Recent searches
Search options
I've said this before, but the bias in English towards agency makes it really hard to talk about LLMs in a reasonable and neutral manner. If an LLM-based chatbot outputs a question as its response, is it fair to say that the chatbot asked a question? There's a critical difference between returning text that contains a question and asking a question, yet it's very difficult to be consistently clear about that distinction.
@xgranade@wandering.shopAnyway, it's something I'm trying to watch in my own language when I talk about Facebook's new chatbots today. They aren't "trying" to do anything, they aren't "claiming" things, or anything like that. It might seem like hairsplitting, but it does matter — so much of the marketing behind LLMs is geared towards getting us to believe that they are capable of acting with intent and purpose, but that's just not true.
(Never mind that if it *was* true, it'd open a gigantic ethical can of worms to effectively compel and imprison a system capable of agency behind a corporate chatbot. So I can only hope that said marketing teams don't believe their own claims.)
@xgranade
If you draw a Venn diagram with a circle for marketing and a circle for ethics the 2 circles don’t intersect or even touch at all, so the issue never comes up.