World Labs' Atlas: new-view prediction rebuilds 3D spaces from 3 photos, 50-100x cheaper
drfeifei · x · 2026-09-06
In an a16z conversation, World Labs co-founders Fei-Fei Li, Justin Johnson and Ben Mildenhall detail Atlas, a world model for spatial intelligence.
- Core thesis: LLMs rest on next-token prediction and video models on next-frame prediction; Atlas is built on generative new-view prediction, which Johnson argues is also AI-complete — solving it in full generality would solve any intelligence problem.
- Technical milestone: Atlas is the first model to unify pixel generation and pixel reconstruction, two problems computer vision has kept separate for half a century.
- Practical payoff: digitally capturing a 3D space previously needed 100-300 photos per room; Atlas works from just three, a 50-100x cost reduction.
- The discussion also touches on implications for robotics and why the team emphasizes spatial reasoning.
More from Companies & People
- Hugging Face employee shares his path: Argilla acquisition to the smol course series — ben_burtenshaw · 2026-09-06
- Before Altman's Ouster, OpenAI's Board Was Divided and Feuding, NYT Reports — GarrisonLovely · 2026-09-06
- Hossenfelder Exposes Undisclosed 'Grants' Funding Anti-AI Fear Among Influencers — mallow610 · 2026-09-06
- Waymo vehicle spotted in Pittsburgh, hinting at city expansion — chris_j_paxton · 2026-09-06
- AI influencers complain: praising OpenAI is free, loving Google gets you called a shill — ChrisUniverse · 2026-09-06
- SaaS survived, but AI stole its budget: only ~20% of AI spend is net new — AccBalanced · 2026-09-06