What innovation paths for AI to become a GPT?

Abstract Early commercial applications of artificial intelligence technologies (AITs) were narrow but extremely profitable. Comparable uses of those technologies throughout the economy would lead to a growth boom. Firms which emulated the early applications successfully would make tremendous strategic gains. This is a situation familiar from earlier rounds of information and communication technology. However, for AITs to become a general‐purpose technology across many commercial applications sectors will require some new innovations. This paper examines the innovation paths that could lead to that desirable outcome, the ones that have stalled, the ones in the process now, and the ones that might occur in the future. Strikingly, early AIT use, both commercial and with technical customers, occurred where Digital Transformation was not required for it to succeed. The innovation paths all require Digital Transformation as key steps.