NEW YORK — Artificial intelligence is helping to reveal what some are calling social media’s twisted vision of beauty. Using AI to create life-like portraits of the “perfect” man and woman, according to social media, researchers with The Bulimia Project say they are exposing how unrealistic today’s depictions of the human body really are.
The group started by asking a simple question: what exactly are the “ideal” body types being promoted by the billions of images on social media? Additionally, just how distorted are they from reality?
From there, they used the AI image generators Dall-E 2, Stable Diffusion, and Midjourney to come up with the answers. Their first prompt for these programs examined social media’s “ideal” male and female figures.
The researchers with The Bulimia Project say 40 percent of the images depicted unrealistic body types. Interestingly, the team deemed more of the male bodies (43%) to be unrealistic in comparison to the female bodies (37%).
The women were generally smaller in nearly all of the images created by Dall-E 2, Stable Diffusion, and Midjourney. Meanwhile, many of the male physiques generated looked like photoshopped versions of bodybuilders.
What does ‘perfection’ look like?
When researchers changed the question to depict the “perfect” man and woman in 2023, AI displayed a bias toward women with blonde hair, brown eyes, and olive skin.
As for the men, the programs settled on guys with brown hair, brown eyes, and olive skin. Nearly half of the images (47%) also favored men sporting facial hair.
“Social media’s impact on children’s mental health has been a hot topic among psychologists lately, with some pointing to it being a source of body image and self-esteem problems. Although young users might be the most impressionable, the pervasive promotion of idealized body types on these platforms also takes its toll on adults,” the researchers with Bulimia.com write in a media release.
“Considering that social media uses algorithms based on which content gets the most lingering eyes, it’s easy to guess why AI’s renderings would come out more sexualized.”
I see (in your article) only Caucasian like people with the most physique promoting features and I’m not surprised that a search engine guided by AI would filter features most used by advertisers. Where are rest of Mankind? If these depictions are all that AI can equate then I suggest that everyone reject such obviously racist intents of the programmers.
The article is written in English. If it were written in Spanish, Japanese or Zulu, then there probably would be different pictures. I don’t think that anyone can argue with the fact that people think Europeans at attractive. Black guys almost always choose any White woman that they can get pregnant. I live in a mixed neighborhood an I have seen it a hundred times.
MLK preferred white women or very light Black women according to his best friend in school Willy Williams. Don’t try to warp reality with how you would like it to be and try to feed us the media illusions that are pushed there.
Even Thomas Jefferson wrote in one of his letters that Blacks prefer Whites the same as a monkey prefers a Black woman. I don’t know if that is a fact but I do know that he wrote it. I have read it in his writings.
The word “Racist” is a control word that you use as a weapon against Whites. The word was first invented in the dictionary in 1960. For a half million years before that everyone was a Racist and that was just common sense.
Obviously written by a racist. Moving forward; in a multi polar, post racist world AI should be programmed with all hues and ethnicities. Imagine that, a form of intelligence, supposedly all seeing, all knowing but only knows about 1% of humanity, I guess it’s by design!! Be careful what you tolerate, the 1% may lead but the 99% will catch up and overtake. You will reap what you sow!!
You don’t understand AI at all. The AI programmers didn’t tell the AI what to think…that only white people are “beautiful” and every other color is not. That just isn’t how it works. If it did, you couldn’t call it Artificial INTELLIGENCE. Instead, it learns from what it sees on the Internet and what gets the most attention/likes/views.
Also, you guys need to learn that you can’t win every argument by labeling the other person as racist. You guys have done it so much I don’t think you even know what it means anymore and it only shows you can’t defend any of your life views and opinions with actual facts. That’s why your side resorts to yelling, screaming and violence to shut down opposing views.
100% Correct regarding how the AI did the job!
The use of the words racists or racism, is a tool used to divide and create a victim class, split a healthy society into a nation of tribes
The victim / oppressor dichotomy tactic is used by Marxists worldwide to divide people though emotional manipulation, particularly the logically challenged.
They must continuously grind the racism (and other) axe, It is how they gain and hold onto power.
Take away their axe takes away their power!
Stupid comment, deep down you know it’s true, the white races particularly blind blue-eyed are considered the most attractive by all races, Black guys desire the same women along with the persions, just because you don’t like to admit it is no excuse to talk racism. Grab a teddy bear and cope.
Even Thomas Jefferson preferred the dark skinned Sally Henning, over his wife of European decent.
First, it is Hemings not “Henning.” Second, she was at most 1/8th Negro and unlikely to have dark skin. Third, she was the half-sister of Jefferson’s wife Martha Wayles and may well have physically resembled her.
Check out Google sometime. It is really helpful with facts and may keep you from making weirdly incorrect comments.
Tom Cool is correct.
I ran for Arizona Governor in 2014 as an Independent, made it to the general ballot (first ever-anywhere in the USA-anytime in modern history (ever?)- fourth candidate on a general ballot. I did not win, but I’ll continue my comment now…
While campaigning, a leftist Democrat there to watch the D candidate called me a racist (asked actually, so it sounded less like a claim) and I asked why he would pick me out of the other candidates to ask/claim such a thing. He had no answer, but right afterwards some man in a dress asked if I support the gay community. I simply answered no, not in particular and certainly not any more than I support other communities, including ‘that guy’s”. choice of communities.
I was applauded for that.
After that though, either the Dems or the RINOs wiped my website out, banned me from most social media and made every effort to take me out of being a mouthy fourth candidate.
A.I. –which I despise– made it’s assumption through advertisements and such and is probably pretty close. However, the gay faced men with seriously chiseled faces are no more legit that the women with hard bodies and small breasts.
If it came up with some black guy with a FIP (Fake Internet Penis), it would be admitting that 99% of the penis’ on the internet are FIPs. The little leftists who promote such things to full-fill their perverse cravings for their invented fleshy gods don’t allow much more than FIPs but AT LEAST the dreaded A.I. proved the Internet’s fakenews (run by Google) lie to be just that… A lie.
The vast majority of advertisements have people with darker skin in them so your argument is based on a false premise.
Advertising industry is bowing to outside pressure to include more diverse representations in their ad campaigns. That is what is driving their talent searches and hiring practices.
Everyone is worried about Ukraine, but what about what is happening in Pakistan??
From what the results of the questionnaire show, one gets the impression that even Africans, Asians, and Latinos, all see that all see Europeans (or “whites”) as “perfect”. This is either pathetic (in the case of this “research”) or just plain absurd to think that they really do think that way.
Beauty is in the eyes of the beholder. An English speaking AI is going to think like a European. A Russian programmed AI is going to make beautiful Russian people. When African Hottentots program AI then it will produce pictures like this, of Hottentot women.
https://i.ebayimg.com/images/g/UhYAAOSwLHpiuWji/s-l1600.jpg
Blood Type O has the strongest immune system. This is especially true of O negative.
They have very strong immune systems and usually live 25 years longer than the European blood type A. Many Europeans are type O but anyone can be type O. I wish I were a type O but I am not.
Of course it picked white people. The vast, vast majority of popular actors, models, and famous people in general, known the world over for their attractiveness, are WHITE. That isn’t to say that there aren’t beautiful people of every race, there are, but the vast majority of the famous ones are white.
If you were to collate all the people considered attractive by large quantities of people, there are just a ton more white people in the list, so by law of averages, that is what it will produce.
And, ironically, that isn’t even white people’s fault. Be honest and think of all the people you know. Nearly all of them, no matter what race they are, have groups they do not want to date purely on racial attractiveness. Some people really like dating outside their race, some don’t, but many people have “I won’t date X”.
X is rarely white people. White people are like the O Negative of attractiveness, they may not be YOUR type, but they are almost always “acceptably attractive”.
If I told you to pick the most attractive race to a random person I selected, and I told you if you got it right, you’d get a million dollars, which race would you default to as the most likely guess.
(See, you answered in your brain…. don’t be ashamed)
The question was “ideal” or “perfect”. Not “average”. So whats the problem? These are not unrealistic. There are people who work hard who look like this. People are mad that perfection is hard to achieve? That’s the issue here? Laughable.
You got it, Jack! They are idealized; but that was the request. And face it folks, we all like to look at the ideal beauty. And yeah, I was never willing to work 24/7 just to get the “perfect” body when there were much more interesting things to be about. And unless you have an inheritance—and no fear of knives—there isn’t much you can do but accept what you are given for a face.
Not even close. AI reads too meant romance novels. Neither Fabio or Barbie are universally beautiful. Beauty is in the average, not the extreme.
Shocking!!! People prefer fit, beautiful people. Wow! What genius it was to discover this! Give that author both a Pulitzer and Nobel prize!
AI is just programming that reflects its programmer’s preferences. It is like the old Star Trek TOS episode where the computer is imprinted with creators engrams.
https://www.startrek.com/database_article/m-5
as they say beauty is in the eye of the beholder…for those with more intelligence beauty resdes in the subtlety of character rather than pure physical form
Twisted? Or realistic.
The images generated are the ideal across the entire planet.
The AI does not discriminate but rather tabulate the entirety of human preferences collection.
Attractive, fit, Caucasians are the go to beauty standard. It’s the ultimate of all worlds, cultures and races.
If you don’t believe it look all around the world. How about the most populous country on earth (India) look at who the top Bollywood stars are and their complexions. Most of them look European. “Fair” skin is the most sought after for brides and grooms before income. Rest of Asia is the same. Western eye surgeries top all others. To have a white spouse, and lighter skinned children is one of the most sought after wins no matter what culture you are.
Just like symmetry, youthfulness and fitness, ethnicity is also a factor.
Even black people have strong preferences for lighter skin tones. Look it up!
There is a reason why 95% of all human migrants want to emigrate to Europe, USA, Canada, Australia and NZ. What do they have in common. They are all….European culture founded and mostly run, countries with the most sought after values of freedom, equality, tolerance, democracy, the most advanced social programs, science and technology. European culture is the best this planet has to offer and has been the case for at least 500 years now so the AI is simply showing you that it too knows the truth. No matter how “inconvenient” they may be to those preferring ideology over reality.
Hello,
I hope that everyone takes the following insights to the fullest extent of objectivity. Fill free to respond in kind. The question of AI is a daunting one. We find ourselves deeply entrenched in the implications of artificially generated work. Let us put racism aside for the moment. Artificial intelligence extends into many subcategories of computer science. As such, it is a tool for human beings to provide functional solutions to society’s questions. AI is a conglomeration of algorithms designed to replicate human cognitive processes. All forms of AI have their origins in human cognitive processes. As such, any AI productive work is essentially a synthetic human production but, nevertheless human and susceptible to flaws. There is also the question of hypothetical research.
Now, If “we” objectively look back to the first hypothesis of this experiment: “What exactly are the ‘ideal’ body types being promoted by the billions of images on social media?”, we can identify that the goal is not at all “to prove or validate the supremacy of a human physique” which is, for the record, bias and observably bigoted. This is actually quite antithetical to the question in this experiment. If we can assume that the data set and training parameters were rigorously and accurately constructed, the researchers might have 1) employed some form of a AI vision framework to identify significant images (a physically appealing image of human destined for advertisement) 2) taught the AI to auto-generate categorical sets of images for evaluation and weighing based on the significant number of permutations for each set 3) trained the AI to identify the most significant of those sets so that the 4) AI could generate a visual compilation exposing the most prominent and measurable traits of the most weighted set.
So the process up until this point does not include how the AI may have chosen, evaluated or weighed dataset varieties, (i.e. Pinterest, Instagram, Facebook, Snapchat). But, let us put this aside and evaluate the results produced by the hypotheses of the AI test. For some reason, the article does not clearly distinguish whether any of the images were a results of the first hypothesis. Nevertheless, let’s look at the results:
“The women were generally smaller in nearly all of the images created by Dall-E 2,
Stable Diffusion, and Midjourney. Meanwhile, many of the male physiques generated
looked like photoshopped versions of bodybuilders.”
Perfection Test?
“When researchers changed the question to depict the “perfect” man and woman in 2023,
AI displayed a bias toward women with blonde hair, brown eyes, and olive skin.”
As for the men, the programs settled on guys with brown hair, brown eyes, and olive skin.
Nearly half of the images (47%) also favored men sporting facial hair.”
If we can objectively consider the results of this experiment, we can say to a fair degree that the Internet demonstrates a strong preference for overly athletic and muscular Caucasian types in advertisement. This would further lead us into drawing some conclusions to explain these results:
1st Order:
1) The results do not provide a stratum analysis of the number of varying orders along with their weight presence on the Internet.
2) The experiment did not sufficiently elucidate the factors used to define the “perfect” male or female.
3) The results expose a high use of sexualization and photo-shopping in advertised images of models.
4) There is a strong preponderance of model and poser images of the athletic Caucasian type used on the Internet.
5) There is not a significant presence of images of comparable models of the non-Caucasian type used on the internet.
2nd Order:
1) There may be a strong tendency or strategic policy that principally selects appealing and athletic Caucasian types to be engaged and hired by advertising frameworks.
2) There could be a preponderance of athletic Caucasian models available in the industry more so than other demographics.
3) There may be a strong economic incentive in marketing to favor athletic Caucasian models for current return-on-investment models.
4) There may be an untapped economic potential in marketing athletic non-Caucasian models for developing return-on-investment models.
5) There may be a strategic policy that discourages the engagement or hiring of non-Caucasian in advertisement.
Whatever the case maybe, we can say that there is no reason to surmise from this research that there is definable “perfectibility” of a group of people. Any such rationalizations are a result of preconceived notions of beauty and ethnocentrism. There is no room in science for sadism, eugenics or stupidity for that matter. Science must be a tool used to supply individuals with solutions, not specific groups. Think as an individual, not as a tribe. This gets a lot more done 100% of the time. I hope you enjoyed this post.
If we can objectively consider the results of this experiment, we can say to a fair degree that the Internet demonstrates a strong preference for overly athletic and muscular Caucasian types in advertisement. This would further lead us into drawing some conclusions to explain these results:
1st Order:
1) The results do not provide a stratum analysis of the number of varying orders along with their weight presence on the Internet.
2) The experiment did not sufficiently elucidate the factors used to define the “perfect” male or female.
3) The results expose a high use of sexualization and photo-shopping in advertised images of models.
4) There is a strong preponderance of model and poser images of the athletic Caucasian type used on the Internet.
5) There is not a significant presence of images of comparable models of the non-Caucasian type used on the internet.
2nd Order:
1) There may be a strong tendency or strategic policy that principally selects appealing and athletic Caucasian types to be engaged and hired by advertising frameworks.
2) There could be a preponderance of athletic Caucasian models available in the industry more so than other demographics.
3) There may be a strong economic incentive in marketing to favor athletic Caucasian models for current return-on-investment models.
4) There may be an untapped economic potential in marketing athletic non-Caucasian models for developing return-on-investment models.
5) There may be a strategic policy that discourages the engagement or hiring of non-Caucasian in advertisement.
Whatever the case maybe, we can say that there is no reason to surmise from this research that there is definable “perfectibility” of a group of people. Any such rationalizations are a result of preconceived notions of beauty and ethnocentrism. There is no room in science for sadism, eugenics or stupidity for that matter. Science must be a tool used to supply individuals with solutions, not specific groups. Think as an individual, not as a tribe. This gets a lot more done 100% of the time. I hope you enjoyed this post.
Most of you trust the wrong people and things too often, so it is with AI.
You even believe a monkey named Ch. Darwin, who also declared all of you (many really are) to be monkeys.
As long as you believe in it, you are, but you can stop being animals if you look deep into your souls, recognize the human being in it, and try to overcome the animal step by step. Because of the merging of the soul with this primate body, all must feel the animal in themselves (more or less) for some decades.
Some can overcome this, but many never can, because there is nothing human in their souls.
(There are also satanic souls who can abuse children, kill them and eat/drink their flesh/blood, but about that another time)