Sergey Levine

Sergey Levine is a co-founder of Physical Intelligence, a company building robotic foundation models — general-purpose models intended to control any physical robot across any task. He is also a professor at UC Berkeley, where his research centres on robotic learning and reinforcement learning.

In the wiki

• Sergey Levine on Robot Foundation Models, Generalisation, and Embodied AI — Invest Like the Best, 31 March 2026: why robotics needs a general foundation model rather than task-specific systems, the tension between genuine generalisation and demo-optimised results, and how chain-of-thought reasoning and reinforcement learning combine to handle long-tail scenarios.

See also

• Patrick O'Shaughnessy — host, Invest Like the Best
• Caitlin Kalinowski on Building Hardware, the Robotics Frontier, and Supply-Chain Reindustrialisation — a hardware-side view of the same robotics frontier
• Andrej Karpathy — a parallel voice on reinforcement learning’s limits and genuine generalisation
• Large Language Models — the architecture Levine’s robotic foundation models adapt from