Up to now, machine-driven image recognition hasn't just been a science-fiction artifact brought to life, it's been a practical upgrade. Programs that can identify objects and even people help us learn about the world around us — or at least save time when labeling photos.
But Facebook's bout of bad publicity over a facial-recognition feature called Suggested Tags suggests how we can wind up on the other end of the image sensor. In the nightmarish future some critics of the social network now evoke, our mug shots become yet another searchable, indexable item on the Internet.
An op-ed on the Guardian newspaper's web site argued that Suggested Tags puts us well on the primrose path to the zero-privacy future of the film Minority Report, with "corporations bombarding passersby with holographic advertisements which crawled up the sides of walls, addressing them by name."
A post on PC World's site called this feature "downright creepy": "imagine, a world in which someone can simply take a photo of you on the street, in a crowd, or with a telephoto lens, and discover everything about you on the internet."
Ugh. But let's rewind first.
Back in December, Facebookannounced in a blog post that it would suggest tags for new photos. If it saw a Facebook friend in a picture whom you or others had tagged in earlier shots, it wouldsuggest that person's nameand invite you to keep, correct or drop that label.
Suggested Tags works like the facial-recognition options that have been in photo software since at least 2005 (when a since-shuttered Bay Area startup named Ojos launched one such application) and now constitute major features in Google's Picasa, Apple's iPhoto and Microsoft's Windows Live Photo Gallery. In all of these systems, once you identify people in enough photos, the software will take it from there and try to find them in other pictures.
Facebook has the advantage of a far larger set of tagged photos with which to start. It has the disadvantage of a history of privacy missteps.
This time, however, Facebook doesn't warrant the hand-wringing. Suggested Tags only work among friends (spokeswoman Jaime Schopflin e-mailed Monday that the site only suggests tags for closer friends, not every casual friends-list acquaintance), require human action to take effect and aren't shared off the site.
She added that Facebook's technology can't find one face out of the "billions" of portraits on the site: "It can only pick one face out of a hundred."
The true photo-privacy risk on Facebook remains unchanged: the demonstrated willingness of friends to tag us in possibly unflattering shots, a move we can only undo after the fact. You've edited your privacy settings to limit who sees photos tagged of you, right?
(The process to opt out of Suggested Tags could be simpler, though: In the Account menu at the top right of Facebook, select Privacy Settings; on that settings page, click the Customize Settings button; on that second settings page, click the Edit Settings button to the right of "Suggest photos of me to friends"; in the small window that appears, pick "Disabled" from the menu.)
But what about the future of image recognition? As it becomes increasingly available — on Tuesday, Google added a "search by image" option to its Chrome browser — some companies will be tempted to misuse it.
And they won't need to know your name to put you in potentially creepy scenarios. Would you be bothered by an ad that merely matched ads to your perceived gender and age? That was demonstrated in Tokyo last year.