict which firms are likely to follow advice but that these “coachable” firms are not more likely to achieve positive outcomes. The reason is that advice may be rejected either because the firm is stubborn and uncoachable, or because the firm possesses capabilities related to the advice which is hard for outsiders to observe.