Physical Intelligence

Physical Intelligence is a San Francisco-based AI research company developing general-purpose foundation models for robotics, aiming to give robots broad physical intelligence and dexterity across many different tasks and hardware platforms.

Physical Intelligence was founded in 2024 by a team including Karol Hausman, Sergey Levine, Brian Ichter, and other researchers with deep robotics and machine learning research backgrounds from institutions including Google DeepMind and UC Berkeley, pursuing a research agenda focused on building general-purpose "foundation models" for robotics, conceptually similar to how large language models like GPT have been trained on enormous amounts of text data to develop broad, general-purpose language understanding and generation capabilities, but applied instead to physical robotic manipulation and control, with the ambitious goal of developing AI models that can generalize across a very wide range of different physical tasks, environments, and even different robot hardware platforms, rather than requiring separate, narrowly trained models for each specific robotic task and hardware configuration. The company's research addresses a significant historical limitation in robotics AI, where models trained to perform a specific manipulation task on a specific robot typically generalize quite poorly to different tasks or different robot hardware, requiring extensive additional training data and effort for each new task or robot platform, a limitation that has significantly slowed the pace of practical robotics deployment relative to the more rapid recent progress seen in language and image-based AI models that have achieved much broader generalization capabilities. Physical Intelligence has released research and demonstrations of its foundation models performing a range of physical manipulation tasks across different robot hardware platforms, positioning its work as an important step toward more broadly capable, generalizable robotic AI that could eventually support a much wider range of practical robotics applications without requiring bespoke AI development for each specific use case and robot configuration. The company has attracted very substantial investor interest and funding despite still being firmly in a research and early product development phase, reflecting significant broader investor enthusiasm for foundational robotics AI research given its potential long-term implications across manufacturing, logistics, home robotics, and many other physical task domains. Physical Intelligence has raised very substantial venture funding from investors including OpenAI, Jeff Bezos, Thrive Capital, and Lux Capital, reaching a valuation exceeding $2.4 billion within its first year of operation.

News

No news available for this company.