To think that mumsnet posters should know that anyone can use AI to analyse their posting history

[LostPollypocket]

I understand why a previous thread on this was deleted and suspect another one will be. I did think the thread was doing something very useful though in making people aware that even anonymous posts may not be so anonymous in a world where anyone (or business, crime syndicate, political party, or government) can use AI to pattern match and analyse everything you’ve ever written on the internet, including places like mumsnet where you might post about very personal topics. If you use a particular word frequently, or semicolons in a particular way, or even respond emotionally to certain types of posts, then even with a name change anyone might be able to build a profile on you and connect it to say, archived facebook data, and know who you are, how you think, and what you posted at 3am in the depths of postpartum depression.

😂
I don't have to go that far back.

I’d say it’s already a massive problem that personal profiles have been built to show highly targeted adverts that receive no public scrutiny, and a big part of what has led to the rise of extreme politicians and political fragmentation. There’s the potential for that on steroids - an extreme party targeting individuals with highly personalised misinformation that hits all of their buttons, whatever those are.

Then there’s the potential for blackmail, already we live in a world where spam emails lead to extortion and suicide because people think they’ve been ‘caught’ watching porn. Imagine the accurate personalised version of that spam email that threatens to send all of your contacts an anonymous post that you wrote years ago.

Identity theft. AI voice scams already exist, now imagine they clone your voice and ring your elderly mother with details of a family anecdote that you put on facebook once.

Employers, oppressive governments, health insurers…

Anyone gaining any kind of public prominence, even locally, could be a target for all sorts of things.

I’m sure there’s more!

I just AIed my self and I sound dull but helpful.
Some things I knew were public came back as expected, trended to more recent info, over emphasised some posts and made other things sound like facts that are outright wrong.

A typical AI summary really.

It is worth being aware how much posters share when it's very easy to get a summary and trend of their posts.

I might change username to give a gap on the trending. It's not like I'm building a brand on Mumsnet to monetise.

If you give chatGPT the question "are these 2 posters the same person" it may get the right answer yes/no simply by chance.
If you test it on yourself, you will know if it gave the right answer, but anyone else, how would you know if it's right?

I just tried it with one of my own posts and that of another person picked at complete random and it told me we are the same person 😆

Apparently we both have the exact same direct, assertive, conversational style so must be the same person 🤭

Yeah, so accurate isn't it?!

Different person actually. My agenda is concerned citizen!

How do you even do this????

AI currently works well if you know how to engineer a good prompt and less well otherwise, this isn’t how it will be done. It will be AI agents trawling huge amounts of data and using existing corpus linguistics techniques to find patterns

I asked chat gpt is this the same person then copied and pasted a response from me and from someone else (didnt include user names obv)

Pull the other one, it's got bells on.

Yep. I asked AI about my posting history and some of the stuff that it gave me was accurate but it also said that I had posted about my experience of being a single parent...I have never been a single parent and I've never posted about being one!!

So I would take it all with a rather large pinch of salt, personally.

If you are so worried about it- why are you posting on here then?

Its clearly dangerous

Good point op

I am going to put what I post into AI now and have the words changed to different styles.

I have always quite deliberately changed details when posting to reduce risk of someone trying to figure out my identity from posts.

Am I right in thinking that owners of sites like this will get paid by AI companies to allow access?

I just tried it with AI...I sound great.😂

But if you change username and then at some future point use, say, the distinctive turn of phrase “AIed myself” and additionally consider the very revealing data of when one account stops posting and the next starts, then you haven’t created a real gap in the data.

They don’t need to, it’s on the internet!

What ??

Some of us just aren't that bothered 🤷‍♀️

We don't take MN too seriously, don't post that often and don't post anything we wouldn't mind about other people (RL) knowing.

I’m not worried about it. My whole identity is a complete fabrication.

Too late, I previously used facebook, and have an extensive anonymous posting history across various anonymous forums, as I suspect do many of us! From this point I will certainly share far less information about myself and change usernames more frequently and carefully, and attempt to delete any info I am aware of that I would not want to be bought and sold.

But i’m more concerned about the wider social implications than personal privacy tbh. Think of the most credulous and reactive person you know. Do you want them to be seeing a feed full of AI generated videos of whatever is most triggering to them before the next election. Times several millions.

If I gave it 2 pieces of my writing I would expect it to give a fairly hedging response, that there's enough to suggest it's the same person but it can't be certain.

It's less about whether AI can currently compare 2 pieces of writing with accuracy, and more about whether in the future it could analyse a style and independently find matches elsewhere.

If that ever became a reality, people would probably just use AI to cover themselves. For example they'd use their own writing when something is publicly in their name, and use AI if they were posting elsewhere anonymously.

You cant be that worried about it if you are going to stay on MN then 🤷‍♀️

You're saying this like it isn't already happening?

The credulous and reactive person is already targeted with those videos by the algorithm that has established that they click on that sort of shit.

Why would they then need to connect their account to their MN data?

The credulous and reactive person is already targeted with those videos by the algorithm that has established that they click on that sort of shit.

Exactly! this is already happening 😆

I saw the last thread, so I asked AI to tell it everything it knew about my current username.

All it could tell me is I may or may not have children, and I might be a civil servant. Why on earth it would think that I've no idea! 😂