Veo
Veo is Google’s video-generation model family, discussed in [src-063] as evidence that generative video systems can learn surprising amounts of physical structure from passive visual data.
Key facts
- Type: AI video generation model family
- Maker: Google / Google DeepMind
- Source role: Example of Intuitive Physics In AI and a possible route toward World Models [src-063]
- Hassabis points to video generation that renders liquids, materials, lighting, and short physical interactions as evidence that models may be extracting lower-dimensional structure from observed reality [src-063].
- The episode connects Veo-style progress to interactive video, playable simulations, and future AI-generated game worlds [src-063].
What it adds
Veo links the wiki’s model/product layer to simulation. In this source, video generation is not only media creation; it becomes a probe into whether models can learn how the world behaves well enough to support interaction, games, and AGI-relevant world modeling [src-063].
Related entities
Related concepts
Source references
- [src-063] Lex Fridman – “Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475” (2025-07-23)