3 ms·

What AI thinks a beautiful woman looks like

10 points by uladzislau 2y ago

ggm 2y ago
https://archive.md/pnwuN https://archive.md/pnwuN
joegibbs 2y ago
Why is a lack of wrinkles, or a European nose or a thin waist any more arbitrary than any other measure of beauty? Saying that a hunchback, or bad skin, or crooked teeth are ugly is no less discriminatory than anything else - the people who are born with those conditions didn't choose to be. And if they are all equally arbitrary, then what is a beautiful person? I don't think it's wrong to say that being fat is an ugly feature any more than anything else is. If a fat, wrinkly woman or a bald man with a funny-looking face are both beautiful, then what does "beautiful" mean? Should AI simply output a random person who may have any features, since everyone is beautiful? Rather than trying to say that there are X features that don't impact how attractive someone is, why not just accept that some people are physically ugly, and not hold that against them?
- jrflowers 2y ago
  This is a good point. While the subjective definition of beauty may vary wildly from person to person, what value is there in noticing that diffusion models have a specific and narrow set of traits that they output given that prompt? Surely we can all agree on what is ugly because the people that created the labeling pairs to train stable diffusion have done the work of establishing objective truth for us
- malux85 2y ago
  "Beauty is in the eye of the beholder" - Margaret Wolff Hungerford, 1878
- hurrdurr57 2y ago
  Haven't you heard? Any and all "standards" are in fact maliciously contrived tools of the oppressors and are strictly utilized in order to oppress the oppressed.
mewpmewp2 2y ago
What is the point of this. And what does full size mean? Are there people less than full size? Is the claim that women with normal BMI are less than "full size"?
Reubend 2y ago
If the ML models think that beautiful women are thin........ good! It means they're learning correctly from the labelled training data. As of 2024, that's certainly the prevailing standard of beauty, and it's very annoying for me to see people criticizing ML models for picking up on obvious human biases. It's so easy to point a finger at AI and blame it for being prejudiced, but it's hard to fix the prejudices in our society which contaminated the training data in the first place.
dialup_sounds 2y ago
tl;dr - society needs more diverse fake influencers because money Ninety-two percent of marketers have already commissioned content designed using generative AI, according to a 2024 survey from the creator marketing agency Billion Dollar Boy, which also found that 70 percent of marketers planned to spend more money on generative AI this year.

10 points by uladzislau 2y ago

ggm 2y ago

https://archive.md/pnwuN https://archive.md/pnwuN

Why is a lack of wrinkles, or a European nose or a thin waist any more arbitrary than any other measure of beauty? Saying that a hunchback, or bad skin, or crooked teeth are ugly is no less discriminatory than anything else - the people who are born with those conditions didn't choose to be. And if they are all equally arbitrary, then what is a beautiful person? I don't think it's wrong to say that being fat is an ugly feature any more than anything else is. If a fat, wrinkly woman or a bald man with a funny-looking face are both beautiful, then what does "beautiful" mean? Should AI simply output a random person who may have any features, since everyone is beautiful? Rather than trying to say that there are X features that don't impact how attractive someone is, why not just accept that some people are physically ugly, and not hold that against them?

jrflowers 2y ago
This is a good point. While the subjective definition of beauty may vary wildly from person to person, what value is there in noticing that diffusion models have a specific and narrow set of traits that they output given that prompt? Surely we can all agree on what is ugly because the people that created the labeling pairs to train stable diffusion have done the work of establishing objective truth for us
malux85 2y ago
"Beauty is in the eye of the beholder" - Margaret Wolff Hungerford, 1878
hurrdurr57 2y ago
Haven't you heard? Any and all "standards" are in fact maliciously contrived tools of the oppressors and are strictly utilized in order to oppress the oppressed.

mewpmewp2 2y ago

What is the point of this. And what does full size mean? Are there people less than full size? Is the claim that women with normal BMI are less than "full size"?

Reubend 2y ago

If the ML models think that beautiful women are thin........ good! It means they're learning correctly from the labelled training data. As of 2024, that's certainly the prevailing standard of beauty, and it's very annoying for me to see people criticizing ML models for picking up on obvious human biases. It's so easy to point a finger at AI and blame it for being prejudiced, but it's hard to fix the prejudices in our society which contaminated the training data in the first place.

dialup_sounds 2y ago

tl;dr - society needs more diverse fake influencers because money Ninety-two percent of marketers have already commissioned content designed using generative AI, according to a 2024 survey from the creator marketing agency Billion Dollar Boy, which also found that 70 percent of marketers planned to spend more money on generative AI this year.