Post by @pronoiac@mefi.social

2023-04-07 04:33 UTC

@DrewKadel@social.coop alt text: Something that seems fundamental to me about ChatGPT, which gets lost over and over again: When you enter text into it, you're asking "What would a response to this sound like?" If you put in a scientific question, and it comes back with a response citing a non-existent paper with a plausible title, using a real journal name and an author name who's written things related to your question, it's not being tricky or telling lies or doing anything at all surprising! This is what a response to that question would sound like! It did the thing! But people keep wanting the "say something that sounds like an answer" machine to be doing something else, and believing it *is* doing something else. It's good at generating things that sound like responses to being told it was wrong, so people think that it's engaging in introspection or looking up more information or something, but it's not, it's only, ever, saying something that sounds like the next bit of the conversation.

Replies (4)

@pronoiac@mefi.social 2023-04-07 04:36

@DrewKadel@social.coop alt text in previous reply, @diazona@techhub.social @clacke

Open ##4802384
@diarra@framapiaf.org 2023-04-08 05:59

@pronoiac@mefi.social @DrewKadel@social.coop I would not call it as #ChatGPT but as BullshitGPT, because this is what it does.

Open ##4802385
@bgtlover@linuxrocks.online 2023-04-09 15:27

@pronoiac@mefi.social @DrewKadel@social.coop perfectly correct! after all, that's what gpt was doing all along, from gpt 1 all the way through 3 and maybe beyond. Expand the framework all you like, but the underlying principles you started with will be the same, unless you changed the principles entirely, but then, why would it be called a new version of an old framework, rather than a new framework altogether? it's true that now, if you ask it questions that require understanding of the world, you get answers that at least attempts more to demonstrate said understanding, but the fact remains though, the answers are indeed mostly emulated. With code it's a bit different, but I don't yet know why, more research is required.

Open ##4802386
@sheldonrampton@mas.to 2023-04-09 23:02

@pronoiac@mefi.social @DrewKadel@social.coop People should not think that ChatGPT is a reliable research tool, but it can actually be pretty helpful as a down-and-dirty _approximation_ tool. It gets some details wrong, but most of the facts is presents have been correct in my experience. If you make the effort to fact-check it a little, you can get good results from it with less effort than other methods.

Open ##4802387