Deep Geography: Implications of the Socio-Spatial Structure in Artificial-Intelligence Research for Financial Institutions

Abstract

As a leading branch of artificial intelligence, deep learning promises competitive differentiation and more efficient operations for financial institutions. But the geographic and social structure of the deep-learning community complicates access to its most cutting-edge research for many financial institutions. We explore how this socio-spatial structure presents obstacles for incumbent institutions, which disadvantage them against new entrants to financial markets – from whom incumbent institutions are facing threats of disruption. We also probe how geographic distribution of deep-learning research affects the hegemony of the New York-London axis in financial innovation. Consequently, we find a need to revisit how the importance of financial centers is measured and viewed in economic geography. Finally, we investigate how ‘at-a-distance’ involvement by incumbent financial institutions in deep learning generates black-box risks that could harbor significant ramifications for both regulatory policies and systemic stability.