@Glorfindel Sorry for abusing superping, but you seem to be online and I wanted to briefly discuss your flags, if you're willing and available.

@0xC0000022L no problem! I can't prove they're generated by ChatGPT, of course, and even if it was, and the user has the knowledge and took the time to validate the answer and provided attribution, then it's a fine answer in my book.

I don't know your site's policy on ChatGPT, and it varies between SE sites. If you think you can contact the user, sort things out, make sure the posts get attributed and they're not overly promoting their own product, fine with me.

@Glorfindel I think we have no stated policy as of now, frankly I thought the whole "ChatGPT answers are banned for now" applied across SE 😁

I commented on the two flagged answers for now and in one case also downvoted.

@0xC0000022L we had some discussion in the TL and AFAIK only attribution is required (both by SE policy and the terms of usage of ChatGPT). The ban was necessary on Stack Overflow because of the volume vs. the number of caretakers (moderators and regular users) - I guess that's less of a problem on Reverse Engineering.

In case of the answer linked above at least the single spelling issue that stood out ("witch") speaks against "AI-generated". Could of course be an obfuscation technique of its own to disguise AI-generated content. So far all AI-generated stuff I got to see from ChatGPT was at least using correct spelling, albeit not always correct grammar.

@0xC0000022L witch one? Not the one you linked to, it seems ...

(sorry for the pun)

@Glorfindel Got it. It would probably be less of an issue, yes. Alas, I had some colleagues at work "ask questions" to ChatGPT and all of the stuff that was related to reverse engineering the answers sounded legit, but if you did more than just read them (e.g. trying the "advice") it became clear very quickly that this AI was bullshitting very eloquently 😐

oh, this one reverseengineering.stackexchange.com/a/31373/13506

@Glorfindel yep

@0xC0000022L ... meaning that while the answers were quick and sounded legit at face value, upon closer inspection always turned out to be made up -- but very eloquently so.

The detector says 0% chance of being auto-generated, true. So maybe they do have some knowledge (more than I at least :P) and could be a valuable contributor to the site.

@Glorfindel That was my thought, when you pointed out the affiliation (which also isn't bad per-se) which the person failed to mention.

@0xC0000022L yeah, that's the main problem. Especially on SO where many correct answers suffer from poor grammar.

Very true, but also understandable given the majority of users probably aren't native speakers of English (nor am I). And even if they were, I know plenty of native speakers with worse grammar and spelling than I usually manage for English ;)

Side-note: here the affiliation has been disclosed.

Yeah, noticed that, hence I knew they were affiliated. I am not sure if their product is directly relevant to the question at hand, but if it is, then it's definitely in line with the rules for self-promotion.

@Glorfindel I guess at RE.SE it's more about deobfuscation than obfuscation -- which their product claims to do -- but it's still good to have an overview of the techniques and certainly the user seems to have some expertise in the subject matter.

Not really surprising, since they work for a company in that area :)

@Glorfindel Indeed. Anyway, thanks for flagging and also for being available for a brief discussion regarding the flags. Appreciate it. Oh and I wish you a happy new year. I guess with less than 10 days into the new year it's still alright to say that ;)

Thanks, Happy New Year for you too!

Thanks

@Glorfindel one more question, what tool did you use to assess whether it was AI-generated?

@0xC0000022L it's a user script I wrote, using an API published by Hugging Face:

@Glorfindel Thanks so much!