Spatial Intelligence in AI

Current large language models understand text and patterns, but they lack spatial reasoning, the ability to mentally model 3D space, physics, movement, and how objects interact in the real world. AI pioneer Fei-Fei Li argues this is a critical gap: humans naturally understand that a cup on a table doesn't fall through, or that walking backward into a wall is blocked. Without spatial intelligence, AI struggles with robotics, AR/VR, autonomous vehicles, and physical reasoning. The next generation of AI needs what neuroscientists call 'embodied cognition', understanding the world through physical simulation.