A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in robotic manipulation. Researchers have developed a system that allows robots to learn complex, dexterous manipulation tasks, such as spinning a pen between their fingers, using a more data-efficient method. The approach, known as ‘Diffusion Policy,’ leverages diffusion models—a type of generative AI—to translate camera observations directly into precise robotic actions. This method shows improved performance and generalization over previous techniques, requiring less specialized data. The research indicates progress toward robots that can perform a wider array of delicate, human-like tasks in unstructured environments. Read the full article at https://technologyreview.com/2024/05/20/1093035/robot-learns-pen-spinning-diffusion-policy/.