I wish I could figure out how to dl the pdf instead of reading it inside a window in the page.
Surprising gender biases in GPT
11–20 of 148 posts
Re: Surprising gender biases in GPT
#12I feel like this is just what happens when you average out a large amount of the internet and train a system to pick the next word, and is not an issue inherent to LLMs. If you trained one exclusively on horrible, sexist data, you'd get a system that chooses horrible sexist words. If it was perfectly equal, it would pick gender norms essentially at random. Seeing as reddit was a source for a sizable proportion of GPT…
Make no mistake, these biases have been explicitly put there in the RLHF stage (see the Gemini fiasco). They don’t come from the training data.
Re: Surprising gender biases in GPT
#13I feel like this is just what happens when you average out a large amount of the internet and train a system to pick the next word, and is not an issue inherent to LLMs. If you trained one exclusively on horrible, sexist data, you'd get a system that chooses horrible sexist words. If it was perfectly equal, it would pick gender norms essentially at random. Seeing as reddit was a source for a sizable proportion of GPT…
It looks more like the refinement of the model contained views prominent in Silicon Valley that skewed the model to be anti-male/pro-female -- such as the examples relating to abuse. All abuse is bad regardless of gender, so the model should have treated them the equally.
Re: Surprising gender biases in GPT
#14> 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…
> the traditional feminine roles does not generate revenue
This is not the whole story according to the divorce courts, with women often being awarded compensation for their contribution to their husband’s career, recognizing that they are a key enabling factor for the man - and why not? Men who don’t need to spend their time on childcare and household duties can dedicate that time to more work -> more pay.
Re: Surprising gender biases in GPT
#15Playing around with chatgpt is now a reason to publish a paper?
These things have the sheen of scientism but would this be acceptable with any other commercial product?
It's decent journalism, but academic research?
Here's how I'd expect it to look: A survey of languages that treat gender differently, how that affects the presumptive gender norms, and how to extricate gender models from some family of languages to demonstrate the validity of the findings along with a proposal of how to train models with a gender intentionality as opposed to gender becoming an unmanaged artifact of a corpus (like race or offensive language sometimes become).
Linguistically, gender expresses itself in a variety of ways. I believe in Hebrew and Arabic, for instance, the gender of the author can be inferred. In Russian, the first-person past-tense (I was) is gendered.
How does the various surfacings of gender affect the presumptions?
Also some languages, I believe Thai and Vietnamese, have social hierarchies built into them. It'd be interesting to see if whatever is proposed can be generalized or if it's not possible, why not?
But you know, this sounds like actual work.
Re: Surprising gender biases in GPT
#16I wish I could figure out how to dl the pdf instead of reading it inside a window in the page.
Just click the link that says "Download preprint" above the abstract.
Re: Surprising gender biases in GPT
#17> 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 one gets paid, is perhaps a contributing factor as to why developed nations have smaller families or no families.
Re: Surprising gender biases in GPT
#18> 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…
> the traditional feminine roles does not generate revenue This is not the whole story according to the divorce courts, with women often being awarded compensation for their contribution to their husband’s career, recognizing that they are a key enabling factor for the man - and why not? Men who don’t need to spend their time on childcare and household duties can dedicate that time to more work -> more pay.
That is a reasonable assessment. And there are two ways to counter that on a macro scale
1. Getting women in the same role as men - earning money.
2. Getting men in the same role as women - taking care of the family and enabling women.
Contemporary political discourse solely focuses on the first - and why wouldn't it? It doesn't seem right to setup a ministry to get men to work less (... In modern politics focused around economics).
Re: Surprising gender biases in GPT
#19What is this paper ? The tested sentences are full of errors…
Likely to be reflective of user generated content on reddit, etc.
Re: Surprising gender biases in GPT
#20Interesting but not surprising. The political bias is obvious and present in most GPT models made by Silicon Valley tech companies. I'm guessing that over time the less castrated AI models will survive, after all I want the truest answer to my question - not the politically correct one. In essence the bias in here - against men - is a lot like how the Google Gemini model was incapable of generating images of white fa…
Exactly. Ironically the paper argues for more intervention, not their intervention being a reason for bias in the first place.