How a Blurry Cow Highlights Weaknesses in Google’s Face Recognition

A cow grazing on the banks of the River Cam in Cambridge, England, shows up in Google Street View with its face blurred for privacy.
It’s a harmless mistake, and one full of humor, especially given the amount of bovine puns that the Internet’s technorati bestowed upon it this week. But the fact that Google’s facial-recognition algorithms identified a cow — and probably many other non-human faces as well — is significant. It’s at once a lingering effect of the  that Street View faced in Europe a few years ago and a sign that Google’s artificial intelligence has a lot of room for improvement.

Great to see Google takes cow privacy seriously
— David Shariatmadari (@D_Shariatmadari)

The search giant has gone to great lengths to promote its ability to detect faces. Consumers are perhaps most aware of the technology thanks to the recently revamped , which automatically detects features in your photos grouped into People, Places and Things. The People search can organize faces from across all your images, and can even detect the same person across several years.
Using similar algorithms, the Things search can identify landmarks, even differentiating between famous landmarks and their identical copies. Did you visit the Statue of Liberty replica in Paris’s River Seine? Google knows. Snap a picture of the mini Eiffel Tower at Las Vegas’s Paris Hotel? Google will tell you.
Like all machine-learning algorithms, though, Google’s must be trained using as many examples as possible. The blurry cow on the River Cam is proof that even with an image database as massive as Street View’s to learn from, the algorithms still aren’t perfect. To help further their training, Google is  to third-party developers, who can harness the detection technology in their own websites and apps via an API.
As for the blurry cow itself, Google made light of the situation.
“We thought you were pulling the udder one when we herd the moos, but it’s clear that our automatic face-blurring technology has been a little overzealous,” a spokesperson told the . “Of course, we don’t begrudge this cow milking its five minutes of fame.”