It’s highly likely your friends could spot you in a crowd, even if your face were obscured or your back turned. Rather eerily, a recent report revealed that your friends aren’t the only ones who can pick you out in a sea of people—Facebook can, too.
In fact, Facebook’s algorithms can accurately identify an individual user from among 800 million pictures—in less than five seconds—98% of the time.
Using its artificial intelligence lab—a rather Orwellian acquisition for a social media platform—Facebook researcher Yann LeCun wanted to determine if the same algorithms already used for facial recognition could be adjusted to scrutinize other cues. LeCun, an expert in computer vision and pattern recognition who came to Facebook in 2013, explained that identifying cues—like body type or the way a person stands—could be taken into consideration by the technology the way humans already do.
“There are a lot of cues we use. People have characteristic aspects, even if you look at them from the back. For example, you can recognize Mark Zuckerberg very easily, because he always wears a gray T-shirt,” he explained.
From a pool of 40,000 public photos found on Flickr—some with faces fully visible and others partially or completely hidden—the tweaked algorithm successfully identified people 83% of the time, though the company says that figure has been as high as 98%. Facebook claims the system will help users of the new Moments feature, which automatically sorts uploaded photos by event—such as a graduation or birthday party—and tags the people who attended. LeCun believes the tool could eventually be a boon for the privacy-minded by alerting them whenever their picture appears anywhere online—even if their face is obscured.
But this has ignited a rather obvious controversy.
“If, even when you hide your face, you can be successfully linked to your identity, that will certainly concern people,” said Ralph Gross of Carnegie Mellon University. “Now is a time when it’s important to discuss these questions.” Though the algorithm is strikingly advanced, its very concept allows a rather gaping potential for abuse. When identifying someone in public is as simple as an automated algorithm, there is no opting out.
Facial recognition is yet another example of technology outpacing law. Talks held last week by U.S. government agencies to establish guidelines for the new technology failed miserably when privacy advocates stormed out in protest. At issue was the answer to a seemingly simple question: “If you are walking down the street, a public street, should a company be able to identify you without your permission?”
Alvaro Bedoya of the Georgetown University Law Center and other privacy advocates in attendance thought this was an easy no, but tech industry representatives didn’t share the sentiment. “We asked if we can agree on this edge case,” he explained, “but not a single company would support it. What facial recognition allows is a world without anonymity […] Companies are already marketing products that will let a stranger point a camera at you and identify you by name and by your dating profile. I think most reasonable people would find this appalling.”
It’s arguable that anyone who doesn’t lacks understanding of the technology—or doesn’t fully grasp the implications. The “if you haven’t done anything wrong then you have nothing to hide” aphorism only goes so far—especially when the choice to hide is no longer yours to make.
“You walk into a car dealership and the salesman knows your name and how much you make,” Bedoya warned. “That’s not a world I want to live in.”