How robots learn to understand and act in the physical world

A panel at the Imagination in Action Silicon Valley AI Summit explores the foundations of physical intelligence with Stanford researchers Gordon Wetzstein and Marco Pavone. The discussion examines how robots perceive their environment, learn from data, and translate observations into safe, reliable action.

Topics include world models, simulation-based training, tactile sensing, and agentic AI systems. The researchers address why a powerful model alone isn't enough for useful robotics, and where humanoid robots might deliver real-world value.