Nvidia’s GR00T 2 learns robot tasks from human video instead of teleoperation
🤖 AI-generated content — The title and summary were produced automatically by artificial intelligence, without human editorial review.
Nvidia’s GR00T 2 foundation model trains robotic arms by transferring human movements from video rather than hours of remote control. As a world-action model it predicts the next movement and a short future video; training used about 400,000 hours of head-mounted camera footage on a cluster of more than 1,000 GB300 chips. Reported success rates range from 10% to 80% by task. New skills can be taught via demonstrations with data gloves, specialized grippers or a finger exoskeleton.
Continue in the app — vote & join in ➔Source: The Decoder (AI) · via ahirlevel.hu