Google's foray into face recognition raises privacy concerns
Newly acquired PittPatt software could add facial recognition elements to Google+, YouTube, image searches
Eric Schmidt and I have at least one thing in common: we both find facial recognition software creepy.
He went on to explain that Google had actually developed facial-recognition software as part of its Google Goggles product but withheld the technology because of privacy concerns.
Imagine being able to identify a stranger simply by photographing them with your smartphone: up would pop their name, age, social networking profile. From a technical perspective, with modern face-recognition algorithms and a large enough database of faces, it's entirely possible. It's also creepy.
That's something Schmidt and I agree on. So, then, given Schmidt's stance, I was somewhat surprised to learn that Google recently acquired PittPatt, a software company that specializes in — you guessed it — facial recognition software.
PittPatt's software, spun off from research done at Pittsburgh-based Carnegie Mellon University, is part of the growing computer-vision industry. According to the company, the software can "search images for faces, determine if faces are the same person, track faces in video sequences and pinpoint constituent landmarks in faces."
The software does both facial detection (simply identifying the presence of faces) and facial recognition (identifying individuals based on pattern matching). The latter is what has privacy advocates most worried.
So, what will Google do with PittPatt's software? In short: we don't know, yet.
Could be used in image search, YouTube, Google+
A spokesperson told me the company has "nothing to announce at this time." But it's not hard to imagine some likely applications.
For example, Google has a photo-management program called Picasa, which already uses facial recognition software to help organize images by person.
It's also not difficult to imagine facial-recognition software being rolled into a service like Google image search or YouTube. But it's also possible — and this, for me, is where things begin to cross the line into the realm of the creepy — that PittPatt's technology could be folded into Google+, the recently launched social network.
Google+ already has photo sharing and photo tagging functions that are very similar to ones used by the rival social networking site Facebook. If Google does add facial recognition to its photo-sharing functions, it certainly won't be the first site to do so.
Late last year, Facebook did exactly that and got itself into hot water with a feature called Tag Suggestions. It used facial recognition to automatically identify people in Facebook photos in order to make tagging them easier. The feature isn't currently available in Canada, but in parts of the world where it is available, it's been quite controversial.
Critics take issue with the fact that facial recognition is enabled by default, that the feature is opt-out (rather than opt-in) and that the settings are confusing. This past June, in the EU (which has strong data-protection laws), the Tag Suggestions feature was probed by regulators. If Google does incorporate facial recognition technology into its own social network, it would be wise to pay especially close attention to the Facebook example.
"We've said that we won't add face recognition to our apps or product features unless we have strong privacy protections in place, and that's still the case," a Google spokesperson said.
I have to wonder, though, might Google consider its much-touted "Circles" feature, which allows Google+ users to organize people into separate groups, as having "strong privacy protections"?
Are the granular sharing options built into Google+ enough to make web-scale facial recognition software any less intrusive? It's hard to tell.
For me, the key will be transparency. As creepy as I find facial recognition, I completely understand that others really like it. I understand that it can have practical (even fun) uses. And I have no problem with large companies automatically scanning photos and videos for faces — as long as users are aware that it's happening and have the opportunity to make informed choices.
I'll take the creepy face recognition I know over the creepy face recognition I don't any day.