Research
Nvidia's model GR00T 2 watches humans on video, then imagines itself as the robot
The Decoder · By Maximilian Schreiner ·
The Brief
- Nvidia's new AI model GR00T 2 is designed to teach robots new tasks without laborious teleoperation.
- It transfers human hand movements from videos to robotic arms and was trained on about 400,000 hours of video footage.
- Success rates still vary widely across tasks.
Why it matters
If video-based learning works reliably, robot teams could train new manipulation tasks from recorded human demonstrations instead of costly teleoperation sessions. The wide variation in success rates means it is not yet a dependable replacement.
Who’s affected
- Robotics teams working on manipulation tasks
- Developers building on Nvidia's robotics models
- GR00T 2
- #robotics
- #machine-learning
- #nvidia
- #imitation-learning
- #foundation-models
Nvidia's new AI model GR00T 2 is designed to teach robots new tasks without laborious teleoperation. Instead, the system transfers human hand movements from videos to robotic arms, trained on about 400,000 hours of video footage. Success rates still vary widely, though, ranging…