A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates that large language models (LLMs) can be used to generate complex robotic task plans directly from natural language instructions, bypassing traditional intermediate coding steps. The research team developed a system where an LLM, given a high-level command like ‘make a sandwich,’ can autonomously break it down into a sequence of executable actions for a robot, such as locating bread, picking up a knife, and spreading condiments. This approach significantly reduces the need for manual programming of each sub-task. The system was tested in simulated kitchen environments, showing promising results in handling multi-step objectives with varied objects. However, challenges remain in ensuring precise physical execution and adapting to unexpected real-world conditions. The work points toward a future where robots can more flexibly interpret and act on human intent. Read the full article at: https://technologyreview.com/2023/09/llm-robotic-task-planning