The famo AI gaydar study was repeated – and, no, n't tell if you're straight or not jt om your face • The Register

Contents:

• 'I WAS SHOCKED WAS SO EASY': ​MEET THE PROFSOR WHO SAYS FACIAL REGNN ​​N TELL IF YOU'RE GAY
• GAY MACHE LEARNG – DO EXIST?
• THIS PSYCHOLOGIST’S “GAYDAR” REARCH MAK UNFORTABLE. THAT’S THE POT.
• THE FAMO AI GAYDAR STUDY WAS REPEATED – AND, NO, N'T TELL IF YOU'RE STRAIGHT OR NOT JT OM YOUR FACE
• CAN THE MACHE TELL IF YOU ARE PSYCHOTIC OR GAY?
• WHY STANFORD REARCHERS TRIED TO CREATE A ‘GAYDAR’ MACHE
• A STANFORD SCIENTIST SAYS HE BUILT A GAYDAR G “THE LAMT” AI TO PROVE A POT
• EXPLORG WEIBO ERS’ ATTUS TOWARD LBIANS AND GAYS MALAND CHA: A NATURAL LANGUAGE PROCSG AND MACHE LEARNG APPROACH☆

'I WAS SHOCKED WAS SO EASY': ​MEET THE PROFSOR WHO SAYS FACIAL REGNN ​​N TELL IF YOU'RE GAY

Weeks after his trip to Mosw, Kosski published a ntroversial paper which he showed how face-analysg algorhms uld distguish between photographs of gay and straight people.

”In a paper published last year, Kosski and a Stanford puter scientist, Yilun Wang, reported that a mache-learng system was able to distguish between photos of gay and straight people wh a high gree of accuracy. Prented wh two pictur – one of a gay person, the other straight – the algorhm was traed to distguish the two 81% of s volvg imag of men and 74% of photographs of women. “I was jt shocked to disver that is so easy for an algorhm to distguish between gay and straight people, ” Kosski tells me.

GAY MACHE LEARNG – DO EXIST?

’ Photograph: Jason Henry/The GuardianNeher did many other people, and there was an immediate backlash when the rearch – dubbed “AI gaydar” – was previewed the Enomist magaze.

) There was also anger that Kosski had nducted rearch on a technology that uld be ed to persecute gay people untri such as Iran and Sdi Arabia, where homosexualy is punishable by ath. His fdgs are nsistent wh the prenatal hormone theory of sexual orientatn, he says, which argu that the levels of androgens foet are exposed to the womb help terme whether people are straight or gay.

“Th, ” he wr his paper, “gay men are predicted to have smaller jaws and chs, slimmer eyebrows, longer nos and larger foreheads... (A uple of days later, Kosski tells me he has checked his slis; fact, he says, he didn’t tell the Rsians about his “AI gaydar”. This blog post explor the possibily of gay mache learng – that is, artificial telligence that is specifilly signed to intify and rpond to queer people and issu.

THIS PSYCHOLOGIST’S “GAYDAR” REARCH MAK UNFORTABLE. THAT’S THE POT.

However, there are a few key rearch areas that are likely to fall unr the umbrella of gay mache learng, cludg natural language procsg (NLP) mols for analyzg text data related to LGBTQ topics, predictive molg for unrstandg the health needs of LGBTQ people, and remendatn systems for fdgLGBTQ-iendly bs and servic. The aim of the study was to see if mache learng uld be ed to intify gay dividuals based on facial featur alone. The rults of the study were mixed, wh the algorhm rrectly intifyg gay men 81% of the time and lbian women 74% of the time.

Not fac taken om a datg webse), the accuracy rat dropped signifintly, wh the algorhm only rrectly intifyg gay men 54% of the time and lbian women 38% of the time. There is a lot of bate surroundg the existence of gay mache learng, wh some people argug that don’t exist and others assertg that do. This is bee gay mache learng allows for data to be traed on a wir range of people, which n ultimately lead to more accurate predictns and better rults.

Gay mache learng n therefore help bs and anizatns to make better cisns based on data, which n ultimately lead to improved out.

THE FAMO AI GAYDAR STUDY WAS REPEATED – AND, NO, N'T TELL IF YOU'RE STRAIGHT OR NOT JT OM YOUR FACE

There are not many exampl of openly gay people the world, which mak difficult to tra a mache learng algorhm on this group of people. Another challenge is that sexual orientatn is a fluid ncept, which means that n be difficult to fe what means to be “gay. There is a lot of discrimatn agast openly gay people many parts of the world, which n make difficult for them to participate mache learng studi or to be open about their sexual orientatn when participatg such studi.

CAN THE MACHE TELL IF YOU ARE PSYCHOTIC OR GAY?

As rearch ntu to explore the possibili of gay mache learng, we are likely to see even more amazg and novative for this technology.

This is particularly worrisome the se of gay people, as there is a risk that gay people uld be exclud om the benefs of mache learng simply bee they are not part of the domant group. About 7000 people participated the march, of sexual mori of the LGBT (lbian, gay, bisexual, and transgenr) muny and their supporters parad through the streets of downtown Tokyo to promote a society ee of prejudice and discrimatn.

And where data do go far enough to try and unrstand our muny when led by mercial imperativ, will pat a particular – mostly gay, while, male – picture:. When choosg between a pair of photos, the rultg program accurately intified a gay man 81 percent of the time and a gay woman 71 percent of the time. More ntroversially, Kosski and Wang’s paper claimed that the program based s cisn on differenc facial stcture; that gay men’s fac were more feme and lbian women’s fac were more mascule.

WHY STANFORD REARCHERS TRIED TO CREATE A ‘GAYDAR’ MACHE

A smart person wh a puter and accs to the ter n judge sexual orientatn of anyone the world, or lns of people simultaneoly wh very ltle effort, which mak liv of homophob and opprsive regim jt a ty b more easy. Unsurprisgly, that origal work kicked up a massive fs at the time, wh many skeptil that puters, which have zero knowledge or unrstandg of somethg as plex as sexualy, uld really predict whether someone was gay or straight om their fizzog.

The Stanford eggheads behd that first rearch – Yilun Wang, a graduate stunt, and Michal Kosski, an associate profsor – even claimed that not only uld nral works ss out a person’s sexual orientatn, algorhms had an even better gaydar than humans.

A STANFORD SCIENTIST SAYS HE BUILT A GAYDAR G “THE LAMT” AI TO PROVE A POT

“The paper propos replitg the origal 'gay fac' study a way that addrs ncerns about social factors fluencg the classifier.

In some untri, homosexualy is illegal, so the technology uld endanger people’s liv if ed by thori to "out" and ta spected gay folk. “Moreover, this entire le of thought is premised on the ia that there is value to be gaed workg out why 'gay face' classifiers might work – value further scribg, fg and settg out the methodology for any tpot dictator or bigot wh a puter who might want to opprs queer people. When the mache had to intify homosexualy om a randomly selected pair of imag featurg eher a homosexual or heterosexual dividual, the AI was on pot 81 percent of the time for men and 71 percent for women.

"Media headl that claim AI n tell if someone is gay by lookg one photo of your face are factually accurate, " wrote Drew Anrson, Director of News and Rapid Rponse for GLADD.

EXPLORG WEIBO ERS’ ATTUS TOWARD LBIANS AND GAYS MALAND CHA: A NATURAL LANGUAGE PROCSG AND MACHE LEARNG APPROACH☆

What their technology n regnize is a pattern that found a small subset of out whe gay and lbian people on datg s who look siar. Stanford's study is based on a key assumptn of an unvetted theory that explas homosexualy lled the prenatal hormone theory (PHT) of sexual orientatn.