The puzzle of a powerful new technology spreading everywhere while the economy’s measured output per worker barely improves. The economist Robert Solow put it sharply in 1987: you could see the computer age everywhere but in the productivity statistics. The same question is now asked of AI: if it saves so much time on individual tasks, why does that not yet show up clearly in national figures?
Technical view
Proposed explanations fall into four camps: mismeasurement (gains in quality and free services that output statistics miss), lags (organisations need years of complementary investment in processes, skills and data before a general-purpose technology pays off), redistribution (gains captured by some firms at others’ expense), and simple overestimation of the technology. The information technology version eased with faster productivity growth in the late 1990s, which supporters of the lag explanation cite. For AI, task-level studies often report sizeable time savings while aggregate effects remain contested, and economists disagree on how large and how quick the eventual effect will be.
Related
Part of AI Terminology Index, section: Society and Impact.