Google executive chairman Eric Schmidt said it’s “unlikely” that Google would create a facial recognition service because it would make people uncomfortable, in the same week the company filed a patent for recognizing celebrity faces.
Schmidt’s comments were in response to a question about a phrase he coined: “crossing the creepy line.” He reportedly said that the “surprising accuracy” that facial recognition systems have attained is “very concerning.”
Despite this, Google filed a patent this week for “Automatically Mining Person Models of Celebrities for Visual Search Applications.” The patented process would use media reports that include pictures of famous people to automatically generate the facial data needed to identify new pictures of that individual.
Creepy? It depends on who’s a “celebrity.” The patent defines that as anyone appearing in “one or more articles.”
Google is in a particularly good place to launch a general facial recognition service if it wants to. The company’s free Picasa photo editor has a feature that automatically scans users’ photos and picks out the faces in them, grouping them into batches it guesses are the same person. It invites users to name those individuals.
It could be possible to use that data — the “person model” mentioned in the patent — to identify new pictures of the same person online, or snapped on a camera phone.
Facial recognition would also slot neatly into Google Goggles, a mobile app that identifies whatever is in front of the camera. Pointing the smartphone at the Eiffel Tower, for example, might bring up information about the structure. Google reportedly didn’t give Goggles the ability to ID people for fear of crossing “the creepy line.”
Another company in a position to build a facial database is the aptly named Facebook. The social network has access to millions of pictures of people, many of them containing faces helpfully identified by users. Facebook could launch an option to automatically tag new pictures, for example. If people had to opt-in to be auto-tagged, there might not even be much objection.
The question of universal facial search is quickly becoming more an issue of social norms than technical hurdles. It’s still hard enough that the companies with the resources to do it are worried about alienating users and more maverick entities can’t yet pull it off.
But as the technology inevitably trickles down, someone with less to lose and everything to gain will almost have to give it a shot. Schmidt acknowledged as much, saying, “some company, by the way, is going to cross that line.”