Meet the Robot Trained on 100 Million Internet Videos
Rhoda AI is developing a new approach to robotics training, utilizing its Direct Video Action (DVA) model, which learns physics and real-world interactions from over 100 million internet videos rather than traditional lab data. This method allows robots to adapt to unexpected situations and perform complex tasks like decanting and sorting, with some tasks requiring only 10-20 hours of robot-specific training after initial video-based learning. The company plans to enhance the model's reasoning capabilities to predict and select optimal future actions.
This training method could significantly accelerate robot development and adaptability, making robots more versatile in manufacturing and other industries by enabling them to learn from a wider range of real-world scenarios.