So your answer doesn’t contain any elements from LLM the other answer contains some elements but the language doesn’t match an AI. However, ChatGPT detection on ChatGPT4 is more difficult, most detection tools are based off the less capable ChatGPT 3.5 model. If enough users report a comment it’s automatically deleted

The 2nd one contains elements from AI? What's your exact proof for that?

Hundreds of hours playing around with ChatGPT. I said it contained elements not that it was written by ChatGPT. What I will say while it fits a specific pattern the actual language used doesn’t suggest ChatGPT. I won’t get into specifics.

Yes, I asked exactly about that. I don't think hours are any sort of good proof. I can say the same and claim I don't think it's ChatGPT. Although I've used mostly 3.5 and for fun. What parts do you think are related to AI and why? And do you think that answer should be deleted/publicly criticized?

It has 3 upvotes and zero downvotes it’s not in any danger of being deleted. It’s ineligible to be deleted by anyone except a moderator. I am not going to identify what parts of that answer, match a ChatGPT response, so in the event I am wrong and it actually was generated by ChatGPT then I give individuals who intentionally use ChatGPT to generate answers for knowledge to bypass detection.

@Ramhound: I would go further than simply saying that "most detection tools are based off [...] 3.5." The reality is, most detection tools are seriously unreliable for any modern generative AI (i.e. you can probably write a good detector for ELIZA, but not for a modern AI like ChatGPT).

@Kevin - ChatGPT 3.5 responses are actually really really easy to identify. ChatGPT 4 responses are more impressive but they in my experience have similar style. The difference between ChatGPT 3.5 and ChatGPT 4 is the quality of the response (ChatGPT 4 is more natural) but in either case in my experience they are very inorganic.

@Ramhound: Maybe a human can do better. But I was talking about detection tools, not things that a human can do.

@Kevin - I understand. I caught an answer that has several elements, and was clearly in my mind based on hundreds of responses from ChatGPT which detectors failed to identify, generated by a LLM. As my first response suggested those tools haven’t been really updated to detect ChatGPT 4. This reads like a LLM response. Honestly, in 2023 any answer that contains a list of information and no actual detailed explanation of the “how”, I almost always instantly suspect of being generated by ChatGPT.

None of their detection methods work for detecting ChatGPT from the hundreds of human hours to their other undisclosed secret algorithms. This is an example of another flawed system IMO since there are rules that something is not allowed, you have trigger happy sheriff's that shoot first and ask questions later. If people copy and paste right from ChatGPT, then that's probably proof since it'll put a reference link in it. I suspect whatever they think is working, is only working to see the 1% of what is really out there by ID10T humans. They think AI is going take their purpose and rep IMO.

If a human wrote this answer I would be shocked.

Its like they pick and choose what does and does not apply to rules. Here's a post when the answer posted was from the person that asked the question, waited 2 hours, then posted a self-answer referencing that they got the answer from ChatGPT: serverfault.com/questions/1150289/… This is on an SE community site too, it is explicitly stated that the answer is from ChatGPT, the post is still there. Uncertain example when the rules apply or not, what exceptions there may be, and just stating you got from AI is the workaround.

@VomitIT-ChunkyMessStyle - It’s not actually a workaround, everything I have seen from any moderator is that any use of ChatGPT isn’t allowed. Regardless, that obviously (self admitted answer) was flagged by this user.

Understanding the applicability of rules across various SE communities can be confusing. While I try to embrace emerging technologies, I also prioritize adherence to established rules. In the context of Q&A involving ChatGPT and similar AI, I recognize the potential perception of these advancements as a threat. IMO... Unsubstantiated accusations of ChatGPT usage can be seen as a threat, especially when wrongly targeted. Engaging in voting or flagging without concrete evidence is equivalent to making baseless accusations.

It takes as much effort to find these as it does to find profanity within posts. I've seen racist content generated only by the SU flawed audit review system with their algorithms, reported it, and have everything saved with proof from long ago even though I'm a YT. They so worried about AI stuff, they still got their own flaws promoting racism and things like that to fix. IMO.... They need to give the mods access to fix the real problems of the system, that are threating to the community and world as a whole for that matter.

“Engaging in voting or flagging without concrete evidence is equivalent to making baseless accusations.” - This is the reason I flag a moderator. If discussion about AI happen I’ll point specific examples

For years, people have been informing moderators about various issues. However, moderators find their hands tied a lot, and they too still face the same scrutiny regardless of their status within this level (e.g. concrete evidence). In my opinion, the impact of advocating for movement or change is greater when reaching a broader audience, rather than confining the message to the limited scope of Stack Exchange community moderators.

@VomitIT-ChunkyMessStyle yeah, I don't think any approach to AI they had so far is correct. AI can be helpful when answering some questions. IMO. Other - not at all. Personally I think AI answers should be allowed, as long as the rule of linking proper source (ChatGPT) is done. Of course the poor quality/wrong ones can always be judged and downvoted. Just like any answer.

But then you still have a problem of linking ChatGPT to being the source. Which is not too solvable at the moment, just like current AI content removal detection, which would be impossible to defend in e.g. court as there's simply no proof other than "common sense" and AI to... detect AI. None of this makes any sense to me and as can be seen by my post it leads to less knowledgable people spamming "Is it ChatGPT?" nonsense under many longer answers, which SE exactly wanted to avoid.

What is the purpose of this post past to whine about something asking. “Is this ChatGPT?”