Could Machine Learning Be a General-Purpose Technology? Evidence from Online Job Postings

General purpose technologies (GPTs) push out the production possibility frontier and are of strategic importance to managers and policymakers. GPTs are typically identified with many years of hindsight. For managers deciding on technology strategy, this classification will come too late. In this paper, we use data from online job postings to provide an approach to assessing the relative likelihood that a technology is general purpose before it has widely diffused. Our approach is comparative, ranking different emerging technologies in the three dimensions the literature identifies as defining GPTs: (1) widespread use, (2) ongoing technical improvement, and (3) innovation in application sectors. We find that machine learning and related data science technologies are relatively likely to be GPTs.