AI or human? Any smart ways to tell?

[Suzanne O]

Yes that is unnerving when people start taking on the mannerisms of AI speak. I keep telling people writing formal letters - don't say "Given that" or they'll think it was written by AI (maybe it was!). In fact it really annoys me when my server ticket is replied to by "person name" and it's obviously been written by AI.

Of course you could set a new forum rule "Only post in your ancestor's local dialect" 'appen tis easier to tell when blithering idiots are human or wastrels - daft 'apeths.

Or you could have a rule that simply says do not post AI slop in any thread that is not your own.

If they test you close their access off from the thread via a permanent threadban

 

[Wildcat Media]

As I understand it a lot of universities and school systems subscribe to software (AI driven itself, ironically) that they can run papers and such through. Whether that's cost effective for a forum site is probably questionable.

One of my online friends is an instructor at a university in his area, and he say that those system fail at detecting whether the writing is AI or not. The problem is....think about it. Where did these Great AI Slop Machines get their content? Yep, from scraping millions of pages of existing writing. So of course AI is going to say something is AI-produced, as it only refers back to the writings it has already plagiarized. So yeah, many false positives, and the less intelligent instructors will "Fail" the paper rather than consider that AI is flawed, prone to error, and that the sources for its determination are things it assimilated (like the Borg).

Back when I took courses over 15 years ago, it was easy to find essays posted online which were turned in for grades. Many were via paid sources, but plenty of others were free. Hell, our Econ instructor gave us take-home exams, and we could find those same questions online, word for word, with the correct answers. I can only imagine the number of exams, essays, etc. available online today as opposed to what I encountered back then.

The big AI companies are now at the point of buying up books, destroying the bindings, scanning them, and sending them to the landfill. So that would only make an AI detector even more susceptible to false positives.

And everyone asks why so many of us hate AI...

 

[Wildcat Media]

This just shows how easy it is to emulate the way a bot writes (or did write a year or two ago). When you see something like that it could easily be a human who has just got used to, or even trained to, write that way. Meanwhile bots are getting better at writing like a human so being able to tell will soon be, or already is, very difficult to impossible.

I've seen some truly atrocious AI-generated text in recent times.

I should send you a sample of one of them...

 

[Suzanne O]

I've seen some truly atrocious AI-generated text in recent times.

I should send you a sample of one of them...

Oh i believe you here. Some of the stuff on Social Media is AI Slop. The geniune stuff isn't taken more seriously than it should be. Instead you have bots repeating other people's geniune stories.
I see it on Threads every single day.
You see somebody on your feed that's popped up and their start date was yesterday for example and they already have 17k followers.
You read their profile and they're going on about some misinformation they're spreading.

 

[Anthony Parsons]

I watched something the other day... a teacher did something tricky with a test, where if a human answered it, the outcome is their response, but if an AI answered it, it had a hidden message within the test that only an AI would see and print, so everyone with the printed word failed. It was quite ingenious to say the least.

 

[ArtG]

Thanks all. Very interesting if not disturbing.

The question I haven't answered is why a bot would simply answer threads without posting any SPAM? What's the point?

 

[Alvin63]

Thanks all. Very interesting if not disturbing.

The question I haven't answered is why a bot would simply answer threads without posting any SPAM? What's the point?

Maybe that bit hasn't started yet! A test to see if they're accepted?