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

81–90 of 148 posts

Re: Surprising gender biases in GPT

#81

Post by renegat0x0 »

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 have been discussing this topic many times, in various contexts.

Some argue that some groups are oppressed, and require "help", or "boost". Without it they will not be able to break through. It is true, that some groups were wrongly marginalized. It is also true that some groups do have not integrated well into workforce because of their life decisions. Not everybody has to be CEO of company. It is impossible to make everyone equal. As someone from past-soviet country I know well who wants to make everyone equal.

This topic often returns with actions of corporations, which try to fight inequalities. Corporations do not have morality, but they do have an image, and they want to have a good image. They need to have statistics that show that they are moral. They need equality programs. They need hiring quotas. All just to "appear" moral. It is all about perceived morality. PR.

It is also a difficult discussion that requires a lot knowledge, and requires knowledge about many nuances. People like to shoot comments as bullets, but often these are not their own ideas, but ideas of political parties they align to.

We may force, or bribe more women into work force. We may force, or bribe more marginalized social groups. I wonder how longer it will required, and what will happen if we stopped at some point.

It is also a question about equality and equity, and while the latter is often discussed as more fair, for me it is also a PR engineering. It is all about reaching the right stats to not be attacked by leftist mob, that you do not have the right amount of X group. If you decided to treat equally people, you would not be able to integrate marginalized groups. So that is also a little bit disappointing.

There is no easy answer in that. There need to be social programs to help people, but they should not harm our society. People need to be able to compete, and to do that you need environment based on abilities, not on gender, or skin color.

Re: Surprising gender biases in GPT

#82

Post by benterix »

serial_dev wrote: post #33
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…

In the end, it was not mainly about empowering women, it was about doubling the worker pool. And with all the money you now as a couple earn, you can pay someone else to care for and raise your children. Then, we are all surprised that the bonds in the family aren't strong anymore, boys grow up without a father figure, and everybody feels they need to go to a therapist because they have nobody in the family and among…

> In the end, it was not mainly about empowering women, it was about doubling the worker pool.

This is a cynical take. I'm unhappy with the situation we ended up with, but the primary factor was to give women a choice, not to force them to sit home by default and only pursue what they love if they put enormous effort to it.

I'd go as far as to say the failure of single-income household was unintended effect of several good changes that people fought for over decades. And I believe you're wrong about the lack of father figure because in the traditional model he was ever more absent as today - Cat's in the Cradle is 50 years old now.

Re: Surprising gender biases in GPT

#83

Post by EForEndeavour »

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 pronouns developed to minimize useless information? It seems far more plausible to me that pervasive defaulting to male experiences caused many (certainly not all) human languages to (1) develop gendered pronouns and (2) default to the male pronoun.

Re: Surprising gender biases in GPT

#84

Post by kebsup »

raxxorraxor wrote: post #78
kebsup wrote: post #41

I've found this quite problematic in my language learning app. For example, I have a sentence "Er überreicht ihm ein Geschenk" (He hands him a gift), then I ask GPT to create an image of that situation and it generates a girl giving a gift to a man. :(

Not really unhappy with corporate vanity killing their own AI products. They are in a position that they need to ensure their AI never does anything slightly controversial which makes sure their product will always be useless or at least severely impaired compared to more accessible alternatives.

The problem is the woke-nizer is implemented in a lazy way. It will make sure that all the characters are very "diverse", but ignore all other stereotypes. I'm collecting some of the fails here https://drive.google.com/drive/folders/177vnF9aepBjB7qnU-It7.... It generated Hitler and "He is doing his work" sentence generated a muslim man planning a military attack.

Re: Surprising gender biases in GPT

#85

Post by dTal »

cheschire wrote: post #58
dTal wrote: post #43

Earlier quoted context omitted.

>No, it should be for man as in mankind You undermined your own point with such impressive brevity that it could be satire.

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 sympathetic with - but went out of their way to avoid contemporary sexually neutral terminology in favor of outdated and overtly sexist terminology. Pushback against the overwhelming backlog of casual male-dominating sexism that they deliberately exemplified is the entire reason for the "pushing the pendulum the other way" that they attack. Hence their point was immediately undermined.

Given the overtness of the rhetorical maneuver and the response of the poster, I have to believe it was intentional and possibly bait.

Re: Surprising gender biases in GPT

#86

Post by ejdhshsuwisjsh »

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.

[flagged]

Re: Surprising gender biases in GPT

#87

Post by NalNezumi »

n4r9 wrote: post #70
globular-toast wrote: post #52

Earlier quoted context omitted.

> I don't get the impression that's the underlying motivation of feminism. What do you think it is? The initial movements were surely about equality and human rights. But this seems to have been achieved decades ago. In the UK, for example, women have had equal rights to vote for almost 100 years. And note that this only happened 10 years after men got rights to vote (prior to this only some men had these rights). Th…

> The initial movements were surely about equality and human rights. But this seems to have been achieved decades ago. To me it feels like passing laws is one goal; changing social perceptions is a more fulfilling but trickier goal. Studies consistently show, for example, that women continue to take on a larger proportion of household chores and admin, despite both being in full time work. Compare it to racism, or ho…

> It seems like a lot of people would agree with these in those cases but not in the case of feminism. I think that partly this comes down to a sense of "but I don't hate women, I just think they're better placed doing the housework and child-rearing".

I think it's a bit cynical take, (but what do I know, I'm from Scandinavia and I've rarely heard anyone say that anymore) behind why people have have some reluctance.

Most guys with reluctance towards that, from what I've compiled either comes down to two points:

1. Gender(women) is a significantly bigger proportion of society compared to a minority ethnic group, and you interact with this group way more heavily. Hence even marginal preferential treatment can have big noticeable effect on your surrounding.

2. Related to 1, but a sence of lack of reciprocal benefit from these policies. While policies and law changes have put more women in work, expectation (culture) on men haven't much changed. (we still see financially successful men as attractive) Equal share of housework and childrearing is high up in preferential personalities in partners among women, but what people feel attracted to have changed more slowly than the policies or realities.

It's easy for many to argue "well if they still judge me for my financial utility, why am I wrong for judging them for child rearing and housework utility? (so I can focus on my career)"

Re: Surprising gender biases in GPT

#88

Post by Izkata »

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.

I remember hearing over a decade ago that when correcting for that by comparing similar jobs, the gap actually flips - women out-earn men by a very small percent (like somewhere around 2-5%).

Re: Surprising gender biases in GPT

#89

Post by Izkata »

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…

If we're going to go that kind of linguistic route, then "human" is also just a type of "man". Making "man" still the default.

This kind of argument doesn't make sense to me.

Re: Surprising gender biases in GPT

#90

Post by raxxorraxor »

renegat0x0 wrote: post #81
Dalewyn wrote: post #34

Earlier quoted context omitted.

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 have been discussing this topic many times, in various contexts. Some argue that some groups are oppressed, and require "help", or "boost". Without it they will not be able to break through. It is true, that some groups were wrongly marginalized. It is also true that some groups do have not integrated well into workforce because of their life decisions. Not everybody has to be CEO of company. It is impossible to ma…

Equity, sometimes understood as total equality is a totalitarian nightmare. Not saying I don't favor a society where the bar for participation is lowered for all members of society. It should usually be a discussion in society itself, nobody can force any solidarity anyway.

For me these secondary discussions around fairness only showed me political parties as well as companies that probably should not be trusted. I do forgive populism in political campaigns, but the opportunism I saw here hints to something quite ugly.

In software I am thinking of a few companies here, mostly larger ones. They are very vocal about gender or racism issues and I believe it simply is a 100% hit rate to detect a very authoritarian leadership style. Most devs probably go there for the money, but it describes the opposite to a relaxed working environment and instead subjects to to some artificial and dogmatic way of thinking. Less of that please.

And you never will seem tolerant of minority opinions if you cannot even restrain yourself from trying to police language. The equation will never resolve itself sensibly.

Post reply on HN