🤖 Fei-Fei Li Proposes Dividing World Models into 3 Functions
The co-founder of World Labs has proposed classifying world models based on three functions: renderers (visual fidelity), simulators (physical accuracy), and planners (action output). This will allow a transition from purely text-based models to an understanding of the spatio-temporal structure of reality.
🌍 This division helps structure research in the field of Spatial Intelligence, separating visual video generation from the creation of physically accurate digital twins for training Embodied AI.
👤 This is the foundation for creating truly autonomous robots and interactive 3D worlds that understand the laws of physics.
Source 1: https://drfeifei.substack.com/p/a-functional-taxonomy-of-world-models Source 2: https://drfeifei.substack.com/p/from-words-to-worlds-spatial-intelligence