Meet the Professor Who Says Facial Recognition Can Tell If You’re Gay

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
According to Michael Kosinki, psychologist and assistant professor of organizational behavior at Stanford University, facial-recognition software can be used to determine whether someone is gay. Last year, Kosinki and a research associate used a machine-learning system on 35,326 photographs from dating websites and found that it was able to distinguish between photos of gay and straight people with a high degree of accuracy.

There was an immediate backlash when the research – dubbed “AI gaydar” – was previewed in the Economist magazine. Two of America’s most prominent LGBTQ organizations demanded that Stanford distance itself from what they called its professor’s “dangerous and flawed research”. Kosinski received a deluge of emails, many from people who told him they were confused about their sexuality and hoped he would run their photo through his algorithm.
 
But are you gay if you don't think you're gay, but the AI says you are?
 
upload_2018-7-7_19-11-12.png
 
I'm curious how they controlled their research. Was it put up against a human group? That's usually how AI is judged (percent accuracy above humans).

Btw, let's not forget who this guy is:

Russian prime minister, Dmitry Medvedev, and several members of his cabinet convened in an office building on the outskirts of Moscow. On to the stage stepped a boyish-looking psychologist, Michal Kosinski, who had been flown from the city centre by helicopter to share his research.

And ...

Kosinski’s “stuff” includes groundbreaking research into technology, mass persuasion and artificial intelligence (AI) – research that inspired the creation of the political consultancy Cambridge Analytica.

So he's being flaunted around by Russia (whose government is homophobic to the point of persecution) for his technology to rapidly find gay people. This is a modern version of the Jewish tests.

His research was probably biased (via funding, patriotism, or upbringing) and I'd be surprised if it was more accurate than a human.

You see, photos scraped from dating sites likely came from straight and gay groups. Naturally, I'd assume the gay guys who are out of the closet likely don't hide the fact they're gay on these sites, so the algorithm may have just picked up on poses, colors, and other things that a human could perceive. The real test is how accurate it is identifying gay people who are not openly gay.

And to be quite frank, I wouldn't doubt if in 30 years we find out that Putin was gay or bi. It's like the anti-gay preacher who is really a closet homosexual. They hate the fact that they're like that and decide to go out of their way to overly prove they're not. Putin exhibits this type of behavior in a lot of his photo ops, they're all overly "manly" (shirt off, doing pure man things, etc.) photos and it's become a meme because of it.


TLDR;
- Researcher is connected to Russia and has a special interest in pleasing them and their homophobic tendencies.
- This could be dangerous tech because of Russia's recent treatment of homosexuals.
- I wouldn't be shocked to find out Putin is gay.
 
Last edited:
According to Michael Kosinki, psychologist and assistant professor of organizational behavior at Stanford University, facial-recognition software can be used to determine whether someone is gay. Last year, Kosinki and a research associate used a machine-learning system on 35,326 photographs from dating websites and found that it was able to distinguish between photos of gay and straight people with a high degree of accuracy.

There was an immediate backlash when the research – dubbed “AI gaydar” – was previewed in the Economist magazine. Two of America’s most prominent LGBTQ organizations demanded that Stanford distance itself from what they called its professor’s “dangerous and flawed research”. Kosinski received a deluge of emails, many from people who told him they were confused about their sexuality and hoped he would run their photo through his algorithm.

This reminds of older research that pretended to identify potential criminals from face analysis. The research was a failure.
 
I will say it's a flawed test based on their source data. They used photos from dating websites. People on dating sites have a huge tendency to flaunt what they deem to be their strongest traits to promote their sexuality. That's the entire god dammed point, finding someone with similar sexual preferences because sex is the desired outcome.

I'm sure any one of us can get on Tindr right now and guess with relative accuracy what each person is based on just their profile pic. We do it subconsciously anyways, "I bet she's a slut", "He looks like a douche", "They are definitely gay." For all those other cases of people with relatively conservative pics, there's a bio page for everyone to read.

I'm willing to bet that if they did just head shots of people and they would have nothing special about them in the photo, the test would have much worse accuracy.
 
When people used phrenology (head shape) to judge people's intelligence, the theory didn't have much to stand on other than being a "super handy" tool for discrimination purposes.

However, since hormones directly affect your physiology (even bone structure), surely anyone with some experience with people can derive assumptions about hormonal conditions of another person by merely looking at them. Just because this theory does have legs to stand on, it doesn't mean that it's any more useful. It's just a "super handy" tool for discrimination. Once a theory is designed into a functioning system with some degree of accuracy doesn't immediately become codified as a lawful tool, but I'm sure in some areas this _will_ find a way around other fields of broken logic, and be used to round up gays where it's illegal to be gay. The problem is not found in the tool, it's found in broken logic to ratify use of the tool.

Basically, I get why people are upset by this research, but it's clearly being upset at the wrong things. It's basically the gun-control debate with high cheek bones.
 
Last time I used a facial-majiggy it though I was a girl in her teens.

I had a beard then. It may have been blonde, but it was a beard.

I don't know how much the tech has developed in the last six months or so, but, yeah... Not sure I'm 100% believing it can stare into my soul.
 
I've always known I was straight, never attracted to men or that kind of thing.

I was playing TF2 against some 12 year old kids last night and doing a fine job winning.
That's when I found out from the internet that I was "gay"


Seriously though, so what if AI can tell if someone is homosexual?
It's not like the rest of the world can't tell in 5 minutes or less.
 
If the image is Professor Michael Kosinki. he looks... well, gay
don't need AI for that
 
Kosinki's efforts aside, it's probably a fairly safe bet that most self respecting proctologists could spot one a mile away.

Note: I purposely avoided using the phrase "up close and friendly" simply because I'm not sure how friendly such doctors get when involved with their front facing tasks.

In news elsewhere, at one major hospital, an emergency room nurse servicing a patient who showed up with a stuck foreign object was overheard giggling "there's a light at the end of the tunnel and it looks like a battery operated one."
 
Last edited:
I can picture some dudes writing this guy with, "hi, whenever I watch the Rock in a movie, I feel soooo................ Please can I send you my pic. I gotta know."
 
I'd be surprised if it was more accurate than a human

i would be VERY surprised if it was LESS accurate than a human.

gay-dar is a task that males and females fail a lot.
many males have tried to hit lesbians, and many women wasted time trying to seduce gay men.
for an AI to be LESS accurate than humans, it would need to perform worst than random. :ROFLMAO:
 
I've always known I was straight, never attracted to men or that kind of thing.

I was playing TF2 against some 12 year old kids last night and doing a fine job winning.
That's when I found out from the internet that I was "gay"


Seriously though, so what if AI can tell if someone is homosexual?
It's not like the rest of the world can't tell in 5 minutes or less.
I never knew how promiscuous my mother was until I started playing call of duty.
 
This is certainly interesting, and it is not neccesarily a bad thing for the gay community, like they seem to be suggesting.

It would seem to prove that there is a genetic component to being gay, not just a "phase" or something that can be "prayed away".

Genes often have different features that go hand in hand. For instance, when foxes at fur farms in Russia were bred based on how docile they were (in an attempt to avoid the aggressive ones) the offspring quickly started developing floppy puppy dog ears, for instance.

So if there is a genetic or a combination of genes that results in homosexuality, it would not be surprising that it also expressed expressed itself in other ways, like in facial features.

Unless - of course - the algorithm is just picking up on the fact that gay men tend to be better groomed :p
 
Last edited:
So if there is a genetic or a combination of genes that results in homosexuality, it would not be surprising that it also expressed expressed itself in other ways, like in facial features.

In the future, there is a single sex and gender, and it's name is Robert Paulson.

Jokes aside, look up "methylation gay", should find the info in no time.
 
I just checked my genes and other than needing a good washing, they seem to be just fine.
 
for an AI to be LESS accurate than humans, it would need to perform worst than random. :ROFLMAO:

I agree that people get it wrong. I disagree that people get it wrong on the order of random. I also think, most importantly, that it generally does not really matter.
 
I don't care so much about the gay people.
Duckface guy is easy to identify.

What's scarier for me is how many trannies I can't tell are trannies by looking at their face on dating websites.
Real women have more manface than these trannies.

Plastic surgery is getting too good, and they even have an adams apple reduction surgery.
Gimme an AI face rec for that.
 
I wouldn't doubt if in 30 years we find out that Putin was gay or bi. It's like the anti-gay preacher who is really a closet homosexual. They hate the fact that they're like that and decide to go out of their way to overly prove they're not. .
True that, better said by Christopher Hitchens:
I always take it for granted that sexual moralizing in public is a sign of hypocrisy or worse, and usually a desire to perform the very act that is being condemned.
 
That's nothing. I wrote a program that uses facial recognition to determine if someone has a really big penis and is really good at having sex with lots of women all the time, sometimes even three ways (and always the 2 girls 1 guy type). It correctly identified me as having all these traits and also said I was one of the manlier men on the planet.
 
I'm curious how they controlled their research. Was it put up against a human group? That's usually how AI is judged (percent accuracy above humans).

His research was probably biased (via funding, patriotism, or upbringing) and I'd be surprised if it was more accurate than a human.

if you had bothered to read the article:

"the algorithm was trained to distinguish the two in 81% of cases involving images of men and 74% of photographs of women. Human judges, by contrast, were able to identify the straight and gay people in 61% and 54% of cases, respectively. When the algorithm was shown five facial images per person in the pair, its accuracy increased to 91% for men, 83% for women. “I was just shocked to discover that it is so easy for an algorithm to distinguish between gay and straight people,” Kosinski tells me. “I didn’t see why that would be possible.”"

Doesn't mean it's something that should be used, it wasn't even really intended for this, the gaydar part of it was an effect he noticed.
 
if you had bothered to read the article:

"the algorithm was trained to distinguish the two in 81% of cases involving images of men and 74% of photographs of women. Human judges, by contrast, were able to identify the straight and gay people in 61% and 54% of cases, respectively. When the algorithm was shown five facial images per person in the pair, its accuracy increased to 91% for men, 83% for women. “I was just shocked to discover that it is so easy for an algorithm to distinguish between gay and straight people,” Kosinski tells me. “I didn’t see why that would be possible.”"

Doesn't mean it's something that should be used, it wasn't even really intended for this, the gaydar part of it was an effect he noticed.

Here is my whole issue with AI system...the algorithm is nearly 100% dependent on the quality of the input. This means, in reality, most AI algorithms are biased since the data inputs are no truly random and diverse. It is the very reason why Facebook's facial recognition software is "racist"...because that is how it was trained (mostly wide dudes made up the input set).
 
So he's being flaunted around by Russia (whose government is homophobic to the point of persecution) for his technology to rapidly find gay people. This is a modern version of the Jewish tests.

TLDR;
- Researcher is connected to Russia and has a special interest in pleasing them and their homophobic tendencies.

Don't believe everything the State approved media tells you.

There is a gay guy running for Mayor of Moscow. He's not dead yet, his HIV might kill him before government goons do.

Yes Russia banned homosexuals outside of Russia from adopting children, but homosexuals WITHIN Russia are able to adopt children. Something only recently allowed in all 50 states here in the US.

The ban on "propaganda" towards minors, simply says you can't encourage children to be homosexual. It doesn't mean homosexuals aren't allowed to be around children or have children. I would think even here in the States, we don't particularly want heterosexual adults encouraging young children to seek out certain sexual acts.

Don't accept the paradigm that you are either fully open to accepting homosexuality -or- you are homophobic. That is not a realistic expectation. They believe it is a deviant behavior. You might not agree with that, but that belief is not unreasonable. It's somewhere between those 2 paradigms.
 
Back
Top