Facebook’s head of artificial intelligence, Yann LeCun, told New Scientist that the experimental algorithm could correctly identify people with 83 percent accuracy when confronted with 40,000 public photos on Flickr, even when only some of the photos had the person’s face clearly showing.
The system works by looking for unique characteristics of each person, using attributes such as clothes, body shape, typical posing and even individual hair styles. LeCun told New Scientist: “People have characteristic aspects, even if you look at them from the back,” citing how Facebook CEO Mark Zuckerberg is identifiable “because he always wears a gray T-shirt.”
Although the algorithm is not yet publicly available on Facebook, it looks set to arrive in the near future and could be used for automatically tagging friends in photos. As interesting as it may be, the technology is likely to understandably concern many people though because of the implications of being identified even when your face is away from the camera.
Ralph Gross of the Carnegie Mellon University in Pennsylvania said: “If, even when you hide your face, you can be successfully linked to your identity, that will certainly concern people.”
The algorithm is powered by a complex neural network that replicates the way in which the human brain works. Such circuitry is becoming increasingly common-place as it becomes the way to progress machine learning technologies.
Its success is certainly impressive but, in a world that is increasingly conscious of personal privacy, it remains to be seen whether the public will be accepting of such a potentially intrusive technology. Although it could be useful for law enforcement and suspect tracking, people may be critical of the potential of yet-more surveillance when friends can still be tagged manually in photos anyway.