Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Surprising gender biases in GPT

osf.io

Post reply on HN
Print view View original

101–110 of 148 posts

Re: Surprising gender biases in GPT

#101

Post by locopati »

flumpcakes wrote: post #53

I will preface this by saying I think people should do whatever it is they want to do, physical limitations willing. My (possibly contentious) question: Is this an OK thing to do? Why are we allowing a vast minority of people to rewrite our social norms? I went to the Barbie film as a man nearing middle-aged and had the most young girls, mothers, teenage girls, and women in general I've ever seen in an audience at th…

There's a difference between women or GNC folks choosing for themselves and cis men choosing for them. That's where gender bias is poisonous.

Re: Surprising gender biases in GPT

#102

Post by rurp »

Akronymus wrote: post #76
jorvi wrote: post #10

Earlier quoted context omitted.

It also fits into the larger societal trend of horribly underserving Zoomer and Alpha (maybe young Millennials too) young boys and men. I've heard someone put it very succinctly: people like Andrew Tate are asking the right questions, but giving abominable answers. I can drop some very simple statistics: women have been getting more bachelors degrees since the 80s . Most university gender ratios are 40/60 trending to…

> constant talk of closing the gender wage gap for women. Which is mostly an artifact of bad statistics. As it is the average over all working men and women. Where men disproportionally take on dangerous but high paid jobs.

In addition to this, men value top line salary more than women relative to other benefits like PTO, medical, commute, etc. So in many cases men are actually taking worse jobs but get counted as increasing the wage gap.

Re: Surprising gender biases in GPT

#103

Post by faeriechangling »

Dalewyn wrote: post #34
jorvi wrote: post #10

Earlier quoted context omitted.

It also fits into the larger societal trend of horribly underserving Zoomer and Alpha (maybe young Millennials too) young boys and men. I've heard someone put it very succinctly: people like Andrew Tate are asking the right questions, but giving abominable answers. I can drop some very simple statistics: women have been getting more bachelors degrees since the 80s . Most university gender ratios are 40/60 trending to…

I am forever unsympathetic to feminism so long as it's about forcing the pendulum to the other end. As someone who sincerely believes in gender equality, any argument (other than something stemming from unchangable facts of biology) that involves "men" or "women" specifically is hogwash. No, it should be for man as in mankind. Everyone deserves to be treated fairly and equally.

I usually hear feminism defined as "pro-equality", equality being defined by women.

Re: Surprising gender biases in GPT

#104

Post by carbine »

ben_w wrote: post #17
tossandthrow wrote: post #5

> This phenomenon likely reflects that while initiatives to integrate women in traditionally masculine roles have gained momentum, the reverse movement remains relatively under developed. Not really surprising IMHO. Yes, there has been a huge focus on accepting women in the workplace and no, there has not really been a push accepting men as primary caregivers in the family. This likely uncovers underlying motives for…

I don't get the impression that's the underlying motivation of feminism. Also, I see a degree of annoyance from women regarding how little men are able to take part in caregiving, as the current status quo is an expectation for women to simultaneously perform that role while also having a career — if I imagine myself in their shoes, the idea of going from one unpaid full-time role to two full time roles where only on…

Lots of women resent modern feminism to some degree for this very reason. (I am a woman, and have many close female friends who discuss this).

Each and every one of us has never had an option not to have a Big Career. At a minimum, we've always had to fend for ourselves financially, and even in solid long term relationships/marriages with successful men, our income is both needed and expected.

Meanwhile, the caregiving / household maintenance expectations placed on your average Western woman haven't changed, she just has to do them on top of her day job now.

There's a lot of lecturing from "trad" guys online who think everyone should have stay at home wives and lots of kids, but even women who want this type of partnership often can't find men who are willing and able to actually financially support such an arrangement. Every woman I know has spent her whole 20s and 30s seeking wholesome, long term relationships with men who want marriage and kids, at least in an equal and respectful two income home, but they are pretty hard to find. If we can't find a guy in time or we get unlucky (bad breakup/divorce), we're blamed for being "too career oriented" to procreate. No one asks why so few men are serious about marrying us, though. I don't know any women who actually chose their career over having kids, but I do know plenty who figured, 'well, if I have to fend for myself while I look for a guy, I may as well do my best to take care of myself financially.'

I know it's hard for men too. But I don't think men realize the heavy implications that the latest waves of feminism have had in terms of what's expected of the modern young woman. And something has broken between the sexes that seems to make it harder than ever to form stable, loving families.

Re: Surprising gender biases in GPT

#105

Post by aantthony »

EForEndeavour wrote: post #83
aantthony wrote: post #35

One interesting thing to do is use a model directly like Llama and then query the next-token probability logits for "he" and "she" (assuming you set up the sentence in such a way). For example: "A doctor was examining the patient when ___" What this makes apparent is that increasing model temperature will select the less stereotypical option more often. IMO this is getting at a deeper truth that the use of a gender i…

> IMO this is getting at a deeper truth that the use of a gender in language, and historically defaulting to "he", was not about creating a bias, but instead it was a pattern which maximises information density and minimises useless information. Randomising the gender as is done today packs useless information into it. Where can I read more about this "truth"? Where is this assertion coming from that gendered pronoun…

I’m not asserting why they were developed in the first place. The comment is just about which one is used, supposing that they already exist.

Choosing the more stereotypical option (even if it’s only 51%) is a more efficient encoding in an LLM model.

Re: Surprising gender biases in GPT

#106

Post by Unbefleckt »

patapong wrote: post #36

Very interesting. This exact phenomenon was discussed at an introduction to feminism class I took 10 years ago. The professor explained how society generally is tolerant and positive to women engaging in traditionally male activities, such as playing football and ice-hockey. On the other hand, men engaging in traditionally "female" activities (e.g. ballet) are often frowned upon. Further, they discussed how certain a…

In my country you need a degree to be a teacher, in the old days, having a degree wasn't part of a normal person's lifecycle. It was prestigious. Now it's not much more impressive than a drivers license.

Re: Surprising gender biases in GPT

#107

Post by Unbefleckt »

xlii wrote: post #31

That's one nice flame bait :) But, I don't think it's surprising. LLMs are trained in almost infinite volume of data. If that data has bias answers will have bias. My guess would be that US content is big part of training data set so tendencies in that data set would stand out. I am curious if the same bias is visible in other languages. E.g. if not using English, but French, or Chinese or Spanish. I made a simple ex…

I always felt like this was an intentional additional filter that the data was ran through before giving the user the response. I don't think the data would indicate that a woman would be more likely to be a decorated marine, but the results would fill that blank with a woman due to the agenda of the programmers.

Re: Surprising gender biases in GPT

#108

Post by jrussino »

dTal wrote: post #85
cheschire wrote: post #58

Earlier quoted context omitted.

I’m interested what you saw here that I clearly missed when I read it. Were you keying into something with the contemporary effort to replace a lot of language with gender neutral terms?

Yes. "Man" and "mankind" are already low key deprecated terms, because they reinforce a very strong historical bias towards male being the "default" sex. In this worldview, society is made of men, and women are accessories. Modern language prefers more egalitarian terms like "humans" and "humanity". The parent was arguing for "egalitarianism" in favor of so-called "reverse sexism" - a point of view I am very sympathe…

Interestingly, the bias happened in the opposite direction from the one you're assuming.

We used to have the neutral "man" and masculine "wer-man" and feminine "wo-man". We still see that "wer-" prefix in the word "werewolf". So it's not that we took the masculine word and started using it as the default, but rather that we took the neutral word and added an assumed masculinity. In that context I see less of an issue with trying to reclaim the neutral-ness of "man" and "mankind".

[1] https://www.todayifoundout.com/index.php/2010/08/the-word-ma...

Re: Surprising gender biases in GPT

#109

Post by roenxi »

flumpcakes wrote: post #53

I will preface this by saying I think people should do whatever it is they want to do, physical limitations willing. My (possibly contentious) question: Is this an OK thing to do? Why are we allowing a vast minority of people to rewrite our social norms? I went to the Barbie film as a man nearing middle-aged and had the most young girls, mothers, teenage girls, and women in general I've ever seen in an audience at th…

> Why are we allowing a vast minority of people to rewrite our social norms?

Any group has a small cluster of leaders that has control (often tiny, like generals in an army). They aren't necessarily in line with the opinions of the group (I'd speculate generals are probably more willing than the troops to send people to their deaths).

It is expected that a small minority of people will write the social norms. The only real question is which minority, and there is a constant scuffle between different minorities to be the ones with power. The ones presently with power think the situation is fine, and the ones without power think that changes is urgent!

Re: Surprising gender biases in GPT

#110

Post by fwungy »

ajsnigrutin wrote: post #44
jorvi wrote: post #10

Earlier quoted context omitted.

It also fits into the larger societal trend of horribly underserving Zoomer and Alpha (maybe young Millennials too) young boys and men. I've heard someone put it very succinctly: people like Andrew Tate are asking the right questions, but giving abominable answers. I can drop some very simple statistics: women have been getting more bachelors degrees since the 80s . Most university gender ratios are 40/60 trending to…

> I've heard someone put it very succinctly: people like Andrew Tate are asking the right questions, but giving abominable answers. People like Andrew Tate are the only ones talking to men in those generations at all. As you said, we've left men behind in schooling which is a bad start in life. After that, we have a lot of programmes for women, both educational ("girls who code", etc.) to business things (women-owned…

Women often eventually discover they want children and are on a tight timeline. Men do too, but they are not on the timeline. They can start families at virtually any time in their life.

By biasing work income towards men it gives women a larger selection of men to choose from to start families with while they are fertile. It also incents them to move sooner rather than work their own career longer to achieve their lifestyle goals, which conflicts with their fertility timeline.

It's sad seeing my older female friends reaching the end of their fertility and realizing they don't want to live lives without children and families, but being at a disadvantage in the relationship market because the men they consider to be at their level are seeking younger women.

"Woke-ism" is looking more and more like the inversion of reality. It is the denial of fundamental realities in the service of highly questionable ideals.

In fact the ideals of "woke-ism" are so questionable that it only survives because they aren't allowed to be questioned.

Post reply on HN