When AI Understands the World, Not Just the Frame

Most AI platforms still produce convincing pixels without understanding what's actually happening in the scene - physics, continuity, cause and effect. Amit Jain explores how multimodal "world models" close that gap by learning to simulate reality itself, not just imitate it - and why that understanding is what separates a flashy demo from AI reliable enough for real production. He'll also discuss what comes next - extending that same world-understanding from video and creative work into robotics.