If you’ve ever uploaded a picture onto Facebook in the last two years, then you’ve probably been prompted to tag someone in the image based on an accurate, auto-populated list of possibilities; that’s Facebook’s facial recognition hard work. But after years of data-mining its user base, Facebook’s artificial intelligence lab has recently updated the underlying computer vision algorithm, enabling it to discern users even if their faces are concealed.
Instead, the algorithm analyses users’ posture, hair, body type, and fashion based the history of images they’ve uploaded, creating a distinct identity profile. While facial recognition technology is not exclusive to social media, finding it ways into security cameras, shopping centers, and even a church in Germany, Yaan LeCun, head of Facebook’s artificial intelligence division, wanted to the impact such technology can have on Facebook.
“There are a lot of cues we use. People have characteristic aspects, even if you look at them from the back,” LeCun says. “For example, you can recognize Mark Zuckerberg very easily, because he always wears a gray T-shirt.”
To test out the software, the team obtained 40,000 public pictures from Flickr — beware, users forfeit ownership when setting the images to public — and ran them through the algorithm. Some images displayed users’ faces in clear via, while others were partially concealed or fully concealed. The end result? Facebook’s machine vision recognized people’s identities with 83% accuracy.
Presumably, its accuracy will further increase once the software is set loose on Facebook’s vast image database, where it can absorb significantly more reference material.
For privacy advocates, such a prospect is truly frightening; taking into consideration that Facebook is possibly the world’s largest database of identifying information, and that information shared on platform becomes the company’s property data, it wouldn’t be too far of a stretch to assume that the data may sold to other third parties with equally capable facial recognition technology.
Facebook’s tools are not exclusive, but the underlying data fueling them is. Combine facial recognition with image geo-tags and smart phone-geotracking, and all smart phones become the ultimate surveillance tools. Hello Philip K. Dick’s Minority Report .
Conversely, LeCun does mention that the technology will eventually alert users whenever an image of themselves appears up on the Internet.
Source: Newscientist, Image: Siri Stafford/Getty
Advertisement
Learn more about Electronic Products Magazine
