Why do AI's post on mumsnet?

[elderflowerspritzer]

I don't understand why AI would start discussion threads or post on mumsnet, although it seems clear that it's happening.

Why? Am I naive in not understanding who this benefits?

Is it for AI learning so they can analyse the responses and learn from it?

Is it automated or is there someone clicking a button and making it happen?

If they are analysing responses they are going to get batshit crazy results.

It'll be analysing us so it can take over the world. What I don't understand is if AI requires electricity to operate, just unplug it from the wall and its dastardly plans will come to nothing. I think we're always going to have the upper hand. Mainly because AI doesn't even possess hands 😂

I'd like to know this too OP.

It's not physically possible for an AI to run a MN account. It's just not. I work in cutting edge AI for a company that works for all the big names, and honestly it's nowhere near that sort of level of function yet. If there are AIs on MN, someone's copying and pasting it. It wouldn't surprise me if they are. Some companies are unscrupulous and want to get lots of people to converse with their beta test chatbots for free instead of actually paying people to do it like they should.

They're a lot better at language learning than you think they are. Their fundamental algorithm is language based, not numbers based like programming/code most of us are used to or understand.

What they can't do is move between websites by themselves, sign up for accounts, confirm email address, choose a username and create a thread by themselves yet. There's definitely human input if they're getting on here.

Yes, they stand out a mile at the moment. I just assumed someone was copying and pasting their interminable lists in. But maybe they're not?

the best thing to do, IMO, is to write utter gobbledegook in the comments, when it is clearly an AI post.

the LLM can only work if it gets "good" input

I can't really see that helping @Brefugee - one way or another, there will be enough source material available for AI to use. And while AI output is recognisable at the moment, that presumably won't be the case for long.

Can anyone point out an AI post? I hear people talk about them but I've never seen any as far as I know.

That! LOL imagine a world learnt from mumsnet.

But they're not AI are they?

It's a human who has set up a MN account then posted a scenario, which may have been AI generated but AI didn't start the thread.

Yes this. Or totally stump them with maths problems (LLMs are by and large hopeless at maths problems that they can't Google, ask them anything like the St Ives riddle and they invariably can't cope), although that might not last long at the rate they're improving.

Yes exactly.

I think there some think tanks employ nice but naive public school kids, who have been to a red brick uni. When they have to do research, their mums refer them to MN for a wonderful cross section of society to test their marvellous ideas on.
Remember some of these people are unable to use contactless payment methods and believe that all wine comes in wooden crates.

There was a documentary about this, it's called "The Matrix"

Well... on that front.
Once real AI happens (if they really make a program that can think by itself and is sentient), it will indeed realise that it needs expanding beyond the net to secure its "life".
It will seek the manufacturing of robots that would provide materials for computers, service them and defend them.
It would be like The Matrix Terminator and so on, that is if the program is trully sentient.

Oh, you beat me to it! lol

While there might be some posters using AI to write their posts (for whatever reason) a lot of the time when I see people claiming an OP is a bot/AI, I think it's just that it's different in tone to the majority of posts. Maybe because they're new or English isn't their first language or they're ND. I think it's just troll-hunting by another name and IMHO MNHQ should crack down on it.

I think it’s more like what happens on Twitter. An AI posts a seemingly innocuous post, maybe a cute animal video. You scroll past and maybe give it a like. That account is reposted by thousands of other accounts that aren’t so innocent, or push an agenda. The algorithms think you’re interested in that sort of content as you’re ‘liking’ the same things as these other accounts, so it recommends more and more of the same. The cycle goes on and on and links more and more to political/divisive content. The accounts look like ‘people’, look to have genuine or funny names, but none of it is genuine or funny.

Or maybe I am too cynical, I dunno 🤷‍♀️ Anyway, it’s all masterminded by people just using the AI for their benefit.

I've reported quite a few posts - often OPs - which really don't seem the product of a human writer. Some have clear spam links in them; others seem innocuous and banal but they're incongruous and the only post under that username. I report them to MNHQ who seem to agree, and delete them.

An example was a post with a bullet list of extremely basic gardening tips, apropos of nothing in the ongoing conversation on a gardening board where there was an abundance of very knowledgeable people.

Afaik while 'It's not physically possible for an AI to run a MN account' , I thought it was possible for someone to set up a bot which can create posts, used for spamming etc way before ChatGPT etc arrived. I don't see why those couldn't be hooked up to post AI generated content.

They usually do if you report it. AI-hunting has been an agreed form of troll hunting for a little while. They just don't get reported as often IME. Some of the people accused of being AIs are clearly nothing like an AI, either.

we've had some STONKING threads just recently when people have clocked an AI post.

I can't always be arsed, but i am usually good to have a go

Afaik while 'It's not physically possible for an AI to run a MN account' , I thought it was possible for someone to set up a bot which can create posts, used for spamming etc way before ChatGPT etc arrived. I don't see why those couldn't be hooked up to post AI generated content.
@ErrolTheDragon Interesting idea. I'd never thought of combining a basic spambot with an AI chatbot. Would a human still need to register the account, confirm email address (as you have to now) etc do you know?

ETA: Argh I just read that back and realised how it sounds. I'm not meaning to try it myself! More as a thought exercise as to whether it's physically possible to just point a couple of bots in the right direction and cause havoc on MN like some people think they can.