Ming-Yu Liu on Open Models, World Models, and Physical AI at NVIDIA Cosmos
In this episode of Practical AI, hosts Daniel Whitenack and Chris Benson interview Ming-Yu Liu, Vice President of Cosmos Lab at NVIDIA. Liu discusses the evolving physical AI landscape as artificial intelligence expands into robots, vehicles, and other physical systems. The conversation examines why open models are essential for robotics and commercial innovation, as well as the fundamental role of world models. Liu details the distinction between rule-based and learned simulation and explains how world models are used to test policies. Additionally, the discussion touches on multi-agent physical systems, upcoming technical challenges, and how practitioners can get started with open models.
Original sources
YouTube Practical AI
Video:Open Models, Physical AI, World Models, Robotics, NVIDIA Cosmos & AI Agents |Episode 374| ↗
AI Radar wrote this summary from the publisher’s notes with AI, not from a transcript. How it works.