Nvidia has introduced the GR00T 2 AI model for training robots

Nvidia has unveiled a new AI model, GR00T 2, designed to train robotics without the need for complex teleoperation.

The system transfers human hand movements directly from video footage to robotic manipulators. To train the model, developers used a dataset consisting of approximately 400,000 hours of video recordings.

Task success rates range from 10% to 80%, depending on the complexity of the specific task assigned to the robot.

This new development aims to simplify the process of programming and adapting robots for performing practical manipulations in real-world environments.