Just a quick note on nomenclature: going forward, I will not be referring to large language models (LLMs) as “artificial intelligence” (or “AI”) going forward. They are indeed artificial, but I think the word “intelligence” implies things about LLMs that simply are not helpful by and large: it is a very weight-bearing metaphor. It comes with a few centuries of history, a history of both implicit and explicit claims about what it is to be human. Many of those claims themselves are rather suspect; when we transpose them without remainder to these fascinating machines we have built, we often only confuse ourselves.

Better, I think, to be explicit about the technology in question. “LLM” works just fine and does not carry those misconceptions.

I’m working on a project that among other things is interested in exploring other metaphors for these kinds of systems. If and as they suggest themselves to you, I would love to hear from you about them!