Get Timothy Lillicrap on Dwarkesh Podcast

Timothy Lillicrap is a research director at Google DeepMind who is now publicly outlining how reinforcement learning can teach deep reasoning to large language models, a frontier capability reshaping AI in 2025 and 2026. In October 2025 he gave a talk at Yale’s Wu Tsai Institute on “learning to think,” detailing methods to make models reason more deeply about games, math, and real-world problems.

This matters today because reasoning models are the new battleground in AI and fresh research shows smarter, compute-aware ways to scale thinking at inference time, which could change costs and capabilities overnight. With DeepMind consolidating Google’s AI stack and its leadership signaling a five to ten year AGI horizon, Lillicrap can give an insider’s read on world models, planning, and how quickly these breakthroughs will hit products.

He has not appeared on the Dwarkesh Podcast before according to the show’s episode listings, so this would be a first-time, high-signal conversation for that audience.

Support this nomination now so we can hear timely, unfiltered insight from the person shaping how machines learn to think.