Daniela Rus on robots learning through play and physical AI

Daniela Rus, director of MIT CSAIL and cofounder of Liquid AI, discusses how robots can learn about the physical world through playful exploration and experimentation, similar to how children develop skills. She explains why simply connecting a language model to a robot is insufficient, and explores research into AI systems that can design physical tools and build safety requirements during training.

The conversation covers autonomous vehicles, how reliable behavior builds trust with users, and Liquid AI's work on efficient neural networks that can run locally on devices rather than relying entirely on cloud computing. Rus also discusses the path from academic research to building companies, and what the field of physical intelligence still needs to mature.