5 ms·
> the only bad thing is that it's hallucinating an answer instead of admitting it doesn't know. Isn't this a fundamental issue?
by bacon_waffle 3y ago
> the only bad thing is that it's hallucinating an answer instead of admitting it doesn't know.
Isn't this a fundamental issue?
- qup 3y agoOnly if it can't be corrected. How do you rate the likelihood of this problem being unsolvable?
- bacon_waffle 3y agoSorry, no idea.
- scrame 3y agoSince it's made by people who are convinced they're always right when explaining things? Fairly high.
- sieongioetnio 3y agoI don't think it's obviously solvable. All current approaches are plainly incapable of introspection. These GPTs don't understand their own "minds" half as well as we understand them, and we don't understand them very well.
- 11101010001100 3y agoAsk ChatGPT.
- GenerocUsername 3y agoWell it's a language model. Technically its just a really good auto complete, whose factual database is a side-effect of stringing together contextually correct tokens. It by itself is entirely incapable of knowing when it is wrong, despite possibly generating sentences apologizing for being wrong when told it was wrong
- williamstein 3y agoWhen I try this in GPT-4 I don't get a hallucination: "I'm sorry, but as an AI with a knowledge cut-off in September 2021, I can't provide specific information about the number of chapters in "The Haj" by Leon Uris. This book, like many novels, is not primarily structured by chapters and its sections may vary based on the edition of the book. You can easily find this information by checking the table of contents in your copy of the book." (I'm aware that every time you use it the answer is different.)