AI startup secures $400M to develop ‘physical AI’

Groundbreaking AI robotics startup secures major funding: Physical Intelligence, a San Francisco-based company developing general-purpose AI for robotics, has raised $400 million in a funding round that values the company at $2.8 billion.

• The investment round attracted high-profile backers including Amazon founder Jeff Bezos, OpenAI, Thrive Capital, Khosla Ventures, Lux Capital, Sequoia Capital, and Bond Capital.
• Physical Intelligence aims to bring general-purpose AI into the physical world by developing foundation models and learning algorithms for robots and other physically actuated devices.
• The company’s long-term goal is to develop artificial physical intelligence, enabling users to simply ask robots to perform any task, similar to how they interact with large language models and chatbot assistants.

Key innovation: The π0 (pi-zero) foundation model: Physical Intelligence has developed a general-purpose foundation model called π0, which represents a significant step toward achieving artificial physical intelligence.

• Unlike traditional large language models (LLMs), π0 spans images, text, and actions, acquiring physical intelligence by training on embodied experience from robots.
• The model can control various robots and can be either prompted to carry out desired tasks or fine-tuned for specialized applications.
• π0 is trained on broad and diverse data, allowing it to follow various text instructions while directly outputting low-level motor commands through a novel architecture.

Research findings and potential applications: Physical Intelligence recently released a paper exploring the potential of robot learning to enhance flexible, general-purpose, and dexterous robotic systems.

• The research aims to address key challenges in robotics, including data acquisition, generalization, and robustness.
• The company’s approach utilizes “robot foundation models” to overcome these obstacles, potentially revolutionizing the field of robotics.
• To test their architecture, Physical Intelligence had robots perform various tasks, including folding laundry, cleaning tables, and assembling boxes, demonstrating the versatility of their AI-powered system.

Implications for the robotics industry: The development of general-purpose AI for robotics could have far-reaching consequences for various sectors, from manufacturing to home automation.

• If successful, Physical Intelligence’s technology could significantly reduce the need for task-specific programming in robotics, making robots more adaptable and easier to deploy in diverse environments.
• The ability to control robots through natural language commands could make advanced robotics more accessible to non-expert users, potentially accelerating the adoption of robots in everyday life.
• However, the development of such advanced AI systems also raises important questions about safety, ethics, and the potential impact on the job market that will need to be addressed as the technology progresses.