This paper analyzes the unexpected success of AlphaFold V1, an AI-based software program representing a publicly available and novel methodological advance within the “protein folding” (PF) subfield of computational biology. As a novel tool that represented a significant advance, AlphaFold V1 influenced the size and the expertise composition of academic research labs already producing similar software in the PF subfield. Specifically, AlphaFold V1 caused the principal investigators of established labs to reconsider the size, and the depth and breadth of expertise necessary to capitalize on the advance; and even to decide whether their labs should continue work in the PF subfield altogether. Results show that impacted labs adapted to the new methodological advance by becoming larger and broader in expertise. This study contributes findings on: (i) the mechanisms of new opportunity and increased competition that shape lab composition; (ii) the complexity of science at the frontier of human expertise and artificial intelligence; and (iii) the role of companies in the collective effort to produce scientific knowledge. The results carry implications for further research on the integration of disciplinary insights, the scaled computational capabilities enabled by artificial intelligence, and the organization of science itself.
Ansprechpartner: Michael E. Rose