Is AI a General Purpose Technology?

Abstract

A general purpose technology (GPT) is an exceptional link in a complex chain of innovations connected through space and time. Common examples include the steam engine, electrification, and the automobile (Bresnahan & Trajtenberg, 1995; Bresnahan & Trajtenberg, 1995; & Bekar, et. al., 2018). A GPT is set apart from other innovations because of the massive breadth, depth, and duration of their impact on our world (Trajtenberg, 2019; Bekar, et. al., 2018; Strohmaier & Rainer, 2016; Ott, et. al., 2009 & Bresnahan & Trajtenberg, 1995). They are worthy of special consideration. The changes GPTs bring can take decades and even centuries to manifest and, thus, research remains post hoc (Bresnahan & Trajtenberg, 1995; Lipsey, Bekar, & Carlaw, 2005; Jovanovic & Rousseau, 2005). Yet executives, entrepreneurs, researchers, policymakers, and investors would benefit from foresight anticipating a GPT. This research explores this empirically through the tracking of the ‘pervasiveness dimensions’ of a candidate GPT before its full potential economic and societal impact is felt. To this end, our candidate GPT is Artificial Intelligence (AI) and the key enabling technology of Deep Learning (DL). We trace the evolution of AI and DL over the past fifty years with a particular focus on the critical period AI/DL’s diffusion from 2000 to 2020 (Goldfarb, et. al, 2021). We are aided by a rich base of prior GPT research, research in complementary disciplines, and industry analyses and data. We hope to provide the practical benefit of adding to the decision-maker’s toolkit, while also providing a deeper understanding of AI not just as an innovation but as an extraordinary technological, economic, social, and political phenomenon.