Jack Rae (Google DeepMind) on Gemini 2.5 Pro, Reasoning Scaling & Path to AGI
Jack Rae, Principal Research Scientist at Google DeepMind and technical lead on Gemini's thinking/inference-time scaling work, joins The Cognitive Revolution to discuss Gemini 2.5 Pro—a release the host describes as the first time a Google model ranks #1 across many important dimensions, particularly for long-context command (he tested it with a 400K-token research codebase). Rae attributes Gemini 2.5 Pro to a full Gemini team effort spanning pre-training, thinking, and post-training, and notes that long-context capability and deeper thinking/reasoning have stacked unusually well together to unlock new problems.
Original sources
Podcast The Cognitive Revolution
Episode:Scaling "Thinking": Gemini 2.5 Tech Lead Jack Rae on Reasoning, Long Context, & the Path to AGI ↗
YouTube The Cognitive Revolution
Video:Scaling "Thinking": Gemini 2.5 Tech Lead Jack Rae on Reasoning, Long Context, & the Path to AGI ↗
AI Radar wrote this summary from the publisher’s notes with AI, not from a transcript. How it works.