AI's Models of the World, and Ours

Jon Kleinberg

International Conference on Machine Learning 2025 · Invited Talk

Jon Kleinberg, a renowned professor from Cornell University, delivered a thought-provoking talk at ICML 2025, delving into the profound implications of AI's burgeoning capabilities, particularly its development of "models of the world." The presentation explored how machine learning, having long surpassed human performance in domains like chess, is now grappling with challenges ranging from the emergence of **algorithmic monoculture** to the intricate dynamics of human-AI collaboration and the fundamental theoretical underpinnings of language generation. Kleinberg emphasized that understanding these new frontiers requires not only empirical investigation but also a return to foundational computational theory, offering abstract models to reason about AI's behavior and its impact on human endeavors.

AI review

Kleinberg delivers a coherent and intellectually ambitious talk that weaves together empirical observations from chess AI, collaborative system design, and a genuinely interesting theoretical result about language generation. The crown jewel is the theorem with Mullainathan showing generation is achievable in the limit where identification (Gold 1967) is not — that is a clean, non-trivial separation between two fundamental primitives, and it's the kind of result the theory community should pay attention to. The chess material and partner-bot experiments are engaging but function more as…