Conversation · NVIDIA
Intro to NVIDIA Cosmos

Ming-Yu Liu
VP of Research, NVIDIA
49 minWith Kim IsenbergAn NVIDIA Cosmos Labs session featuring Kim Isenberg
Watch on YouTubeAbout this conversation
A deep dive into NVIDIA Cosmos with its creator Ming-Yu Liu, Vice President of Research at NVIDIA, featuring Kim Isenberg of Superintelligence.
The conversation unpacks the motivation behind Cosmos, the science and research powering its open models, and the long-term vision: how Cosmos is redefining physical AI, where the models are headed next, and how developers and researchers can start using them today.
Topics
- Why NVIDIA built Cosmos and the problems it is designed to solve
- The core science and model architecture behind Cosmos
- How open models are accelerating innovation in physical AI
- Selected use-case demos in real systems
Chapters
- 0:00Introduction
- 1:01NVIDIA Cosmos Overview
- 3:03Live Stream Kickoff
- 6:12Limitations of Physical AI
- 9:06Cosmos Model Components
- 13:41Training the System
- 18:55Community Q&A Part 1
- 20:48Open Source Importance
- 24:19Learning from the Community
- 26:52Community Q&A Part 2
- 29:35Current Status and Limits
- 32:03Future Vision and Cookoff
- 35:34Final Q&A and Closing
- 47:44Stream End
Kim Isenberg is co-founder and Editor-in-Chief of Superintelligence, an AI newsletter published every weekday.
Read the NewsletterMore interviews
- Sven BreunerThe hidden data bottleneck in AI factories
- Akshay KothariNotion’s AI agent workspace
- Roeland DelrueAI agents and the new cybersecurity crisis