Scientists from MIT’s Computer Science and Artificial Intelligence Laboratory used AI-generated images to train a robot dog in parkour, without real-world data.
Parkour is a non-competitive physical discipline where participants move through an environment using their bodies to overcome obstacles – basically like navigating an obstacle course.
CSAIL’s LucidSim system demonstrates the potential of GenAI – generative AI, or generative AI – for creating relevant robotics training data, enabling expert-level performance in complex tasks like obstacle navigation and stair climbing. [Read more…] about Can robots learn from ‘machine dreams’?