A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in robotic dexterity. Researchers have developed a system that allows a robot equipped with a simple two-fingered gripper to manipulate a wide variety of objects with unprecedented skill, including items it has never encountered before. The system, which combines a neural network trained on a massive dataset of simulated objects with real-world reinforcement learning, enables the robot to perform complex, human-like tasks such as opening a wine bottle, placing a phone on a charging pad, and using scissors to cut paper. This approach marks a shift from training robots for single, specific tasks to developing general-purpose manipulation capabilities, bringing machines closer to performing useful work in unstructured environments like homes and warehouses. The full details of the research are available in the published article: https://technologyreview.com/2024/07/12/1094755/robot-dexterity-mit-csail/