Subvolts take Google DeepMind explores how "thinking time" and long context improve AI models. Feb 24, 2025 - Developers
- Learn about techniques like "thinking time" and long context to improve your AI models' performance and reasoning.
- Business
- Understand how Google DeepMind is advancing AI capabilities, potentially leading to more sophisticated AI agents and tools.
This is a solid interview with a principal scientist discussing core AI model development concepts at a leading lab.
Developing Google DeepMind's Thinking Models
Audio streams from Google AI. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Google AI: Release Notes host Logan Kilpatrick talks with Jack Rae, Principal Scientist at Google DeepMind, about developing "thinking models." They discuss practical uses, how more "thinking time" improves performance, and the importance of long context for these models. The episode also touches on developer feedback and the future of AI model evaluations.
This episode explains how Google builds AI that can "think" better. By giving AI more time to process information and allowing it to consider longer pieces of text, they make its answers more accurate and useful for various tasks.
This episode offers insight into Google DeepMind's approach to building more capable AI models, specifically "thinking models." It's for AI researchers and developers interested in the practicalities of model development, including the impact of computational resources like "thinking time" and architectural features like long context. This work aligns with the broader industry trend of developing more sophisticated AI systems that can reason and plan, a key focus for all frontier labs.
- HostLogan Kilpatrick
- GuestJack RaePrincipal Scientist, Google DeepMind
- Guest: Jack Rae, Principal Scientist at Google DeepMind
- Show: Google AI: Release Notes
- Length: 64 minutes
- Key Concept 1: Thinking Models
- Key Concept 2: Thinking Time Improves Answers
- Key Concept 3: Long Context Is Key
Jack Rae, Principal Scientist at Google DeepMind, joins host Logan Kilpatrick for an in-depth discussion on the development of Google’s thinking models. Learn more about practical applications of thinking models, the impact of increased 'thinking time' on model performance and the key role of long context.
01:14 - Defining Thinking Models
03:40 - Use Cases for Thinking Models
07:52 - Thinking Time Improves Answers
09:57 - Rapid Thinking Progress
20:11 - Long Context Is Key
27:41 - Tools for Thinking Models
29:44 - Incorporating Developer Feedback
35:11 - The Strawberry Counting Problem
39:15 - Thinking Model Development Timeline
42:30 - Towards a GA Thinking Model
49:24 - Thinking Models Powering AI Agents
54:14 - The Future of AI Model Evals
- Thinking Models
- Jack Rae
- AI systems designed to improve their reasoning and problem-solving abilities.
- Thinking Time
- Jack Rae
- The amount of computational time an AI model spends processing information before giving an answer.
- Long Context
- Jack Rae
- The ability of an AI model to process and remember information from very long pieces of text or data.
27 minKoray Kavukcuoglu on frontier models, coding agents, and building AGI
WithKoray Kavukcuoglu
39 minIntroducing Gemini Robotics 2
WithCarolina Parada, Stuart Bowers, Kanishka Rao, Jie Tan
42 minGemini co-leads on project origins and what's next
WithJeff Dean, Koray Kavukcuoglu, Noam Shazeer, Oriol Vinyals
40 minQuestions people ask
- What are Google's "thinking models"?
- The episode discusses the development of Google's "thinking models," which are AI systems designed to improve performance through increased "thinking time" and long context.
- How does "thinking time" affect AI performance?
- Increased "thinking time" is presented as a method that improves the quality of answers provided by AI models.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Google AI.