4

The Stack Exchange Network has been abuzz recently over the use of Artificial Intelligence (AI) chat bots such at ChatGPT to generate content suitable for posting as answers. Some sites, notably Stack Overflow, have issued blanket bans on their use to generate answers.

I had a completely different question. Since AI chatbots are trained by real humans using real human-generated corpora, the output of one arguably represents a mirror (if a dim one) of the beliefs of those who contributed training content. If the population of people that contributed the content is large enough to meet our general notability guidelines, the output could be considered notable as representing a community consensus.

To what extent can AI construct-generated content be considered notable and in-scope for debunking on Skeptics?

Taking the opposite perspective, in my experimentation with ChatGPT, I found several ways to manipulate it into making statements that I knew were false. Does that destroy any possibility of AI-generated content being inherently notable, or does it only impose some sort of good-faith rule, where only content generated through good-faith questioning of bots should qualify as notable? This could be enforced by requiring posters to post the exact text of the question they used to trigger the chatbot's claim. Triggers found by the community to be craftily written to exploit known bugs or weaknesses in an AI construct could be voted to close as asked in bad faith.

To be clear, I am not asking whether it is acceptable to copypaste content from ChatGPT and post it as my own answer to an existing Skeptics question. I am asking about scenarios where I am chatting with ChatGPT and it makes a factual assertion to me that I feel is dubious and suspect that the Skeptics community might be able to confirm or debunk it.

3 Answers 3

10

There are several reasons for asking for a claim to be notable.

One of them is to ensure that we aren't wasting our efforts debunking the unhinged theories brought up by a random drunk in a bar. We ask that the theories be widely believed to justify the hours of community effort that a good answer takes.

If you are using ChatGPT and it comes up with an answer you doubt, I think comparing it to a random drunk in a bar is a fair analogy. That isn't enough to establish notability.

However, if ChatGPT generates some text and it is somehow widely shared as true and widely believed, then it is notable - it is worth the effort to examine it and see if it is right.

(Another reason for establishing notability is to give context to the claim to help us resolve definitions and ambiguities. If the context is "ChatGPT said it", that will be hard to resolve.)

3

The fact that ChatGPT produced something as output does not demonstrate that the claim is notable. The claim might be notable, or it might not be. To demonstrate that a claim is notable, you must find evidence that it is notable, i.e., that a significant number of people believe it is true.

ChatGPT is not particularly relevant towards determinations of notability. ChatGPT isn't a very effective tool for determining what a significant number of people believe to be true. As such, the fact that ChatGPT outputted it is weak evidence, at best, and you would do better to look for some other evidence of notability. Just like "my friend said X" isn't sufficient to establish notability, so too "ChatGPT said X" isn't sufficient to establish notability.

To ask a question here, you must provide evidence that the claim is notable. ChatGPT doesn't really change anything: ChatGPT is not useful evidence of notability, so you'll need to look for traditional evidence of notability, in particular acceptable references to a real, verifiable source.

-6

ChatGTP is curated by interceders.

https://www.the-star.co.ke/news/africa/2023-01-19-kenyans-paid-2-per-hour-to-make-chatgpt-less-toxic-report/

Otherwise, ChatGTP in general is a fundamentally mysterious technology to the degree that its source remains closed and proprietary, and to the degree which we know its results are curated.

17
  • What does this mean for notability?
    – Joe W
    Jan 27, 2023 at 13:40
  • @JoeW That its notability at best is mysterious, and at worst, curated away from skepticism.
    – Anon
    Feb 5, 2023 at 8:50
  • That doesn't make any sense at all, it is either notable or it isn't.
    – Joe W
    Feb 5, 2023 at 14:21
  • @JoeW Ipse Dixit.
    – Anon
    Feb 5, 2023 at 18:15
  • What is that supposed to mean?
    – Joe W
    Feb 5, 2023 at 18:54
  • @JoeW en.wikipedia.org/wiki/Ipse_dixit
    – Anon
    Feb 5, 2023 at 19:04
  • And that doesn't explain what you are trying to say. That has nothing to do with my question about how your answer says anything for notability regarding ChatGPT.
    – Joe W
    Feb 5, 2023 at 19:50
  • @JoeW that doesn't explain what you are trying to say. Ipse dixit.
    – Anon
    Feb 5, 2023 at 19:57
  • I am simply asking you how does what you posted answer the question of ChatGPT being notable or not for the purpose of asking a question on Skeptics. And it seems that I am not the only one wondering this based on the votes this answer as received.
    – Joe W
    Feb 5, 2023 at 20:43
  • @JoeW You asked me a question, and I gave you an answer. That doesn't make any sense at all, it is either notable or it isn't. is your Argument from incredulity.
    – Anon
    Feb 6, 2023 at 1:04
  • And I clarified that your answer didn't make any sense to me or help me understand your answer.
    – Joe W
    Feb 6, 2023 at 1:06
  • @JoeW to me was not said. help me understand was also not said. If you are saying it now, I need you to be specific about what exactly does not make sense, and why. It could help if you detail the basic frame you operate under, which I gather is [ correct me if I am wrong ] Notability is binary. It is not a matter of degree.
    – Anon
    Feb 6, 2023 at 1:28
  • 1
    I don't understand how this answers the question about ChatGPT being notable as none of what you answered seems to address that. Is your answer claiming it is notable? Is it claiming that it isn't notable?
    – Joe W
    Feb 6, 2023 at 1:39
  • @JoeW Yes & No & Yes & No. Yes, in that a curated answer is notable for its curation. Apply this the proverb for example, To learn who rules over you, simply find out who you are not allowed to criticize. -- you can quickly infer interesting and notable things about ChatGTP. No, in that given the curation and proprietary nature, that even if it seems that no curation is present in an answer, you can never be even reasonably sure that no fingers are pushing the scale. And finally, Yes & No insofar as we defer notability to the most useful answer, even if the answer has issues in itself.
    – Anon
    Feb 6, 2023 at 10:16
  • This is sounding more and more like a rant as you seem to be complaining about various things. And yes and no isn't an answer that will help anyone as the question is looking for information if ChatGPT can be used in questions as a reference to a claim and this won't help anyone determine if that is the case.
    – Joe W
    Feb 6, 2023 at 13:29

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .