A popular assertion among cognitive psychologists is that large language models (LLMs) rely on a “developmentally implausible” form of language learning. There are many points of discordance between AI and human development, but data inefficiency is one of the most commonly discussed.
