Artificial Intelligence researchers working on facial recognition systems now say their machine learning system can predict a person’s sexuality and even their political views just from measuring their face.
The research was published this week in the Nature journal Scientific Reports, was conducted by Stanford University’s Michal Kosinski.
The paper explains than the algorithm studies body language. “Head orientation and emotional expression stood out,” Kosinski writes.
“Liberals tended to face the camera more directly, were more likely to express surprise, and less likely to express disgust.”
But there were several other clues in people’s features – not just race and age but more subtle indicators – which also helped the facial recognition algorithm predict the way a subject might vote.
The photos had all other information, such as clothes and backgrounds removed, so researchers could be sure the AI was just working on facial clues alone.
His system is not yet perfect, he says. As yet, it’s only correct about 70% of the time. But that’s already a good deal better than humans, who tend to have no better than a 50-50 chance of guessing someone’s politics from their facial features alone.
Kosinski admits that this technology could easily be used for sinister purposes. “Don’t shoot the messenger,” he says. “In my work, I am warning against widely used facial recognition algorithms. Worryingly, those AI physiognomists are now being used to judge people’s intimate traits – scholars, policymakers, and citizens should take notice.”
He sounded a similar warning two years ago when his team developed an artificially-intelligent sexual preference predictor – a kind of AI "gaydar": “We were really disturbed by these results and spent much time considering whether they should be made public at all.
"We did not want to enable the very risks that we are warning against.”
He said that systems like the one he had developed could be used to invade a person’s privacy: “The ability to control when and to whom to reveal one’s sexual orientation is crucial not only for one’s well-being, but also for one’s safety.”
With facial recognitions systems already being used in UK shops to identify potential shoplifters and vandals, it’s clear that algorithms of the type that Kosinski and his team have developed are only going to become more widely used, despite his warning.
He writes that his team thought long and hard whether to release their research, but decided in the end that it was only a matter of time before someone else did: “We believe that further erosion of privacy is inevitable, and the safety of gay and other minorities hinges not on the right to privacy but on the enforcement of human rights, and tolerance of societies and governments.”
Source: Read Full Article