DeepMind on AI Interpretability Research
Google DeepMind · youtube · 2026-07-10
Google DeepMind's video "Understanding the inner thoughts of AI" discusses the open research direction of AI interpretability. Featuring a conversation between Hannah Fry and researcher Neel Nanda, the core content includes:
- How researchers attempt to "understand" a model's internal mechanisms rather than just looking at inputs and outputs
- Mechanistic interpretability and discoveries about internal model structures, such as sparse autoencoders
- Chain-of-thought monitoring, model auditing, and safety evaluations
- Why interpretability is considered crucial for building safe / aligned / trustworthy AI
- The limitations of current methods and future research directions
This is a long, research-review-style video aimed at understanding and safety capabilities in the AGI era.
Related event: DeepMind Discusses Chain of Thought and Mechanistic Interpretability(4 posts)→
More from Research
- Datology webinar: Curation Studio matched top data baselines with under 20% compute — josh_wills · 2026-10-27
- KANLab: an interactive browser playground for Kolmogorov-Arnold Networks — Lukas Niehaus · 2026-10-27
- CMU's LLM2Cert synthesizes stability certificates verified via SMT and Lean — CarnegieMellonU · 2026-10-27
- NVIDIA open-sources FLORA, a robotics data engine ~125x cheaper than GPT-6 Astra — nvidia · 2026-10-27
- RSNA 2026 Deep Learning Labs return: eight hands-on AI classes over four days in Chicago — PRouzrokh · 2026-10-27
- Steering two orthogonal concept neurons makes a model compose "gnome" — Sauers_ · 2026-10-27