How adoption of AI differs from the adoption of IT
In this paper Babina et al. develop a a novel measure of AI investments which is based on detailed data on firms’ human capital. The underlying assumption is that AI relies more heavily on human expertise rather than physical capital (as shown in the figure below). They use job postings data from Burning Glass Technologies, which include the near-universe of online job vacancies, and CV data from Cognism Inc, which offer job histories for hundreds of millions of individual employees. By combining these two datasets, they construct firm-level measures of AI investments by observing both the stock and the hiring of AI-related employees. Their results highlight important differences between the adoption of AI and the adoption of information technology (IT) in the 1980s and 1990s:
- While IT investments were associated with economically large productivity increases but mixed results on firm performance measures such as market share, AI-investing firms see growth in market share but not in productivity;
- Rapid adoption of IT was driven by small firms, by contrast it is large firms that are more likely to invest in AI. These new technologies lower the overhead costs of spanning multiple markets and allow the most productive firms to expand;
- Finally, they find a positive correlation between AI investments and an increase in industry concentration.
Artificial Intelligence, Firm Growth, and Industry Concentration
Authors: Tania Babina, Anastassia Fedyk, Alex Xi He, James Hodson
From: Columbia University, University of California, Berkeley, University of Maryland, AI for Good Foundation
