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 agree.

What with the plethora of lengthy ads, weird user names and strange posts I am beginning to wonder if this is some Great Experiment.

I mean it doesn’t seem beyond the realms of possibility but it would be good to get some confirmation / assurance.

I’ve even reported my OP to get it some attention.

Come on @MNHQ - seems like a straightforward question and important to know

I have suspected for a while that some opening posts are AI generated.

I hope that MNHQ would delete under the "not a genuine poster" criterion

Forgive my ignorance of such matters but how would MN know the post is AI generated please? All that is needed to post is a valid email address which must be validated and a username.

I am sure I am a dumb ass, but I've no idea!

Me too @AuntieStella, unless MNHQ has authorised or monetised them. I’ve reported under ‘not a genuine poster’ previously

If I were developing and training LLMs I’d want access to this site and pay ££££ for it.

Interesting question. Contrast that idea with the case against the UK University who it seems made a database of MN previous posts without MNHQ’s permission to do various projects of AI research with.

https://www.mumsnet.com/talk/womens_rights/5066453-thread-2-a-corpus-assisted-discourse-analysis-of-linguistic-transphobia-on-mumsnet?page=14

Is allowing AI posts for £££ the compromise position for MNHQ? Not good for HQ not to be honest with human posters if this is so, if our human responses are effectively being sold to benefit MN financially without our knowledge. Lots of people reply to posts here because they want to help the OP.

Haven't read enough of the thread to know the answer but, anything in here?

https://www.mumsnet.com/talk/site_stuff/5122770-why-were-taking-legal-action-against-open-ai-and-other-scrapers

Do you report them?
I've noticed a few OPs (often containing a link) and other posts within threads which seem to be the product of an AI (often curiously bland and banal) which I report and I think MNHQ have agreed and deleted them all.

I would aways report ones with links as that's just spam

Others (that are just off somehow) I might or might not report - depends on how "off" they are. If I think it's worth a look then I'd say what I think is wrong with it, then it's up to MNHQ to check behind the scenes and decide what (if anything) to do.

Aren't you a clever one ;-)

Sure people can use AIs as an alternative to a conventional search engine to find information. But it's still very unreliable, and it's a really bad idea for anyone to just C&p a chunk of text. Use the AI, read what it says critically and with cross checking, then use the info to write your own post would be fine.

I know Ai is not perfect and has errors, but alot of the time it gives better answers than other posters
But it lacks nuance, and in certain topics nuance is certainly needed as not everything is black or white, eg Relationships.

If I wanted an AI answer, then that's where I'd ask.

Similar for google.

When I ask on MN, it's because I want responses from actual people, who bring different insights, open up non-obvious angles, and spark discussion that could lead off into all sorts of interesting tangents. Because it's a chat site and I want to chat to humans, warts and all.

There is absolutely no point belonging to or posting on a message board if every answer is generated by AI.

I want to interact with real people who may have experience of my problem or issue. I want to ask people if they’ve tried a particular pair of shoes or dress on and how they found the fit. I want to try recipes from people who’ve actually made them. I want to read about holidays from people who’ve actually been to the resort I’m thinking about visiting.

What I don’t want is a generic robot reply that dances on the fence of “you may like this” or “other people think that”.

It’s also incredibly rude to come into a group and proclaim that a bloody computer gives better answers than the people in it.

Hi there,

We don't use AI to generate posts and were we to identify a poster as a bot (most likely because users would notice it), we'd remove those posts and block the poster.

We agree that Mumsnet's strength is that you get highly nuanced and varied responses based on lived experience.

Thanks @LilyMumsnet

It'd be interesting to know if MNHQ have tools that allow identification of bot posting and AI generated content (two different things though the former could use the latter) but I'm sure TSSDNCOP applies.

The problem is that while it's reasonably easy to tell a bot post (multiple accounts under same IP address), there aren't really any (good) tools to verify if something is AI posted. The ones that do exist have a high rate of false positives, so you run the risk of kicking genuine posters off the board.

Plus at what point does something become AI generated? I'll quite often use AI to reword an email more professionally, or give me alternative phrasings for a speech etc. It's my words, but filtered, just like if I had a (fairly crap) editor. I've not done this with a Mumsnet post yet, but I bet there are people who do.

Perhaps there needs to be a rule that every poster needs to be able to either correctly use an MN acronym, or reference a 'Classic'.

Yeah, it is tricky - using an AI as a tool (or to make amusing pictures) is one thing, it's using it uncritically for the content which is seriously problematic.

Thank you @LilyMumsnet

The much-deleted poster on this thread is a classic example of why AI-generated threads don't work. (They are a serial abuser of AI/ChatGPT and never really contribute genuinely to a thread, although they believe that they are offering 'superior' advice). As well as a PBP.