I found the Borges and AI article to be helpful for thinking about what it is that these LLMs are. Or at least one way of thinking about them, by analogy. I appreciate that the article is non-technical and short. I suppose an analogy of the analogy can be also applied to the other ML models too. It makes thoughts more specific and concrete than does just daydreaming about the space of possibilities. The stories give a literary model to push and pull on, something to conrast our reality against, something to compare the near possibilities too.
My interpretation is that, like the situations in Borges’ stories, these computer programs grant us access to arrangements of letters: texts. These texts might be fantastically useful, deviously misleading, or just gibberish in which we see what we want to see. We don’t know which of these the texts are, and can’t trust the texts to tell us honestly. We have to decide that for ouselves.
Always has been / Same as it ever was.
Via MetaFilter.
The Link: https://ar5iv.labs.arxiv.org/html/2310.01425a
See more links on the Interesting Links page.
Comments by email are welcome.
Posted: in Links.
Other categories: link-roll; link article.
Back references: The Hofmann Wobble.
Tags that connect: [[ai]] New to me Facts and Ideas in June 2024; [[internet]] Trust and Internet and Ai, The Hofmann Wobble, Relaxing with YouTube, Add Webcams Link.
Tags only on this post: none.