AI revolution and coordination failure: Theory and evidence

Abstract

This paper analyzes theoretically and empirically the coordination failure problem inherent within the 21st-century automation revolution. First, we build a general equilibrium model with labor-saving artificial intelligence (AI) technology that is developed through R&D investments in automation. The model exhibits multiple market equilibria due to a positive feedback loop between AI investments and general economic activities. The available evidence supports our model's predictions regarding the interaction between AI technologies, income inequality, and wages. We also find strong empirical support for multiple equilibria in AI development-the primary prediction of our model. These empirical and theoretical results suggest that AI development can cause coordination failures, thereby creating leaders and followers in automation. However, according to our policy analysis, R&D subsidies and public-private partnerships are efficient coordination devices to tackle this problem.