David Bau's Cornell Tech talk: interpretability is the human-centered second path for AI
TuhinChakr · x · 2026-11-11
David Bau (Northeastern University) spoke at the Cornell Tech Learning Machines Seminar on "Interpretability, Indirection and Introspection"; the recording is on YouTube. Yoav Artzi praised the work and highlighted its refreshingly human-centered framing.
The slide's core argument: as AI capability grows there are two paths — an "RSI monoculture" where AI eclipses and replaces human cognition, or the "Interpretable AI path" where humans gain insight into what AI knows and agency over how it works. As AI capability grows, human agency should grow with it.
Related event: David Bau discusses interpretability and introspection at Cornell Tech(3 posts)→
More from AGI Musings
- a16z: agents now make up the majority of web traffic — welcome to the B2A economy — evisoft · 2026-11-12
- Robotics researchers in denial about the bitter lesson: "if it can make slides, it's already a VLA" — keerthanpg · 2026-11-12
- AI 2027 authors defend SAR-by-2027 forecast as conditional on no slowdown — eli_lifland · 2026-11-12
- Why mathematicians panic: AI is inflating away their reputation currency — virtuolie · 2026-11-12
- Investor: AI diffusion will be slowest in consumer software — Oura, Strava, Labcorp still have no real AI strategy — MartinGTobias · 2026-11-12
- On AI Progress: When Models Lag on Your Load-Bearing Skill, You Can't See Them Improving — dfrsrchtwts · 2026-11-12