The Spatial Intelligence Value Chain
by Dr. Fei-Fei Li • Co-founder & CEO at World Labs, Co-Director of Stanford HAI at World Labs / Stanford University
Known as the 'Godmother of AI,' Dr. Li is the creator of ImageNet, the dataset that sparked the modern deep learning revolution. She previously served as Chief Scientist of AI/ML at Google Cloud and is a pioneer in computer vision, spatial intelligence, and human-centered AI.
🎙️ Episode Context
Dr. Fei-Fei Li traces the arc of AI development from the 'AI Winter' to the current generative explosion, detailing how her creation of ImageNet shifted the industry's focus from algorithms to data scale. She introduces her new venture, World Labs, and the concept of 'World Models'—AI that possesses spatial intelligence and understanding of physics—arguing this is the missing link for robotics and true AGI. The conversation also covers the necessity of human-centered design and the importance of 'intellectual fearlessness' in career pivots.
Problem It Solves
Bridging the gap between generative AI (text/2D images) and functional, embodied AI (robotics/simulation) that can interact with the real world.
Framework Overview
A framework for building 'World Models' that allows AI to reason, interact, and create in 3D space. Unlike LLMs which predict tokens, this approach models physics and geometry to create actionable environments.
🧠 Framework Structure
Input Versatility (Prompt-to-World): ...
3D/4D Reasoning: The model must infer...
Interactability Check: The output mus...
Cross-Domain Application: Validate th...
When to Use
When building products for robotics, gaming, simulation, or any domain requiring physical reasoning rather than just language generation.
Common Mistakes
Confusing video generation (2D pixels changing over time) with world generation (consistent 3D physics and geometry).
Real World Example
World Labs' product 'Marble,' which generates infinite 3D worlds from a single prompt, allowing users (and eventually robots) to navigate and interact within them.
Spatial intelligence to me is the ability to create, reason, interact, make sense of deeply spatial world... World Lab is focusing on that, and of course the ability to create videos per se could be part of this... but we really want creators... to have in their hands a model that can give them worlds with 3D structures.
— Dr. Fei-Fei Li