Moonlake AI: World Models Need Causality and Objects, Not Pretty Pixels
chrmanning · x · 2026-10-03
Chris Manning, Distinguished Member of Technical Staff at Moonlake AI, argues in his talk "World Models Need Causality, Not Pretty Pixels" that photorealistic video walkthroughs are not world models:
- Pretty pixels aren't a world model: he cites Genie 3 (demoed by Riley Goodside) and Marble — both produce convincing video, but neither encodes what objects are or how you can act on them.
- Objects must be separated from background: Moonlake's simulations split scenes into a background plus foreground objects rather than raw generated video.
- Why code and objects matter: building simulations this way yields causal, actionable structure, which is what you need to extract a robot policy from a simulator.
A clear counterpoint to the "generated video as world model" approach.
More from Embodied
- ETH's Anymal robot completes 10km autonomous alpine hike with VLM-guided ROAM system — ChongZzZhang · 2026-11-19
- iFixit tears down Apple HomeView: lots of glue and 'endless frustration' — Resident-Fall4351 · 2026-11-19
- Removing human video drops robot OOD success from 60% to 7%: generalization comes from human video — gui_penedo · 2026-11-19
- PSYONIC's prosthetic Ability Hand now used by 90+ robotics orgs, raising Series A — DhowCIO · 2026-11-19
- Simulator misspecification breeds overconfidence: sim-to-real study on SPEED+ — Nikolay Dolgov · 2026-11-19
- Doppel world model renders the robot body and beats action-only baselines across 8 embodiments — mangahomanga · 2026-11-19