What Physical Intelligence does
Physical Intelligence (π) develops general-purpose vision-language-action (VLA) foundation models intended to control robots across many embodiments and physical tasks. The company’s stated goal is to build a single model that can “control any robot to do any task,” and it describes its approach as combining web-scale vision-language pretraining with robot data collection and post-training methods (including reinforcement learning and experience/correction loops) to improve real-world dexterity, reliability, and generalization. The company also positions this capability as a reusable “Physical Intelligence Layer,” analogous to how developers consume APIs for language/image models—so that robotics teams can integrate the models into their own systems rather than building a full physical stack from scratch.