Fei-Fei Li's World Labs explains Atlas: new view prediction may be the next token prediction
a16z · youtube · 2026-09-04
On the a16z podcast, World Labs co-founders Fei-Fei Li, Justin Johnson, and Ben Mildenhall discuss Atlas, their latest world model, and the case for spatial intelligence.
Key points:
- Atlas centers on "new view prediction": given images of a scene, the model predicts how that environment should look from a different position in space and time, unifying generation and 3D reconstruction in one model
- The team frames view prediction as a potential primitive for understanding the physical world — analogous to next-token prediction for LLMs
- They cover the technical bets behind Atlas, what it can and can't yet capture, and why dynamics, editability, and simulation matter as world models mature
- The conversation tackles whether Atlas is just a scaled-up video model or a new architecture, video models vs. world models, and whether walkable 4D video is coming
- Applications span creative work, games, architecture, and robotics; Fei-Fei Li argues access to real-world training data is one of robotics' biggest constraints today
Full timestamps and the World Labs blog post on Atlas are linked in the description.
More from Multimodal
- Voxtral 3 ships a LiveKit-based voice agent SDK streaming listen-think-speak — MistralAI · 2026-10-21
- Voxtral Mini TTS 3 hits SOTA expressiveness in 35+ languages with tag-based control — MistralAI · 2026-10-21
- Mistral launches Voxtral 3 speech models and new Voice Agents API — MistralAI · 2026-10-21
- Open-sourced pixel art normals generator ships as Rust CLI and wasm library — HugoDuprez · 2026-10-20
- Fibo 2.0 is strongest for ecommerce and creator content, weakest at UIUX design — ArtificialAnlys · 2026-10-20
- Fibo 2.0 leads old Fibo on all 9 capabilities, best at reasoning and anatomy — ArtificialAnlys · 2026-10-20