After LLMs: Spatial Intelligence and World Models — Fei-Fei Li & Justin Johnson, World Labs

In this episode of the Latent Space Podcast, Alessio and Swyx interview Fei-Fei Li and Justin Johnson of WorldLabs about world models and spatial intelligence. They discuss the origins of WorldLabs, which grew out of Fei-Fei Li's academic lab where Justin was a student, and the evolution of AI since AlexNet. The conversation explores the balance between open science and proprietary research, the challenges of resourcing academic AI research, and the shift in academia's role due to the increasing scale of AI models. They also delve into the concept of spatial intelligence, its relationship to language models, and the potential of Marble, WorldLabs' 3D world generation model, for various applications like gaming, VFX, robotics training, and design. The discussion touches on the fundamental data structures of world models, the integration of physics engines, and the importance of intellectual fearlessness in pushing the boundaries of AI research.