Current Events > New AI can guess whether you're gay or straight from a photograph

(+) Yeah! (+)

https://www.theguardian.com/technology/2017/sep/07/new-artificial-intelligence-can-tell-whether-youre-gay-or-straight-from-a-photograph

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.

The paper suggested that the findings provide “strong support” for the theory that sexual orientation stems from exposure to certain hormones before birth, meaning people are born gay and being queer is not a choice. The machine’s lower success rate for women also could support the notion that female sexual orientation is more fluid.
---

I can do this too. Do you have a penis in your mouth? Congrats, gayboy.
---

While the findings have clear limits when it comes to gender and sexuality – people of color were not included in the study, and there was no consideration of transgender or bisexual people – the implications for artificial intelligence (AI) are vast and alarming. With billions of facial images of people stored on social media sites and in government databases, the researchers suggested that public data could be used to detect people’s sexual orientation without their consent.

It’s easy to imagine spouses using the technology on partners they suspect are closeted, or teenagers using the algorithm on themselves or their peers. More frighteningly, governments that continue to prosecute LGBT people could hypothetically use the technology to out and target populations. That means building this kind of software and publicizing it is itself controversial given concerns that it could encourage harmful applications.
---

TimePharaoh posted...

I can do this too. Do you have a penis in your mouth? Congrats, gayboy.

Are you trying to be funny or something?
---

Nah, he's just tired from getting out of his rallies

It checks out, it told me both TC and this thread came up positive.
---

MabusIncarnate posted...

It checks out, it told me both TC and this thread came up positive.

I am openly gay. Your point?
---

His point is it works obviously dipcake?
---

CountessRolab posted...

MabusIncarnate posted...

It checks out, it told me both TC and this thread came up positive.

I am openly gay. Your point?

I'm just verifying the validity of your post. Don't act like i'm trying to be insulting or anything, jeez.
---

topic title should be: new ai can guess 74% of the time correctly in a tiny sample size...
---

TimePharaoh posted...

His point is it works obviously dipcake?

Grow up.
---

CountessRolab posted...

TimePharaoh posted...

His point is it works obviously dipcake?

Grow up.

My technology can detect CEmen with the maturity level of a 7 year old.
---

What the
---

MabusIncarnate posted...

CountessRolab posted...

TimePharaoh posted...

His point is it works obviously dipcake?

Grow up.

My technology can detect CEmen with the maturity level of a 7 year old.

---

this sounds like a fun game
---

CountessRolab posted...

TimePharaoh posted...

I can do this too. Do you have a penis in your mouth? Congrats, gayboy.

Are you trying to be funny or something?

Well he made a more reliable point than the study did.

Gaydar is such a myth it's a joke. There's so many more gay men who are indistinguishable from typical straight men that the only real way to tell if they're homosexual outside of lgbt hangouts is seeing them sexually ACT on it, not their appearance.

If someone's proud of how "accurate" they think their gaydar is just because they can look at the fit pretty boy with coiffed hair wearing tight shorts winking at other guys and guess he's gay....congratulations, ANYONE can do that. Now try picking out the other dozen gay guys in the room who act and look no different than typical straight men.

You can't without the obvious cheats that anyone else could point out.
Because gaydar is bullshit. This study is no different.
---

Neat stuff

Hopefully it's never abused

Did they post specificity and sensitivity?
---

ClockworkHare posted...

Gaydar is such a myth it's a joke.

lol? just because you are incapable of it doesn't mean anyone else is
---

Guerrilla Soldier posted...

topic title should be: new ai can guess 74% of the time correctly in a tiny sample size...

35000 is not that tiny

the more worthwhile thing to say about the sample is the one they already mentioned:

people of color were not included in the study, and there was no consideration of transgender or bisexual people –

---

ClockworkHare posted...

CountessRolab posted...

TimePharaoh posted...

I can do this too. Do you have a penis in your mouth? Congrats, gayboy.

Are you trying to be funny or something?

Well he made a more reliable point than the study did.

Gaydar is such a myth it's a joke. There's so many more gay men who are indistinguishable from typical straight men that the only real way to tell if they're homosexual outside of lgbt hangouts is seeing them sexually ACT on it, not their appearance.

If someone's proud of how "accurate" they think their gaydar is just because they can look at the fit pretty boy with coiffed hair wearing tight shorts winking at other guys and guess he's gay....congratulations, ANYONE can do that. Now try picking out the other dozen gay guys in the room who act and look no different than typical straight men.

You can't without the obvious cheats that anyone else could point out.
Because gaydar is bullshit. This study is no different.

It's not I knew an American Idol contestant was gay and he came out later. And he's not even that flamboyant stereotype.

CountessRolab posted...

The data also identified certain trends, including that gay men had narrower jaws, longer noses and larger foreheads than straight men,

Well shit, looks like I'm gay. Better go suck a dick.
---

It's true, their definitely is learning biased around the dataset which is the problem. Big data problems require big data. 35,000 is not big data.

The machine’s lower success rate for women also could support the notion that female sexual orientation is more fluid.

Most important thing to take from the study though, imo.
---

MutantJohn posted...

It's true, their definitely is learning biased around the dataset which is the problem. Big data problems require big data. 35,000 is not big data.

The machine’s lower success rate for women also could support the notion that female sexual orientation is more fluid.

Most important thing to take from the study though, imo.

It sounds like a sufficiently powered sample size, but again a specificity of 80% (with no mention of false positives) makes this study a fun little story and nothing more.
---

MrPeppers posted...

MutantJohn posted...

It's true, their definitely is learning biased around the dataset which is the problem. Big data problems require big data. 35,000 is not big data.

The machine’s lower success rate for women also could support the notion that female sexual orientation is more fluid.

Most important thing to take from the study though, imo.

It sounds like a sufficiently powered sample size, but again a specificity of 80% (with no mention of false positives) makes this study a fun little story and nothing more.

I'm considering the entire possible data size, which is like 7.something billion people. 35,000 is on the order of like 1e5, compared to the population which is 1e9. So basically, it's like 1/10,000th the population.
---

people of color were not included in the study, and there was no consideration of transgender or bisexual people

If you have to take out that many variables to get these kind of results, it makes the whole thing seem kind of worthless.
---

guys stop

stop and realize what happened

WE JUST INVENTED A REAL GAYDAR
---

MutantJohn posted...

MrPeppers posted...

MutantJohn posted...

It's true, their definitely is learning biased around the dataset which is the problem. Big data problems require big data. 35,000 is not big data.

The machine’s lower success rate for women also could support the notion that female sexual orientation is more fluid.

Most important thing to take from the study though, imo.

It sounds like a sufficiently powered sample size, but again a specificity of 80% (with no mention of false positives) makes this study a fun little story and nothing more.

I'm considering the entire possible data size, which is like 7.something billion people. 35,000 is on the order of like 1e5, compared to the population which is 1e9. So basically, it's like 1/10,000th the population.

Right but the point of a sufficiently powered study is that a sample size is large and randomized enough to be representative of the sampled population. Even a number that low relative to the global population should be sufficient enough to reliably extrapolate data erroneously not sampled. Idk I've been drinking.
The
---

BLAKUboy posted...

people of color were not included in the study, and there was no consideration of transgender or bisexual people

If you have to take out that many variables to get these kind of results, it makes the whole thing seem kind of worthless.

POC do have different features from white people so it makes sense to do the races separately.

It's bisexual that throws this whole thing off. How do they account for that?

But if a human prejudges based on context clues it's "being a judgmental asshole".
---

So a university came up with an algorithm with no practical application other than a "lol" fun machine in the vestibule of a Walmart?
---

CountessRolab posted...

The research found that gay men and women tended to have “gender-atypical” features, expressions and “grooming styles”

well no shit
---

lowtides posted...

BLAKUboy posted...

people of color were not included in the study, and there was no consideration of transgender or bisexual people

If you have to take out that many variables to get these kind of results, it makes the whole thing seem kind of worthless.

POC do have different features from white people so it makes sense to do the races separately.

It's bisexual that throws this whole thing off. How do they account for that?

They subtract people with bisexual faces, clearly.
---

MrPeppers posted...

MutantJohn posted...

MrPeppers posted...

MutantJohn posted...

It's true, their definitely is learning biased around the dataset which is the problem. Big data problems require big data. 35,000 is not big data.

The machine’s lower success rate for women also could support the notion that female sexual orientation is more fluid.

Most important thing to take from the study though, imo.

It sounds like a sufficiently powered sample size, but again a specificity of 80% (with no mention of false positives) makes this study a fun little story and nothing more.

I'm considering the entire possible data size, which is like 7.something billion people. 35,000 is on the order of like 1e5, compared to the population which is 1e9. So basically, it's like 1/10,000th the population.

Right but the point of a sufficiently powered study is that a sample size is large and randomized enough to be representative of the sampled population. Even a number that low relative to the global population should be sufficient enough to reliably extrapolate data erroneously not sampled. Idk I've been drinking.
The

Extrapolations are exactly that, extrapolations. They only go so far. The most accurate thing to do is sample the entire population size.
---

MutantJohn posted...

Extrapolations are exactly that, extrapolations. They only go so far. The most accurate thing to do is sample the entire population size.

google "data saturation"

the thing that makes this study tenuous is the absence of people of colour, trans people, and other sexual orientations. not the sample size
---

Nah, that's dumb as hell. This reasoning is pretty absolute. You can make gross approximations about human nature in any meaningful and discernable way.
---

CountessRolab posted...

TimePharaoh posted...

----

Are you trying to be funny or something?

idk about him, but I thought it was hilarious. Thanks for quoting him because I would have otherwise missed it since he got modded.
---

This just confirms that you can also gauge if someone is an asshole just from seeing their face.

Topic: So can I.