31–40 of 196 posts
I never liked that the term "hallucinating" ended up sticking. The AI doesn't know something so it just invents something. We usually call that "bullshitting", or in more polite crowds, "lying".
That's how lots of science, innovation, & learning work: generate many superficially-plausible candidates via a fast-and-loose process, then refine with a more rigorous evaluation.
That AIs, in the form of LLMs, are now doing this so well was unexpected, and progress in checking 'hallucinations' is proceeding very fast.
(Fortunately, the article is less dismissive than the headline, recognizing these model's potential & mainly urging an understanding of the limitations.)
I never liked that the term "hallucinating" ended up sticking. The AI doesn't know something so it just invents something. We usually call that "bullshitting", or in more polite crowds, "lying".
This is an odd article. To me, it seems like the "creative" arts are an ideal arena for AI. After all, there's no such thing as "wrong" art. The article says "well, sometimes what it makes is bad". Well big deal. A lot of human-created art is awful too.
Yeah but who wants to consume art purely generated by AI (that is, not human-created with AI support)? Most art sites have had blanket bans, or at least required tagging, on ai-generated art because people hate it so much.
Or to put it another way: why are you in the comment section of Hacker News, and not just asking ChatGPT to generate social media comments on the article?
Curiously enough, I don't get the same result as the author on plain ChatGPT-4. https://chat.openai.com/share/e213e0bd-2838-45e2-9942-e52954... https://chat.openai.com/share/64bc62d3-042c-40c4-8d50-8e28ce... Hilariously, plugging the example in the article into Bing enhanced ChatGPT-4, ChatGPT-4 w/ Bing hallucinates because of that very article! https://chat.openai.com/share/4fe50933-8436-44ad-a778-6297ca... If you t…
Bing does not use the actual GPT 4 model. It's almost certainly a lower parameter model (like 180 billion vs > 1 trillion), or at the very least heavily quantized. That's why it makes more mistakes in your tests.
Key sentence for me: > It might be better to say that everything GPT does is a hallucination , since a state of non-hallucination, of checking the validity of something against some external perception, is absent from these models. I try to explain this to people who are obsessed with using ChatGPT to tell them things. So far I've been telling them something like: "it does not attempt to provide you valid information…
No, but it's quite close - because "reasonable" is positively correlated with "valid, correct information".
Contra the headling, we should "celebrate being wrong" – when it's wrong in fast, interesting, & voluminous ways that can then be filtered or corrected. That's how lots of science, innovation, & learning work: generate many superficially-plausible candidates via a fast-and-loose process, then refine with a more rigorous evaluation. That AIs, in the form of LLMs, are now doing this so well was unexpected, and progress…
We don’t do human trials on random drugs, we do human trials on promising ones. Animal trials are more open but even then candidates are carefully considered as being viable. Things are even more open the earlier you are in drug discovery, but we aren’t testing molecules using Dubnium or most elements on the periodic table.
Similarly Ecology isn’t studying what happens when you introduce each type of salt water fish into lakes because the general assumption is they would just die thus saving you from preforming millions of experiments. That basic check for plausibility is extraordinarily valuable across all sciences.
There’s no sacred cows here. You do need to validate plausibility just like anything else, but you don’t need to do an exhaustive search across every possibility.