LLMs remain a divisive topic these days. Maybe we all know someone who has fallen too much in love with their new robotic friend, or who thinks it has turned them into a genius. [Emily M. Bender] and [Nanna Inie] I’ve written about how people anthropomorphize the LLMs they interact with and have suggested some linguistic tips to avoid it. It’s a couple of months old, but we think Hackaday readers will find it interesting. His analysis is interesting because it looks at the way people talk about LLMs and highlights unconscious anthropomorphism. The LLM is a piece of software, not a person, so why does it “recognize” when it does “speech recognition”, for example? Instead, they suggest “automated transcription.” Even “hallucination” implies knowledge that clearly does not exist. They admit that their suggestion of “undesirable results” is not entirely appropriate. They are on safer ground with “entry” and “exit” instead of “notice” and “response.” Whatever your opinion of them, it is clear that the use of LLM will be a feature of the world for the foreseeable future. However, the language around them is capable of evolving, and it may be worth taking note of some of the suggestions presented here. Dealing with our new electronic overlords? Take a look at our AI for Skeptics series.
