A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in AI-powered robotics. The research focuses on enabling robots to learn complex manipulation tasks, such as assembling furniture or preparing meals, from a relatively small number of human demonstrations. The system, which combines computer vision with a new machine learning framework, allows robots to generalize from the demonstrations to handle objects they have not seen before and adapt to minor variations in the task. This approach addresses a key challenge in robotics: acquiring versatile, real-world skills without requiring millions of trial-and-error attempts or extensive, explicit programming for every possible scenario. The technology has potential applications in domestic assistance, manufacturing, and healthcare. For the full details, read the complete article at https://technologyreview.com/2024/05/15/robots-learn-tasks-from-few-demonstrations.