Become A Member | Forum | Profiles | Personals | Classifieds | See Who's Online ...
 
View Topic
  Message Boards : Computers, Tech & Net : View Topic : 15 Posts, Page 1 of 1
  HomeNewNoticesHot TopicsPollsStatsBlogs Login / Register
 
New AI Can Tell From A Photo Whether You're Gay Or Straight
 
# 1 : Friday 8-9-2017 @ 12:11
 
 
Artificial intelligence can accurately guess whether people are gay or straight based on photos of their faces, according to new research suggesting that machines can have significantly better “gaydar” than humans.

The study from Stanford University – which found that a computer algorithm could correctly distinguish between gay and straight men 81% of the time, and 74% for women – has raised questions about the biological origins of sexual orientation, the ethics of facial-detection technology and the potential for this kind of software to violate people’s privacy or be abused for anti-LGBT purposes.

The machine intelligence tested in the research, which was published in the Journal of Personality and Social Psychology and first reported in the Economist, was based on a sample of more than 35,000 facial images that men and women publicly posted on a US dating website. The researchers, Michal Kosinski and Yilun Wang, extracted features from the images using “deep neural networks”, meaning a sophisticated mathematical system that learns to analyze visuals based on a large dataset.

The research found that gay men and women tended to have “gender-atypical” features, expressions and “grooming styles”, essentially meaning gay men appeared more feminine and vice versa. The data also identified certain trends, including that gay men had narrower jaws, longer noses and larger foreheads than straight men, and that gay women had larger jaws and smaller foreheads compared to straight women.

Human judges performed much worse than the algorithm, accurately identifying orientation only 61% of the time for men and 54% for women. When the software reviewed five images per person, it was even more successful – 91% of the time with men and 83% with women. Broadly, that means “faces contain much more information about sexual orientation than can be perceived and interpreted by the human brain”, the authors wrote.

http://www.theguardian.com/technology/2017/sep/07/new-artificial- etc ...

Interesting - I'd imagine humans are just as good at picking up on the expressions and grooming styles but the AI is better at detecting the gender-atypical features.
Reply
 
 Recent Message Board Topics
Dreams And Meanings
What Song Are You Listening To Now?!?!
Avast, Me Hearties!
Funny Pictures To Brighten Your Day - Funny Edition
The Three Days Of Darkness - November 24 - 26, 2016 A.D.
How Do You Feel Today? Part 3
Rant Thread 7
Insomnia.
 
Hey! If you enjoy shooting the breeze with like-minded people, check out
our Message Boards
• Advice • Coming Out
• Computers • Current Affairs
• Discussion • Food & Drink
• Going Out • Humour
• Health • Music
• Newbies • Sexual Issues
# 2 : Friday 8-9-2017 @ 12:32
 
 
It only works for white people .... apparently
Either they didn't bother sampling outside that group... imagine that
Or the parameters it applied don't apply outside that group .

its a bit frightening to imagine what a group like isis might have done with something like this and Russia and what they've been doing in Chechnya! the Russians could very easily trial a technology like this there before rolling it out across the country.
you'd have to question why they are doing research like this at Stanford and who is paying for it?
there are many countries in the world where this technology is unsafe


Reply
 
# 3 : Friday 8-9-2017 @ 16:00
 
 
You guys aren't reading it correctly.


The computer used images from a dating website.

first of, all of those images are filtered through the human eye and mind thinking "this photo makes me look good" "this is a photo of me dressed up" "this photo makes me look suitable for dating X person"

In short the users are choosing the photos to intentionally identify as, for example, a single gay man by their own definition and partially by a socially influenced definition aswell.

If anything all the computer has been able to do is pick up on the intended effect of the photo by the user who uploaded it ("grooming styles" and "gender a-typical features").

It says nothing about how, for example, a person behaves, looks or dresses, or "chooses photos", if they are in the closet, if they are choosing a photo for a CV, if they are just taking a family picture or a sightseeing shot.

My guess (and the logical thing to assume) is that same computer wouldn't have a snowman's chance in hell of being an effective "automatic outing machine" for people who are not intentionally feeding it "gay or straight photos" from a dating site, and, judging by how it operates, it never, ever will.
Reply
 
# 4 : Friday 8-9-2017 @ 16:02
 
 
Also, 81% is not accurate.
Reply
 
# 5 : Saturday 9-9-2017 @ 12:53
 
 
Someone said :
You guys aren't reading it correctly.


The computer used images from a dating website.

first of, all of those images are filtered through the human eye and mind thinking "this photo makes me look good" "this is a photo of me dressed up" "this photo makes me look suitable for dating X person"

In short the users are choosing the photos to intentionally identify as, for example, a single gay man by their own definition and partially by a socially influenced definition aswell.

If anything all the computer has been able to do is pick up on the intended effect of the photo by the user who uploaded it ("grooming styles" and "gender a-typical features").

It says nothing about how, for example, a person behaves, looks or dresses, or "chooses photos", if they are in the closet, if they are choosing a photo for a CV, if they are just taking a family picture or a sightseeing shot.

My guess (and the logical thing to assume) is that same computer wouldn't have a snowman's chance in hell of being an effective "automatic outing machine" for people who are not intentionally feeding it "gay or straight photos" from a dating site, and, judging by how it operates, it never, ever will.

that data isn't free and nobody does anything in the US without somebody paying for it so who& why?
And the only logical thing to assume about AI is it constantly improves
The application of this technology ?
its facial recognition technology .
Reply
 
# 6 : Saturday 9-9-2017 @ 14:44
 
 
Wonder when the app will become available?
Reply
 
# 7 : Monday 11-9-2017 @ 02:50
 
 
Someone said :

And the only logical thing to assume about AI is it constantly improves
The application of this technology ?
its facial recognition technology .

Yeah but "improve" I mean you can improve a technology, but you can't improve a toaster so it will boil potatoes.

This technology is a pretty accurate "haircut recognizer"; all it will ever get better at being is a "haircut recognizer".
It's not going to ever be able to tell a person's epigenetics variants by photograph.
It's only ever going to get better at knowing the gay people that choose to identify themselves freely in the context of a perceived gender-social aesthetic, like real humans.
Reply
 
# 8 : Monday 11-9-2017 @ 05:56
 
 
Someone said :

Yeah but "improve" I mean you can improve a technology, but you can't improve a toaster so it will boil potatoes.

This technology is a pretty accurate "haircut recognizer"; all it will ever get better at being is a "haircut recognizer".
It's not going to ever be able to tell a person's epigenetics variants by photograph.
It's only ever going to get better at knowing the gay people that choose to identify themselves freely in the context of a perceived gender-social aesthetic, like real humans.

The parameters used were - gender-atypical” features, expressions and “grooming styles” on a sample of 35000 men and women .
If they take a larger sample ( and there are millions of photos of gay people online)there will be many more parameters .


why are they doing this and who is funding it? What's the eventual application? Nothing happens in the us without money ...
Reply
 
# 9 : Monday 11-9-2017 @ 10:43
 
 
Fro Hoffpost
Queer Groups Condemn Study Claiming Computers Can Tell If You’re Gay From Photos

http://www.huffingtonpost.com/entry/stanford-study-sexuality_us_5 etc ...
Reply
 
# 10 : Tuesday 12-9-2017 @ 02:56
 
 
I really don't believe that there are "gay" facial features. This sounds like a publicity stunt to me.
Reply
 
# 11 : Tuesday 12-9-2017 @ 03:54
 
 
Someone said :

The parameters used were - gender-atypical” features, expressions and “grooming styles” on a sample of 35000 men and women .
If they take a larger sample ( and there are millions of photos of gay people online)there will be many more parameters .

I'm sorry it's just, we're talking about parameters of a visual image of a human being.
Someone's homosexuality is simply not ascertainable by how a person looks once you remove social and psychological factors (grooming, expression and such) from the build of their face and body (their DNA, epigenetics and hormones).
This is a fact of... biology, and it can't be pierced just by upping computer power and widening sample pools.

You would probably have a better chance finding someone's orientation by a sample of their body odor, or a blood sample and even that's far fetched, like, it'd be a slightly more believable sci-fi movie but still almost as unlikely as warp speed travel. At least it has a biological avenue to cling too not "scanning patterns and trends in the pixels of an image of a person".
Science isn't even sure yet exactly HOW people are gay. You can even have identical twins one of whom is gay and one of whom is straight (I know some).

The results posted are a false positive because the computer algorhythm is ran like a "dreamscope, line finder" which is picking up on the visual themes of a pool of people subtly and instinctively BRANDING THEMSELVES according to their orientation (and you can bet you ass that a LOT of those images, considering its a dating site, won't even be doing this subtly).
This programme will not work on "regular" images. All you're going to be able to define is if someone has usually feminine eyes or a fancy hair cut, or a square jaw or a round jaw. Then what are you going to do? invent a "If you look genetically girly, your gay" machine?
There is no genetic link between looking girly, and being gay (and vice versa for all genders and orientations).
If you have a man with bright frigid, estrogen laces eyes with long legs and big hips, he could be straight as a arrow. You can have heavy set, deep eyed, colour blind bearded men with short little legs and a square jaw, totally homosexual.

It would at the very least need to be a ludicrously precise DNA test the likes of which the world has never ever seen before.


Reply
 
# 12 : Tuesday 12-9-2017 @ 04:57
 
 
In fact, I suspect as science advances, I reckon they'll discover something like the only way to accurately check someone's sexual orientation against their will would be with

1., a brain biopsy (which would either kill them, or need to be done post-mortem)

2., Both Parent's complete DNA,

3., Both Parent's true sexual orientations.


I suspect might never be a scientific way to know if someone is gay against their will, without these things, becasue DNA is different to gene expression. If "orientation" was decided by your DNA-as-written, Identical twins would have identical orientation.

When we're in the womb, we are a "spud" made up of clusters of cells, that don't know what the fuckin' story is with which bits of the DNA to use, and the clusters all just pick whatever, like each little village making up their own laws and stuff in a pre-telecommunications Ireland.
You might have a clump of lower spine that is usually "dad genes expressed" heavy or a left shoulder blade of "mam's genetics mostly expressed" But rather then come out like a lumpy, fucked up freak, the 'Spud' also gets washed in a coctail of stuff things that handle all kinds of "overall decisions" about what you're going to be; They'll try to force a gender, try to make physical symmetry, skin on the outside, Proper skeleton shape.

All the stuff that means you don't come destined to have your Mother's left leg length and your father's right leg, or one coloured tit and one freckly one.
But none of this change which "genes got expressed" in each of the different clumps of the 'spud' that became your body.
and that (in epigenetics) is where it is currently believed a person's orientation can be effected.

It also explains why identical twins can have opposite orientations.
Reply
 
# 13 : Tuesday 12-9-2017 @ 14:11
 
 
but wouldn't the alogarithm detect the expressive difference between the twins ? And with other key markers ?????


Reply
 
# 14 : Tuesday 12-9-2017 @ 22:46
 
 
Someone said :
but wouldn't the alogarithm detect the expressive difference between the twins ? And with other key markers ?????

what's your idea of 'other key markers'?

With regards expression., yes it's conceivable that if the 'gay one' was expressing himself freely in a free society where it is safe and fine to allow oneself to assimilate such expressional accents., A computer complex enough would be able to accurately guess at one's orientation in that case.

But it would never work on (for example) closeted people who would never allow themselves to adapt those expressional accents.
Reply
 
# 15 : Wednesday 13-9-2017 @ 05:27
 
 
Someone said :
what's your idea of 'other key markers'?

With regards expression., yes it's conceivable that if the 'gay one' was expressing himself freely in a free society where it is safe and fine to allow oneself to assimilate such expressional accents., A computer complex enough would be able to accurately guess at one's orientation in that case.

But it would never work on (for example) closeted people who would never allow themselves to adapt those expressional accents.

I don't understand what you mean by " assimilate such expressional accents"
And what expressions do you think gay people have and closeted people and what faces do they make in photos ?

When I refer to markers I mean they reference three but they are based on numerous measurements , some of these might be key?

I'm still at a loss though on the application and who might be funding this research ... any thoughts?

Btw I'd say the closeted might be most susceptible because having to work at passing at something they aren't they give off so many unconscious signals and the face with so many muscles really is the window of the soul
Reply
 
Prev 1Next