AI Can Detect Sexual Orientation Based On Person's Photo (cnbc.com) 350
ugen shares a report from CNBC: Artificial Intelligence (AI) can now accurately identify a person's sexual orientation by analyzing photos of their face, according to new research. The Stanford University study, which is set to be published in the Journal of Personality and Social Psychology and was first reported in The Economist, found that machines had a far superior "gaydar" when compared to humans. Slashdot reader randomlygeneratename adds: Researchers built classifiers trained on photos from dating websites to predict the sexual orientation of users. The best classifier used logistic regression over features extracted from a VGG-Face conv-net. The latter was done to prevent overfitting to background, non-facial information. Classical facial feature extraction also worked with a slight drop in accuracy. From multiple photos, they achieved an accuracy of 91% for men and 83% for women (and 81% / 71% for a single photo). Humans were only able to get 61% and 54%, respectively. One caveat is the paper mentions it only used Caucasian faces. The paper went on to discuss how this capability can be an invasion of privacy, and conjectured that other types of personal information might be detectable from photos. The source paper can be found here.
Just because you can doesn't mean you should... (Score:5, Insightful)
Re:Just because you can doesn't mean you should... (Score:5, Insightful)
Wait, is the Pacific Northwest homophobic? I have some friends that live up in Oregon and they're extremely gay and I never heard them complain about the region being particularly difficult for them. One just sent me a photo taken from his backyard of a mountain being consumed by fire. I'm pretty sure he had nothing to do with that, though, despite the fact that he's flaming.
(I used this joke with him, too, and he didn't seem to mind. He would have told me if it had offended him.)
Not Significant Accuracy (Score:5, Insightful)
If the AI were to simply assign scores of "Straight" to EVERYONE, it would achieve 90% accuracy for men and 85% accuracy for women, since about 90% of men are straight and about 85% of women are straight. So scores of 91% and 85% accuracy are not statistically significant.
Re:Just because you can doesn't mean you should... (Score:4, Insightful)
Massively Flawed (Score:2, Insightful)
First off, what they claim to have created is tantamount to computer-assisted phrenology -- long since debunked and tossed on the scrapheap of superstition.
The most obvious flaw appears here, starting on line 208:
*headdesk*
Leaving aside the gigantic issues presented by self-reporting and self-selecting samples, these idiots failed to account for a common practice among hetero women on dating sites, which is to falsely claim to be seeking other women as a means to reduce or eliminate an onslaught of tacky propositions from clueless het-boys.
Other glaring flaws include:
An actual sociologist could probably identify dozens of other flaws, any one of which would be fatal to the work.
I would undertake to create a similar piece of software that tries to identify criminals from photographs, and use police mugshots to train it. Surprise! Black people are more likely to be criminals! GIGO.
Frankly, I think they should have taken their theme from the closing paragraphs of their paper: "We created a digital phrenologist out of deep neural networks and other off-the-shelf parts that coughs up results that seem relevant and meaningful to the layman, when in fact they're utter garbage." That would have been a good paper.
Perhaps we can indeed learn new things by letting a DNN stare at human faces. But IMHO this paper is utterly valueless in identifying what those might be. GIGO.
Re: Just because you can doesn't mean you should.. (Score:1, Insightful)
It was only 60 years ago that being homosexual was outlawed in most "civilized western nations" and "treated" with drugs and electroshock. Society doesn't have a direction of continued permissiveness and can slip back at any time. Imagine what this ability would then be used for.
Re:Just because you can doesn't mean you should... (Score:2, Insightful)
It won't have this affect on government regimes that use this for population control. It won't have this affect on algorithms trained with this inherent bias in them.
Re:Just because you can doesn't mean you should... (Score:4, Insightful)
Re:Just because you can doesn't mean you should... (Score:2, Insightful)
What is the need to distinguish gay from straight? It's a false categorization that has limited use in academic studies. It has no use in any social context that respects freedom of the individual. Even asking the question implies some need to put people into slots that we can then later act on. And no, I don't believe that the first question you ask a potential life partner is their sexual orientation. Human behavior is far more subtle and varied.
My comment also reflects the fact that ML/AI algorithms may have embedded biases in them that are overlooked because their internals are not very well understood. You are taking a very complex human issue and trying to reduce it to a single number. People without the will to think critically (80% of humans?) won't prevent this bias from propagating regardless of their own intentions.
Re:Massively Flawed (Score:5, Insightful)
these idiots failed to account for a common practice among hetero women on dating sites, which is to falsely claim to be seeking other women as a means to reduce or eliminate an onslaught of tacky propositions from clueless het-boys.
So how exactly does that gambit work for hetero women seeking men? Is this a thing that clued-in men know about? Some secret signaling that says "my profile says woman seeking woman, but I really want guys?" Do they not get an even tackier group of responses from bros hoping they'll hit a jackpot with a lesbian with a secret yen for yang and possibly a FFM threesome? What about fending off the lesbians who take it seriously?
And then there's the whole potential for lack of response, eliminating non-gross men who think cruising the women seeking women section is tacky and a waste of effort.
I mean, I'm genuinely curious here, if this is really a thing.
Re:Just because you can doesn't mean you should... (Score:5, Insightful)
So in terms of 'homophobia', do I irrationally fear any of them? No. Do I hate any of them? No. But it's obvious at least half of them are screwed up.
Knowing that people hate you for your sexuality and getting abuse for it from strangers and (depending on your background) former friends and family? Or alternately that they'd hate you if you were truthful about an unchangeable and fundamental part of yourself that you have to keep covered up every day of your life? (#)
Hmm, yeah. I guess that sort of thing might screw some people up.
Oddly, this would suggest that the actual issue is how homosexuals have traditionally been treated...
And that leads me to the conclusion that it's best not to encourage or approve of homosexual conduct, because it's self-destructive behavior.
...making people like you the problem, not homosexuality itself.
(#) Both of which were the case in most Western societies until recently, and *still* aren't as bad as the hostility homosexuals in many countries continue to face today- e.g. fear of being tortured or killed.
Re:Massively Flawed (Score:5, Insightful)
So how exactly does that gambit work for hetero women seeking men? Is this a thing that clued-in men know about? Some secret signaling that says "my profile says woman seeking woman, but I really want guys?"
Dating websites generally require you to fill out a profile before you're allowed to approach other members. Women seeking men who fill out women seeking women on the website are among the 80% of women chasing 20% of the men. Those 20% know the deal, because they get approached on an hourly basis, and every single woman disclaims her orientation tag in her approach. Those women don't want to be approached at all. They want to do the approaching.