A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates that large language models (LLMs) can significantly accelerate the process of robot training for complex manipulation tasks. Traditionally, training robots to perform precise, multi-step tasks like packing items or assembling components requires extensive, time-consuming programming and simulation. The research shows that by using LLMs to generate synthetic training data—detailed descriptions of tasks and potential solutions—robots can learn new skills much faster and with less human intervention. This approach, which the researchers call ‘programmatic reinforcement learning,’ allows robots to practice and learn from a vast array of simulated scenarios before attempting the task in the real world. The method has shown promise in improving a robot’s ability to generalize and adapt to new, unseen variations of a task. For the full details on this breakthrough in robotic learning, read the complete article at https://technologyreview.com/2024/03/llm-robot-training.