MNHQ do you allow AI generated threads here?

[AccountCreateUsername]

Not a TAAT(s) but there have been some quite unusual threads on here lately. Would MNHQ ever authorise AI generated content for training LLMs or authorise academic or other organisations from using MN content in research or training of LLMs?

Can you assure us that we’re at least responding to a human when we post on threads please :)

Thank you

I’ve noticed another poster’s replies and (it seems) profile have been quickly deleted today. All this speculation about what’s real and what isn’t spoils the whole experience of using this site.

Please help us post in confidence again @MNHQ

I don’t see what they can do beyond delete obvious and reported AI posts. It’s a problem for the whole Internet, and there isn’t an automated solution.

Indeed but some transparency would be nice. I don’t know anything about IT but if schools and other organisations can check for AI generated work, maybe there is or will be an automated solution.

But I’ll accept what MNHQ say - no AI generated posts areallowed, and carry on reporting anyone who comes across as a bit robotic!

The talk guidelines already make it clear that spamming isn't allowed, but maybe some explicit guidance on acceptable and unacceptable use of AI tools would be appropriate?

Agreed @ErrolTheDragon

I think that would be a good idea - as these posts become more common and harder to spot, the talk guidelines should reflect the times we live in.

But how are they detected? Genuine question. I’m not sure what is meant by an AI post. Is it someone posting content that they’ve put through AI? Or is it an AI bot posting with no human action?

If you ask a poster if they used AI to generate a post, I’d expect them to answer. If they don’t, I think they’re more likely to be a bot.

I reported 2 weird posts yesterday. No links but first just saying what they thought/were pushing; no questions, no asks for help, no hope of interaction. The 2nd seemed to be an advert but no links again, 'flat', copied, no life. Hope this helps others distinguish some false posts.

At the moment it's still quite easy to spot them. I imagine that'll change soon enough as they keep using sites like this to hone their skills.

I think there are a lot of bots doing the rounds, it was obvious last year when there was one posting several threads at the exact same time (with the same user name!), last year it was just inane stuff like "what's your favourite time of day?", recently there's been a large uptick in poorly worded threads goading controversial topics I've noticed, it's the political ones that worry me, deliberately stirring political unrest.

This is a really clear example of an AI-generated spam post. https://www.mumsnet.com/talk/shopping/5199375-unlocking-guest-post-opportunities-a-comprehensive-guide-for-usa-bloggers

@EmpressaurusDelleGatte well that's easy to spot as it's spam too so few if any would bother to respond, but the ones feigning a problem, no matter how obvious to some, always get some people replying.

But how are they detected? Genuine question. I’m not sure what is meant by an AI post. Is it someone posting content that they’ve put through AI? Or is it an AI bot posting with no human action?

There are a few different things.
'Bot posting' is used for spamming, DOS attacks and the like - this preexisted the LLM AIs. The content of these may or may not be AI generated now. We've been spotting and reporting spam for years.

There are posts which look to me to be entirely AI generated. Sometimes there are links to products or services, not always. The content is often quite bland and formulaic, often purporting to give advice on a subject but it tends to be basic or just not the sort of thing any in needs (eg a recent example I reported was purportedly tips on making a card for 'Teacher Appreciation Week' ... actually maybe MNHQ could get one or two like that out of the bin, remove any links and post them as examples? I just can't imagine any sentient conscious being spending the time creating this sort of drivel.

' posting content that they’ve put through AI' is different- if you've written something and then just ask an AI to clean up its SPAG (and checked it hasn't changed the meaning of it) that should be ok .

And then there are AI generated images, poems etc which can be very amusing, they depend a lot on the human prompts and usually appear in context with a thread by a participating poster - if it's not obvious they're from AIs then people generally say.

There's no such user as "MNHQ" so you can stop tagging that name

You can use chatgpt to generate a draft then change it to your own wording. It's obvious when someone has lazily copied and pasted an entire Chat gpt response - I think you're entitled to report those.

However it's not really any different from copying any other chunk of text and pasting it as your own.

I agree that entire posts that are purely Chat GPT - generated are bland and don't really add anything but there is nothing wrong with it per se. I think it should be discouraged but it'd take a lot of time and energy trying to "prove" a chunk of text was likely generated by ai. Just report and ignore it.

That's different from "bot behaviour" posting - as in, posting repeatedly and blindly spamming boards. That is not allowed.

However it's not really any different from copying any other chunk of text and pasting it as your own.

You shouldn't be doing that anyway, obviously. If you c&p content you should reference it.

Yes I agree, it's pointless really. Just trying to get away from the idea that because someone got it off Chat GPT rather than Wikipedia it's automatically symptomatic of bot posting!

what is unusual is some try to link using Ai posts as spamming but the dictorany gives different meanings of spamming so to echo previous users, the talk guidlines could use new sections because as it stands based on a lawer type literal reading their are loopholes to use Ai posts, or if people use the Ai to imporve their posts etc

Sorry @DancingTurtle ive just got home but couldn’t answer your question as well as Errol has. Thanks Errol!

@MilesOfCarpetTiles
There should be! Unless I don’t need to tag anyone in ‘site stuff’. Luddite here.

The posts I meant just felt like they read as a bad chatGTP answer. I just wanted some assurance (from MNHQ 😉) that we’re chatting to another human and not some chatbot in training.

How are posters so sure there are AI generated posts? I wouldn't have a clue!

Yes to this, and same goes for responses to an OP. I've seen several that were clearly AI.

What about people who generate AI posts and are not bots?

The AI content is just as lacking in human input as the work of a bot.

usually its human when they are bun fighting, name calling, arguing with slander, toll hunting, or trolling other posters with random or unhelpful comments.

then you know its human,

No I think on 'Site Stuff' it's expected that MN staff will read the threads! You can tag the individual who has replied though with their username.

AI tends to fall back on stock phrases and linguistic patterns.

"In today's fast paced parenting world..."

They're easier to spot than to explain, once you know what to look for. If you want more examples, have a look at LinkedIn Answers - there's a project completely ruined by AI questions getting mostly AI answers, it's probably poisoned whatever LLM they were hoping to train.