World models need causality, not pretty pixels — Christopher Manning on embodied simulation
AI Engineer · youtube · 2026-09-24
Stanford professor Christopher Manning (now at Moonlake AI) tours AI history — from Dartmouth 1956 to a 2007 Google LM trained on 2T tokens — and argues the next step is embodied intelligence via simulation. Generative video like Genie 3 simulates observations without semantics and can't support planning; Moonlake instead builds action-conditioned world models in code from a single photo, researching objects on the web to fill invisible details, with a Claude Code-inspired loop comparing renders to reality to close the sim-to-real gap and replace 10,000 hours of teleoperation.
More from Embodied
- Duke's Cartesian Hand manipulates with seven prismatic joints and zero rotation — lukas_m_ziegler · 2026-09-24
- Waymo's 270M miles show 82% fewer injury crashes than human drivers — DMaguireARK · 2026-09-24
- Developer: Vision Pro's privacy stance frustrates devs but its value is undeniable — dreamwieber · 2026-09-24
- Robotics contract manufacturer RMFG hiring Head of Factory Software, already cash-flow positive — seanmcdonaldxyz · 2026-09-24
- Two robot arms assemble an engine part with Opus 5.5 handling planning and coordination — ZeYanjie · 2026-09-24
- After Meta's keynote, rjs sees the industry K-shaping—and Apple behind on multiple fronts — _AustinCalvert_ · 2026-09-24