A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates that large language models (LLMs) can be used to automatically generate and execute complex, multi-step robotic manipulation plans. The research team developed a framework where an LLM, given a high-level natural language command, breaks it down into a sequence of smaller, executable sub-tasks. This system successfully performed tasks like sorting objects by color and shape, and constructing complex structures from various blocks, without requiring task-specific training data. The approach represents a significant step toward more flexible and general-purpose robotic systems that can understand and act upon open-ended human instructions. Read the full article at https://technologyreview.com/2023/10/llm-robot-planning.