(Originally published at Yahoo Style)
We’ve all got our reservations about artificial intelligence. From Facebook’s robots having to be shut down because they developed their own language, to the simply weird thought of them taking over our everyday lives, the rise of AI hasn’t been entirely successful.
And now, new research conducted by Stanford University has found something even more concerning.
A simple computer algorithm has been able to accurately predict a person’s sexuality a scarily high number of times. All it took to do this was a quick look at the profiles of 36,630 men and 38,593 women on various dating websites.
By analysing different faces, the algorithm was able to accurately predict sexuality 81 per cent of the time in men, and 74 per cent in women. While some people think they’re able to guess a person’s sexual preference just by taking a quick glance, the computer looked at subtle differences in facial structures, that humans are unable to pick up.
Aside the fact that this is a huge breach of privacy and poses a lot of ethical questions, the research has the potential to cause a lot of harm.
During the study, researchers got humans to predict sexuality alongside the computer. The results were startling – men were judged accurately only 61 per cent of the time, and women 54 per cent.
While this might seem like another study to find out just how capable artificial intelligence is, we should look at the wider context.
Facial recognition has already been in use for a while, from the likes of assisting the police with ASBOs to cross-referencing photo ID. We use some elements of it already in day-to-day life – like Facebook photo tagging or dating profile match ups, but it has the potential to be used far more dangerously. While the idea of paying for things with a selfie makes contactless cards pale in comparison, the dark side of facial recognition could land people in a lot of trouble around the world.
In countries where being homosexual is illegal, the rise of a computer ‘gaydar’ could put thousands of people at risk. And if you’re thinking that AI couldn’t 100 per cent accurately identify sexuality, we should take into account the fact that, after being shown five photos of men or women, the algorithm could accurately predict them 91 and 83 per cent respectively.
“Given that companies and governments are increasingly using computer vision algorithms to detect people’s intimate traits, our findings expose a threat to the privacy and safety of gay men and women,” said Michal Kosinski and Yilun Wang, the researchers behind the project.
With a few more year’s research behind AI, we could soon see a completely accurate ‘gaydar’ on the market.