A Comparative Cognition Approach to Evaluating AI Capabilities

Explore a comparative cognition approach to evaluating AI capabilities in this 45-minute lecture by Lucy Cheke from Cambridge University. Delve into the parallels between understanding nonhuman animal behavior and assessing AI systems, both of which present challenges in peering inside the "black box" of cognition. Examine how psychologists' century-long experience in understanding and predicting behavior can be applied to AI evaluation. Investigate the phenomenon of AI systems exhibiting clever-seeming behavior while being prone to inflexibility and shortcuts. Learn about the benefits of robust cognitive testing of AI, not only for understanding artificial intelligence itself but also for gaining insights into biological intelligence. This talk, part of the "Understanding Higher-Level Intelligence from AI, Psychology, and Neuroscience Perspectives" series at the Simons Institute, offers a unique perspective on the intersection of comparative cognition, AI capabilities, and the study of intelligence across different domains.