A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates that large language models (LLMs) can be used to generate detailed, step-by-step robot task plans directly from natural language instructions, bypassing the need for intermediate code. The ‘GenSim’ system translates high-level commands, like ‘tidy the house,’ into low-level actions a robot can execute by leveraging the world knowledge embedded in LLMs to infer necessary objects and logical sequences. This approach shows promise in making robot programming more accessible and efficient, though challenges remain in handling the physical uncertainties of the real world that differ from the LLM’s purely linguistic training environment. Read the full article at https://technologyreview.com/2024/05/15/robots-llm-natural-language-plans/