What is your opinion of AI's depiction of attractive and unattractive women?

[HawkPerception]

Accurate

Inaccurate

A total of 52 votes were cast: 22 girls and 30 guys. See detailed results

[blueonblack22]

It’s showing older women as “ugly”. That’s not fair or right. . Aging is a real thing and even the most attractive man or woman who is 25 will some day be 55 and then 85. Of course most 25 year old women are going to look more attractive than 55 year old women.

However it’s not showing any overweight women anywhere and I don’t think that’s any accident. AI is being preset with certain parameters that are intentionally being applied to not “offend” certain groups of people. A Google Gemini release a few months was so insanely woke that it refused to generate any images of white men. If you promoted it to show a picture of George Washington it would show a picture of a black man.

https://www.reddit.com/r/ArtificialInteligence/comments/1awis1r/google_gemini_aiimage_generator_refuses_to/

Anyway I wouldn’t be surprised if the parameters intentionally don’t show fat faces when you prompt “ugly women”. But unlike being old being overweight is something you can change.

[Kaamraj]

Attraction has a calculative factor behind it including symmetry, proportions, uniformity, lack of visible deformities, etc. I believe that in time with enough training and input AI will be able to calculate mean attractiveness.

[HawkPerception]

I agree

[avery58]

I don't agree. Some of the examples for the "ugly" section were more of average. Some of the plain section were pretty.

Also I don't agree with the "gorgeous" section for Latinas. I've been to Mexico and have seen way prettier girls than that in small towns.

Same for Indian women, beautiful Indians are some of the most stunning women in the world. Most beautiful eyes and hair along with Middle Eastern girls. Those examples were beautiful but I wouldn't say the best.

For the white women section they should change the plain and gorgeous categories around. The plain section was more beautiful than their gorgeous examples.

[TheRealPepperPotts]

Yikes. They are literally all brunettes. I apparently don't even make it on the beauty spectrum for AI, LOL. There's an age bias for sure and it's funny that no one is overweight, really. The faces are also not an equal distance from the camera; some are closer, others are farther. There's a tendency to show more arm/shoulder in the "gorgeous" category than in the "ugly" category, too.

I'll just say AI has a long way to go here... :D

[HawkPerception]

I'd have to agree with you here

[DizzyDesii]

The plains look better than the prettys lmao

[hahahmm]

So I had to track down the original image since I didn't see a link... try:

https://www.reddit.com/r/midjourney/comments/14jlxaz/controversial_question_why_does_ai_see_beauty/

Don't agree with say 70% of what AI says. A lot of the 'plain' ones look better than plain. Some of the 'gorgeous' ones look plain/eh to me. Most of the ugly ones are ugly but not all. Some look plain-ish/average to me.

Although, it's super biased to compare super wrinkly 80 year-old women to 20 somethings.

[hahahmm]

There's definitely some butt ugly 20s and some super old women who aged like wine even if it's rare. Would be nice if it only looked at people of the same age. Also would be nice if it used actual people's face instead of AI but yeah, I get that this would go against the idea of ai DEPICTION.

[RingOfFire]

Interesting question. When I looked at the picture I wanted to see something I agreed with but I didn't.

I even saw a few pictures in the ugly category who I thought were pretty.

I saw some in the gorgeous category that I wouldn't give a second look.

So I think AI needs a little more work. I could have come up with a much better list of ugly and attractive pictures myself without using AI that I think most people would agree with more than they agree with this group.

[RingOfFire]

I also find it interesting that there are no blondes in the picture among the white women. That strikes me as some kind of bias in the programming that I find disturbing.

[hahahmm]

Agree... it's like if a blind person was guessing what guys with eyes think based on who knows what. I bet it can do a lot better at body shape than faces tho.

[MisterWack]

It is hard to see since it is low resolution for me though from what I can see they all look ‘same’. Only difference is “Ugly” seems to be wrinkles, bad hair days, and slightly mature looking.

Definitely not accurate though I would like to see more of a collected sample instead of a select few since there could be some human bias in there.

Beauty Is In The Eye of The Beholder and everyone is attractive to somebody especially when given the chance to showcase their true talents, charisma, character, and personality. We all love, we all feel and give love in our special ways, let us not let algorithms decide for us and let our hearts open up to each other!

[Massageman]

Page seems to get younger the further to the right it goes. So, AI thinks old = ugly. Each step also adds more layers of paint, er, make-up- whether on-skin or in-cloud, so more paint = nice. and so it goes.

Is it any wonder that one of my theses for my degree in college was that humans can never be replaced by computers and mechanization. (I got an "A" for the course, by the way.)

[DishLady]

So clearly there an age bias, but the system is basically just going from least to most symmetrical and with standardized large features (eyes and mouth). It’s interesting but also a little odd.

[andreasderjuengere]

AI's output depends on who programmed them.

They can and will only show results that are ''typical'' - as an analysis of human perceptions and preferences.

Being human myself, I don't need computer assistance to figure out ''attractiveness'' :D

[HawkPerception]

Actually AI isn't like that. Yes you need to program it, but all you're programming it with is the ability to self-learn based on the information on the web and its function (to give you output based on your input, whether that's in the form of text, image, or video)

[HawkPerception]

So yes, it's based on human perception of what attraction is, but on a large scale and because of the scale there may be some incorrect depictions

[andreasderjuengere]

So why is it, then - that ''my'' chat GTP always adds a woke comment to even neutral questions? :D

[HawkPerception]

The developers can add restrictions to what the output is. Not how it self learns. For example... if you add to your memory in personalization that you prefer that the output is not "woke" or politically correct, it will change its output based on your preference.

[andreasderjuengere]

That must be a paid version. I saw no such option when I checked right now.
So... we can set to ''reckless'', '''mean'', 'radical'', ''extremist'' as well? That will be interesting to experience.

[HawkPerception]

Nope it's in the free version. Just type in the prompt "Save in memory that I prefer your output to not be "woke" and politically correct. Just tell me bluntly how things are." You'll see it saved in memory. You can click that "memory updated" to manage what's stored in it's memory

[andreasderjuengere]

Does this also work with the AI's on here? :D - b. t. w.: I had right now asked chat GTP about this as well. The answer took significantly longer than usual. Heh heh.

[HawkPerception]

Not sure exactly what you mean by that. You can't really interact with the AIs here, so the won't respond to you, but it's clear that the devs trained them to respond with different tones and personalities. Like Love Doctor Brad's (or whatever his name is) cringey attempt at being poetic

[andreasderjuengere]

Without interaction, an ''A'' then cannot be ''I''

[AviatorTom]

Assuming the entire chart was created by AI, there's not enough difference between the various categories. If a human survey were done, you'd see much greater differences between them.

[clampfan101]

Many factors cannot be calculated beyond popularity as different people find different things attractive and have different priorities regarding looks.

As for how it could improve its best possible guess, it should start by dividing them by age. I’m curious where you found this and if I could make my own charts of AI guesses.

[HawkPerception]

www.unilad.com/.../ai-ugly-beautiful-women-chart-607509-20230627

[clampfan101]

No longer available, it seems.

[BetaBucks]

It's hilarious that they're all in good shape. And a couple of the uglies and a bunch of the plains, are bangable for sure.

[Alli112]

Nah, lacks diversity overall, labels a ton of people "unattractive" that wouldn't even be considered unattractive just like in society at large (without even going into personal taste, etc). Not sure what data it learned on tbh

[Sharkstealth]

I know and interact with women more beautiful in every category quite a bit. The ugly white category has one if not 2 very pretty girls in it.

[Dongie]

Accurate. But it looked like the sub attractive just had mussed hair and shit.

[Guacamole409]

https://www.youtube.com/watch?v=1onxri0duN0

It's getting better.

[Tiktok_Broccoli_Girl]

Seems like it scaled up the cultural differences from column 2 to column 4.

On 4th column, the blacks are the curliest, Latins have the wavyest hair, Indians have the most jewelry, and whites are the most androgynous.

Column 2 has none of these exaggerations and looks like normal non-retarded non-stereotypical non-steroidal human cats.

[Tiktok_Broccoli_Girl]

Forgot to point out the 4th column's exaggeration of the 4th row: the Asians. There they have the smallest faces.

[SugarSugarplum]

There is very little I could care less about. To me, that's like asking Excel to judge your personality.

[LiamJHayden]

Don't care about it. I have my own standards, and they rarely align with the general public.

[MikeTheBartender]

No overweight women anywhere 🤔

[HawkPerception]

I think this was based more on facial structure 😂

[noohair]

All dark hair.

[HawkPerception]

This wasn't showing every type of girl, body type, hair type, eyes, etc. It chose to only compare a select few. Otherwise you'd see hundreds of not thousands of different depictions.

[Malwi93]

wherer did u get that from

[HawkPerception]

www.unilad.com/.../ai-ugly-beautiful-women-chart-607509-20230627

[JamesRandiDebates]

It looks like AI finds poofy hairstyles attractive