A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates that large language models (LLMs) can significantly improve the planning capabilities of robots. The research, detailed in a paper published on arXiv, shows that LLMs can break down complex, long-horizon tasks into manageable sub-tasks for robots, a process traditionally difficult for automated systems. The team developed a method where an LLM, such as GPT-3, generates an abstract policy outline for a task. A separate, smaller model then translates this outline into executable actions for a specific robot in its particular environment. This hierarchical approach allows robots to perform multi-step tasks, like sorting objects or preparing a meal, with greater reliability and adaptability than previous methods. The work highlights a promising direction for combining high-level reasoning from LLMs with low-level robotic control. For the full details, read the complete article at https://technologyreview.com/2023/07/10/1076091/llms-robot-planning-mit-csail/.