Researchers have found a surprising similarity between the way large language models (LLMs) like ChatGPT process information and how the brains of people with Wernicke’s aphasia function.