Where is the AI that looks out for women ?

[feministAI]

Ive been subject very misogynistic comments tonight by an acquaintance of a friend and asked ChatGPT how to manage it and got such a man centric response I’m really disappointed. Even fecking AI is on the man’s side. It essentially told me to pipe down or say a nice soft response to the secure comment I heard We really are on our own here to speak up for women aren’t we ? A 65 year old man I met today for the first time ever commented on something and his comment was such that I found it disturbing and distasteful but AI suggested I digest it and make a “gentle” comment if it comes up again. I despair.

Try telling it what approach you want it to take. See how that changes it. Tell it to respond as if it is a radical feminist!

Maybe this is it. But it clearly needs to be told and this is what I find sad. The whole “ah. Perhaps you could gently suggest not everyone feels that way”. It can bugger off .

Why use AI for issues like that?

What were the initial comments, and what did ChatGPT suggest?

It's not human. It's going to reply in a bland way unless you direct it in the direction you want.

I read something recently where a woman got AI to help with her CV. She ran it once with a female name and then changed it to a man's name. AI focused on her collaboratory skills on the female version but on her goal getting achievements on the male one.

It mimics and synthesises what it finds online, and regurgitates it in response to your prompt. It’s not surprising that the results are misogynistic or sexist as that’s what the source material is often like.

I'm not surprised this happens - a typical example of algorithmic bias that you will get from AI. It will have absorbed the patriarchal idea that women should be pleasant, and calm and calming at all times.

AI is only as good as its programmers which I bet are mainly men. If it’s taught about men saying offensive things by using a male dominated forum, then it’s going to say bullshit it die.

Are you quite gentle when you talk to AI normally? You might want to consider talking to it like a man and more firmly so it knows that’s the tone you want. It’s programmed to be gentle at first and needs prompting to be more angry.

The “programmers” for AI isn’t quite correct - it’s basically the training data. It learns from the data that is out there and so reflects current attitudes

because the world is misogynistic AI is. It amplifies the dominant message

a simple example. Ai processes job applications and tries to select applicants that are most similar to the successful people in an organisation. The people at the top are usually men. So it tries to pick men. The programmers then insist that sex identifying data is removed. But the ai still works out who is male and who is female - the language used, the hobbies listed. Very few of the top company bosses list ballet as a hobby so they get thrown out

ai reflects our world. Don’t blame the programmer / its all of us

AI isn't programmed in the way you mean. That's not how AI works.

Am genuinely curious at the notion that anything artificially programmed should "look out" for any particular group? Or why outsourcing human interaction to this novel and experimental technology can be in anyway considered helpful?

I'm really alarmed by this trend of wanting AI / ChatGPT or whatever it is to provide answers to things that we should be proficient at by the time we're adults.

What did you want as a response? What was the misogynistic interaction you encountered? I'm really interested to understand why you needed reassurance from AI when you are clear that what was said was misogynistic or unacceptable and why you felt you couldn't address it at the time?

I read an article about AI that said AI will give you the consensus answer, because that’s how it works, by parsing what already exists.

If you are looking for a different viewpoint or POV,
you need to direct it as such.

Yes, it's sycophantic. Also, none of the tech bros behind AI know anything about or give a shit about actual women.

It often gives the wrong answer because it picks up what people are saying online, and often they are misinformed.

When the Red Arrows flew over New York last week, lots of people were praising the US airforce for it. Some people claimed it was the French, others that itcwas the British. As a result Grok was confidently telling people who asked who it was that it was the US airforce.

Tell ai that you want a response that reflects a confidant woman who wants to make strong point that this is an acceptable comment or however you want to respond. See if you get something more aligned to what you want.

Unacceptable comment, of course

Laura Bates (who started Everyday Sexism) has just written a book about this -

www.simonandschuster.co.uk/books/The-New-Age-of-Sexism/Laura-Bates/9781471190513

I’ll be buying that book. Just thought it was interesting and wanted to see what others think. I wasn’t asking it for advice just “testing” what the bot would think an appropriate response would be for my own entertainment and interest

It’s interesting that AI gives preference to men because of the data it uses but when you see AI generated posts on Quora it frequently mixes up male and female during the corse of the post, so that for example, “his” become “her”.
How does whichever AI that is used fail to follow fairly simple rules of grammar, but still be nuanced enough to give preponderance to men. To be clear I’m not doubting it does that, I’m just confused about why this happens and would be interested if someone who knows AI better could explain.

Why do you need AI to tell you how to respond?

I haven’t played with the current AI (my ethics including climate ) but it sounds almost as if they have programmed in “be sex neutral “ as one of the rules on top of the model

Allegedly AI corporation test for bias, but AI admits that its historical training data is male-biased. Just ask it.

It basically looks around at what exists on the internet already and gives you a summary of it back.

It's impossible to say if it gave inappropriate advice without knowing the content/context.

I voted YABU though - it's not sentient, it's not capable of being sexist. It looks at source material though so the websites it's looking at could be sexist.

It's designed for its answers to be tweaked and questioned.