The episode examines a study from MIT involving a sample of 1,018 scientists working in a large American company. The research investigates the impact of an artificial intelligence system based on graph neural networks in the development of new materials. The findings reveal a significant increase in productivity and patent generation but also highlight a growing gap between more experienced and less experienced researchers. While the former nearly double their productivity, the latter gain only minimal benefits, raising questions about the need for targeted training and a reorganization of workflows. The study underscores the critical role of human domain knowledge in interpreting AI-generated results and explores the implications for job satisfaction and workforce management. Additionally, it delves into the generalizability of these findings to other scientific fields and the strategic challenges these technologies pose for businesses.