From LLMs to LTMs?
A Perspectives article from June 2024 is making the case, as its title indicates, for language as a tool for communication rather than thought. Regardless of where you fall on this longstanding philosophical debate, and irrespective of the conclusion (if there can really be any), this once esoteric question should be of enormous practical importance in the age of hashtag#LLMs: large language models are, after all, language models! If language is not primarily a tool for thought, or if it is just one of many, LLMs are not capturing or rendering human thinking, at least not its full manifestation. In that case, what would LTMs look like, "large thought models"? Not to be confused, of course, with LSTM 😎