Physical Intelligence says its latest VLA model can double robot throughput. Source: Physical Intelligence
While ChatGPT and other generative AI engines can draw conclusions from the entire Web, robots require real-world data and foundation models to transcend more deterministic programming and learn new skills. Physical Intelligence last week raised $600 million in Series B funding as it continues to develop models for robots to understand and interact with the material world.
The San Francisco-based company plans to use the financing to collect more data, make strategic partnerships, and grow its team. Founded in 2024, Physical Intelligence raised $400 million a year ago.
Register now and help us celebrate 20 years of RoboBusiness!Physical Intelligence aims for faster, more reliable robots
With foundation models, AI developers are working to make it easier for robots to learn from a variety of inputs and to generalize behaviors more quickly with smaller amounts of data than previous reinforcement learning (RL) approaches. This has implications for robot performance in unstructured environments, from retail stores to households.
“A typical workflow starts when a developer streams RGB-D camera images from any robot to Physical Intelligence’s runtime,” explained Sacra.com. “The system tokenizes this visual stream along with the robot’s movement history and feeds it to a 3 billion to 5 billion-parameter transformer model. Users can provide plain-language goals like ‘make a flat white’ or ‘pack chocolates into this box.'”
The model takes about 100 ms to predict the next 50 steps, and a hardware abstraction layer converts the tokens into robot-specific joint commands, within force and speed limits for safety, it added.
In February, Physical Intelligence made the code and weights for its π0 or Pi0 robotics algorithms open-source. Earlier this month, the company announced Version 0.6 of the vision-language-action (VLA) model.
The startup used the RECAP (RL with Experience & Corrections via Advantage-conditioned Policies) approach to train a robot by demonstration, coach it through corrections, and improve from autonomous experience.
Physical Intelligence said this doubled throughput on tasks such as inserting a filter into an espresso machine, folding previously unseen laundry, or assembling a cardboard box. It also decreased failure rates over hours of operation and proved superior to imitation learning alone, the company claimed.
Alphabet leads Series B as AI competition continues
CapitalG, Alphabet‘s growth fund, led Physical Intelligence’s Series B round with Lux Capital. Bond, Redpoint, and Sequoia Capital also participated.
“What makes Physical Intelligence’s approach transformative is their approach to universally embodied AI: They are building a single generalist intelligence that manifests in any physical form to solve any real-world problem,” wrote Jill Greenberg Chase and Manmeet Gujral, partners at CapitalG.
Previous and returning investors include Jeff Bezos, the executive chairman of Amazon. OpenAI, Redpoint Ventures, T. Rowe Price, and Thrive Capital have also contributed. Physical Intelligence has raised a total of $1.1 billion to date and is currently valued at about $5.6 billion, according to Bloomberg.
Physical Intelligence is hardly alone in getting millions to pursue physical AI. In September, Dyna Robotics closed a $120 million Series A round to develop its proprietary foundation model.
Earlier this month, Archetype AI raised $35 million for “physical agents,” and Foxglove raised $40 million to scale its data platform for robot developers.
Other companies are also racing to get the data and build the models for next-generation robot AI. For instance, Physical Intelligence partner AgiBot has deployed its Real-World Reinforcement Learning system in a manufacturing pilot with Longcheer Technology.
In addition, 1X Technologies has been using teleoperation to train its NEO humanoid to conduct household chores. Skild AI said it is developing a general-purpose Skild Brain.
Tell Us What You Think!