Every query typed into a large language model (LLM), such as ChatGPT, requires energy and produces CO₂ emissions. Emissions, however, depend on the model, the subject matter, and the user. Researchers have now compared 14 models and found that complex answers cause more emissions than simple answers, and that models that provide more accurate answers produce more emissions. Users can, however, to an extent, control the amount of CO₂ emissions caused by AI by adjusting their personal use of the technology, the researchers said. 

To find the best catalyst for green ammonia, researchers were staring down 8000 lab experiments. With AI, they only needed 28.

MIT researchers discovered the underlying cause of position bias, a phenomenon that causes large language models to overemphasize the beginning or end of a document or conversation, while neglecting the middle. They built a theoretical framework that can be used to diagnose and correct position bias in future model designs, leading to more accurate, reliable AI agents.