

Teaching AI to Understand the Physical World, with Dr. Fei-Fei Li of World Labs
282 snips Jun 5, 2025
Dr. Fei-Fei Li, an AI pioneer and co-director of Stanford’s Human-Centered AI Institute, discusses the intersection of embodiment and intelligence at World Labs. She emphasizes the importance of spatial intelligence in AI and explores the critical gaps in current systems. Li reflects on her role in creating ImageNet, the challenges of 3D modeling, and the significance of building diverse, fearless teams. Her vision for future robotics reveals how human-centric approaches can pave the way for transformative advancements in technology.
AI Snips
Chapters
Books
Transcript
Episode notes
Fundamentals of Spatial Intelligence
- Spatial intelligence means understanding, reasoning, interacting, and generating in 3D worlds.
- It is fundamental to human and animal intelligence and essential for complete AI.
3D Modeling is Hard for Humans
- Spatial intelligence is a hard evolutionary problem as animals must reconstruct 3D worlds from visual input.
- Humans can only build complex 3D models with training; innate 3D mental modeling is limited.
Emotional Intelligence Gap in AI
- Besides 3D and language, emotional intelligence remains a deeply unsolved AI challenge.
- Achieving emotional intelligence is critical for reaching general artificial intelligence (AGI).