Can facial recognition technology distinguish one’s sexuality?
Many have warned and continue to do so about the privacy risks facing our privacy technology.
They once said that at some point the customs authorities would scan the faces of passengers on a flight, crossing them with criminal databases, something that is already happening!
If they are confirmed and what the gospel says about the future, then fears will evolve and the time when a program can discern the chances of one becoming a terrorist or developing a rare condition is not just far off.
Until that happens, the level of current face recognition technology could theoretically distinguish one’s sexuality, with better or worse results.
Researchers at Stanford University therefore claim to have used artificial intelligence to find signs of sexuality. In the study, published in The Economist, they trained a face recognition software on profiles of people they found on dating websites and then asked him to identify the sexuality of a random sample.
Researchers Michal Kosinski and Yilun Wang uploaded 35,326 photos of 14,776 people, categorized into four lists: homosexual women, homosexual men, heterosexual women, and heterosexual men. So when he was shown a photo of a man (out of sample always), the app could identify his sexual identity with accuracy of 81% (according to the man’s public profile on his media accounts).
When the target face photos were five, the program scored 91% success! Women found them a little darker, with 71% and 83% respectively in one of the five images. Needless to say, the program did much better than people in recognizing the sexuality of random people (61% for men and 54% for women).
Researchers have even used one of the current programs on the market, as has Facebook and Google, which have been using such services for a long time.
91% is a success rate that only causes terror for sensitive personal data. Machines are proving to be better than humans here and the future will be extremely exciting…