Skip to Content
Artificial intelligence

Facial Recognition Is Getting Incredibly Powerful—and Ever More Controversial

September 8, 2017

A camera trained at your face can reveal an awful lot about you. We’ve written a lot recently about how facial recognition software has reached an incredibly high standard in the last couple of years. So high, in fact, that the technology is robust enough to be used across China to authorize payments and catch trains.

But some recent technologically impressive advances are causing more of a stir. Earlier this week, Michal Kosinski and Yilun Wang from Stanford University posted a paper on the PsyArxiv preprint servers, soon to be published by the Journal of Personality and Social Psychology, which shows that facial recognition AI is more accurate than humans at detecting sexual orientation from pictures of people.

Trained on 130,741 images taken from dating sites where people explain their sexual orientation, the pair’s neural networks can distinguish between gay and heterosexual men in 81 percent of cases from just a single photo. That compares to an accuracy of 61 percent for a human. Given five images, the AI’s figures rises to 91 percent. (The same numbers for women are 71, 54, and 83, respectively.)

The approach is reminiscent of work described last year by researchers from Shanghai Jiao Tong University in China, who trained neural networks using photographs of known criminals and non-criminals. They were then able to correctly identify criminals from new images with an accuracy of 89.5 percent.

The Economist suggests that the same approach could be used to try to identify other qualities, such as IQ or political leaning. The phenomenon is, clearly, troubling to those who hold privacy dear—especially if the technology is used by authoritarian regimes where even a suggestion of homosexuality or criminal intent may be viewed harshly.

A natural conclusion: maybe we should obscure our faces while out on the streets?

Sadly, another new study may render that idea rather pointless. Late last week, Amarjot Singh from Cambridge University and his colleagues in India published a study on the arXiv showing that in many cases it is possible to identify a person using facial recognition even when their face is obscured. When people are wearing a hat, scarf, and glasses to cover their faces, their algorithms were able to identify people with around 55 percent accuracy. The figure rose to 69 percent when just glasses were removed.

Zeynep Tufekci from the Berkman Center for Internet and Society at Harvard University commented on this final finding on Twitter, described the research as part of an “ever-increasing new capability that will serve authoritarians well.” She likely has a point.


Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

What’s next for generative video

OpenAI's Sora has raised the bar for AI moviemaking. Here are four things to bear in mind as we wrap our heads around what's coming.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.