⁂ L. Rhodes on Nostr: It helps to understand that LLMs are basically machine learning programs for finding ...

It helps to understand that LLMs are basically machine learning programs for finding statistical relationships in language use. It's an advanced form of the sort of ML that digital humanities researchers have used to, say, chart the incidence of sexist language in English literature over time. So there are logical uses for LLMs, but a) they're mostly limited to the study of language, and b) they don't necessarily justify the costs piling up around commercial startups like OpenAI and Anthropic.