New Members with AI-Generated Comments?

[cambookpro]

I've noticed recently an uptick in brand new members who seem to be posting messages in a style very reminiscent of text generators such as ChatGPT. However, the comments do seem to be on-topic, e.g.:

(From: https://forums.macrumors.com/thread...-the-sun.2419037/?post=33301851#post-33301851 )

It made me wonder a couple of things:

1. Firstly, what is MR's policy on AI-generated content? I can see a world where it helps people who aren't fluent in English communicate better, but also one where the forums are overrun with AI bots replying with inane messages to inane AI-generated posts.
2. What is the play here for the person behind this account? Is it to generate credibility before releasing some kind of spam/scam?

Thoughts welcome!

 

[I7guy]

Obviously can’t guess the motive. But ai generated content is going to become more prevalent. The flip side is one may be interested in knowing the answer to some question posed in this forum, so let chatgpt find the answer.

 

[russell_314]

Some people might not have English as their first language so they use AI to correctly word their reply. Some people might just want to sound better and they feel the way they type sounds bad. if it’s on topic and the information is correct does it really matter if they typed every single word? They’re not writing an essay for college credits

 

[Boyd01]

what is MR's policy on AI-generated content?

Please see number 13 under "Things not to do" in the forum rules. If you see violations, just use the Report button under the post.
____________________________________

"Don't post content produced by artificial intelligence ("generative AI"). Even if you use generative AI to research an issue, posts must be in your own words and represent your comments or opinions. Exception: You can post AI-generated content if you are giving examples in a discussion thread about the topic of generative AI."

 

[applepotato666]

What is the play here for the person behind this account? Is it to generate credibility before releasing some kind of spam/scam?

It's possible. It could be a script making an LLM generate an answer completely automated, it could be a person using LLMs to write "better". It's not hard to be on-topic, even if it's automated since you can make ChatGPT read the thread. In any case bot activity is skyrocketing, and on forums it's way harder to tell since the majority of us keep ourselves anonymous.

 

[cambookpro]

Some people might not have English as their first language so they use AI to correctly word their reply. Some people might just want to sound better and they feel the way they type sounds bad. if it’s on topic and the information is correct does it really matter if they typed every single word? They’re not writing an essay for college credits

It's not using AI per se that is bad - more the concern that the motives may not be completely innocent (and that's without considering that the example in the original post wasn't really on topic - the thread was about the potential of damaging the Vision Pro cameras, so the suggestion to use a pinhole camera is completely irrelevant and intimates that they lack true comprehension and understanding of what is being discussed).

In any case, @Boyd01 has at least cleared up MR's policy on this (although I would assume that accurately moderating this specific rule is going to become increasingly difficult).

 

[russell_314]

It's not using AI per se that is bad - more the concern that the motives may not be completely innocent (and that's without considering that the example in the original post wasn't really on topic - the thread was about the potential of damaging the Vision Pro cameras, so the suggestion to use a pinhole camera is completely irrelevant and intimates that they lack true comprehension and understanding of what is being discussed).

I’m not sure how AI could be used for evil in the forums but it’s possible.

In any case, @Boyd01 has at least cleared up MR's policy on this

Yup when it comes down to it the website is privately owned so it’s up to them to figure out what is best. As users we just have to do our best to follow the rules.

(although I would assume that accurately moderating this specific rule is going to become increasingly difficult).

Not at all. I’m sure someone will report the posts and keep the mods busy 😂

 

[adrianlondon]

Please see number 13 under "Things not to do" in the forum rules. If you see violations, just use the Report button under the post.
____________________________________

"Don't post content produced by artificial intelligence ("generative AI"). Even if you use generative AI to research an issue, posts must be in your own words and represent your comments or opinions. Exception: You can post AI-generated content if you are giving examples in a discussion thread about the topic of generative AI."

Hah. This'll get rid of all those posh people with their iPhone 15 pros and iOS 18.1 betas.

 

[zorinlynx]

I would vastly prefer the rule about not posting AI-generated text to remain and be enforced. I like interacting with other people online, not computers pretending to be people.

 

[Nermal]

Indeed. There are basically two "classes" of AI-generated text, and I'm personally not a fan of either of them (and therefore I tend to moderate AI posts rather aggressively).

Sometimes we've seen "member A" ask a question, and then "member B" comes along with "ChatGPT had this to say". I consider this no different from the "let me search for that for you"-type posts that we've disallowed for a long time. If someone wanted a ChatGPT answer then they'd ask ChatGPT.

The other kind is the sort you allude to above, where posts are clearly AI-generated but attempt to fly under the radar. This forum was intended for humans to talk to other humans, not for computers to talk to each other, so we also delete these when we see them (keep those reports coming!).

Backing up to question 2 in the first post, often it is indeed an attempt to make an account "clean" before spamming. We don't document the specifics, but new accounts are subject to more checks and filters than established accounts. It's surprisingly common for a spammer to create a new account and then do nothing with it for months, only to then turn around and post spam. We've twigged to this sort of tactic and it typically doesn't make it through the filter, but I suspect the same technique works effectively on other forums.

 

[I7guy]

It's not using AI per se that is bad - more the concern that the motives may not be completely innocent (and that's without considering that the example in the original post wasn't really on topic - the thread was about the potential of damaging the Vision Pro cameras, so the suggestion to use a pinhole camera is completely irrelevant and intimates that they lack true comprehension and understanding of what is being discussed).

In any case, @Boyd01 has at least cleared up MR's policy on this (although I would assume that accurately moderating this specific rule is going to become increasingly difficult).

So you are worried about disinformation? If that is the case it shouldn’t make a difference AI generated or not. Someone posts something that’s a red flag, be mindful of it.

 

[cambookpro]

So you are worried about disinformation? If that is the case it shouldn’t make a difference AI generated or not. Someone posts something that’s a red flag, be mindful of it.

Not really disinformation - more that, in the long term, the forums become clogged up with less valuable posts and it reduces the overall enjoyment of being here.

 

[BugeyeSTI]

Not really disinformation - more that, in the long term, the forums become clogged up with less valuable posts and it reduces the overall enjoyment of being here.

I'm all in for not allowing AI generated posts on this forum. After realizing the the administrators have rules to address the issue, I'm sure posts like the one you've used as an example will be reported and removed..

 

[alfonsog]

I see that here and elsewhere, perhaps they are bots? But I even see full articles written by journalists now that I can tell are all AI or AI enhanced. If I were to publish something using AI I would 1. check on all the facts and 2. rewrite it in my personal style. I think people are just getting lazier with AI. I’ve asked AI to do things which are specific to my field (classical music) and it totally botches the facts.

 

[theluggage]

If I were to publish something using AI I would 1. check on all the facts and 2. rewrite it in my personal style.

...which would take just as long as doing your own research to start with and writing it from scratch. Heck, I've edited stuff written by humans where it would have been less work to just chuck it and re-write it myself.

The way I feel about current so-called "AI" is this: way back in the 80s, someone handed me a floppy disc for my 8-bit micro. It clunked away for 2 minutes and then proceeded to display a blocky, grainy, ~5-second snippet just about recognisable as being from Star Wars. It was amazing at the time, and - to anybody with imagination - suggested that "one day, we'll be able to watch movies on our home computers - don't buy shares in Kodak!". By the late 90s/early 2000s, that became a reality (earlier if you count DVD) - but it was also, in 1982, about as much practical use as a chocolate teapot c.f. just watching a VHS tape.

That's what "AI" (or, more accurately, LLMs, Stable Diffiusion) feels like at the moment - its a fascinating tech, and 10 or 20 years down the line it's distant descendants will probably be invaluable, but right now it's just an impressive demo... except, somehow, everybody from Apple down are trying to persuade me to, figuratively, buy Star Wars on 100 floppy disks... Unlike the metaphor, though, it does have a current "practical" use of swamping the internet with plausible but nonsensical click bait.

 

[mw360]

...which would take just as long as doing your own research to start with and writing it from scratch. Heck, I've edited stuff written by humans where it would have been less work to just chuck it and re-write it myself.

The way I feel about current so-called "AI" is this: way back in the 80s, someone handed me a floppy disc for my 8-bit micro. It clunked away for 2 minutes and then proceeded to display a blocky, grainy, ~5-second snippet just about recognisable as being from Star Wars. It was amazing at the time, and - to anybody with imagination - suggested that "one day, we'll be able to watch movies on our home computers - don't buy shares in Kodak!". By the late 90s/early 2000s, that became a reality (earlier if you count DVD) - but it was also, in 1982, about as much practical use as a chocolate teapot c.f. just watching a VHS tape.

That's what "AI" (or, more accurately, LLMs, Stable Diffiusion) feels like at the moment - its a fascinating tech, and 10 or 20 years down the line it's distant descendants will probably be invaluable, but right now it's just an impressive demo... except, somehow, everybody from Apple down are trying to persuade me to, figuratively, buy Star Wars on 100 floppy disks... Unlike the metaphor, though, it does have a current "practical" use of swamping the internet with plausible but nonsensical click bait.

By any chance was this a little bit of the Death Star dogfights, on a BBC micro?

 

[CharlesShaw]

I’m sure someone will report the posts and keep the mods busy 😂

And if the task becomes overwhelming they can enlist the help of AI.

 

[russell_314]

And if the task becomes overwhelming they can enlist the help of AI.

True 😂