Physical Intelligence

This talk explores the advancements in robotics driven by the emergence of vision-language action (VLA) models. It highlights the transition from robots operating in highly constrained environments to performing complex tasks in semistructured real-world settings. The core thesis is that significant progress in robotics is now enabled by general AI developments, particularly VLA models, and that the primary bottleneck is shifting from hardware to software and model intelligence.