A young man who got an email from Facebook ‘identifying’ him via Facebook tag in a series of photographs which turned out to be his mother as a young woman, says the incident “opens the door to larger and more difficult questions.”
A young man who got an email from Facebook ‘identifying’ him via Facebook tag in a series of photographs which turned out to be his mother as a young woman, says that the “oddly compelling” incident “opens the door to larger and more difficult questions,” according to a report in The Verge.
Specifically, the incident raises questions over what else Facebook’s algorithms can do.
Clearly in this case, they made an error, Fred Benenson, a data scientist at KickStarter, says, but the inadvertent ‘tagging’ shows off that the algorithm currently in use on Facebook to ‘tag’ photos can, in theory at least, trace people’s families via genetic traits translated into their faces.
“What about the cases where this algorithm isn’t used for fun photo tagging?” Benenson said to The Verge via email.
Facebook tag: What can this technology do?
“What if another false positive leads to someone being implicated for something they didn’t do? Facebook is a publicly traded company that uses petabytes of our personal data as their business model — data that we offer to them, but at what cost?”
NEC’s Neoface biometric software is already being used by police forces in the U.S. and the UK to identify people from video footage, as reported by We Live Security.
Facebook’s photo tagging is currently only used within the site, and is an option the user can control. The site has refused to say how they might use this data in future.
Facebook’s faceprints are already controversial. When Facebook extended the reach of its ‘faceprints’ so it could identify people via profile photos, as well as those they were tagged in, the ‘feature’ was banned in Europe.
Senator Al Franken said in a press release, “How many Faceprints does Facebook have? . “Presumably, this would lead to a significant expansion of Facebook’s faceprint database. It would also likely capture some of Facebook’s least active users—those who are visible in their public profile photo but are not tagged in any other photos. These people are often less active users who may not be aware of Facebook’s privacy changes. I urge Facebook to reconsider this change.”
Facebook has already extended the ‘reach’ of tagging, by allowing brands to reach into people’s news feeds by ‘tagging’ other brands or celebrities, according to Marketing Land, and thus reaching the news feeds of people who did not opt to follow them.
Benenson’s case shows off, The Verge says, the power of such algorithms to identify people by family affiliation, race, and even regardless of age: if someone has posted a picture on Facebook, the site will be able to identify them years later.
Facebook’s current face-matching algorithm is limited in scope, at least compared to an algorithm unvveiled as part of one of the networking giant’s AI research projects.
Deepface was one of these – and can match two previously unseen photos of the same face with 97.25% accuracy – humans can do the same with around 97.5% accuracy, a difference which TechCrunch describes as “pretty much on par”.
Deepface: The alarming ‘next step’
It’s a huge leap forward in the technology, which some see as having potentially alarming implications for privacy.
Although Deepface is a research project, and unrelated to the technology used on the site, it “closes the vast majority of the performance gap” with human beings according to the Facebook researchers behind it (PDF research paper here), and can recognize people regardless of the orientation of their face, lighting conditions and image quality.
Publications such as Stuff magazine describe the technology as “creepy”, saying that were it implemented “in the wild” it should make site users “think twice” about posting images such as “selfies.”
Deepface uses deep learning to leap ahead of current technology – an area of AI which uses networks of simulated brain cells to ‘recognize’ patterns in large datasets, according to MIT’s Technology Review.