Four ETH embodied-AI lectures in 3h17m: what Malik, Koltun, LeCun and Song say models still miss
le_james94 · x · 2026-10-13
lejames94 watched all four ETH embodied-AI lectures from 29 May (3h17m total) — Jitendra Malik (vision and robotics), Vladlen Koltun (spatial cognition), Yann LeCun (world models), and Shuran Song (manipulation) — with one takeaway per speaker in follow-up posts.
His summary graphic: what each says a model still misses before it can use a hand — Malik: language is the easy interface, physical interaction is the real intelligence problem; Koltun: be empirical, not ideological — being wrong about the path is how you find a better one; LeCun: predict what can be predicted and throw away the rest; Song: collect the smallest thing that still carries contact and make it portable. The thread's throughline: two gaps in sequence — representation, then information.
Related event: ETH's Four Embodied AI Lectures: Malik, Koltun, LeCun Each Point to a Gap(3 posts)→
More from AGI Musings
- Economist Josh Gans: Writing PowerPoint slides is thinking — joshgans · 2026-10-13
- SSI's Aschenbrenner: frontier AI pacing is game-theoretically unstable, acceleration is inevitable — beffjezos · 2026-10-13
- Math debate follow-up: burden of proof lies with whoever makes the claim — _onionesque · 2026-10-13
- X debate: does human math still matter when AI can prove theorems? — _onionesque · 2026-10-13
- minilek urges policymakers to take AI risks seriously; Ilya Sutskever replies 'Yet!' — minilek · 2026-10-13
- Researcher predicts truth and morality will become mathematical artifacts on the path to superintelligence — ryunuck · 2026-10-13