Slashdot Mirror

← Back to Stories (view on slashdot.org)

Face Recognition — Clever Or Just Plain Creepy?

Posted by Soulskill on from the can't-it-be-both dept.

Simson writes "Beth Rosenberg and I published a fun story today about our experiences with the new face recognition that's built into both iPhoto '09 and Google's new Picasa system. The skinny: iPhoto is fun, Google is creepy. The real difference, we think, is that iPhoto runs on your system and has you name people with your 'friendly' names. Picasa, on the other hand, runs on Google's servers and has you identify everybody with their email addresses. Of course, email addresses are unique and can be cross-correlated between different users. And then, even more disturbing, after you've tagged all your friends and family, Google tries to get you to tag all of the strangers in your photos. Ick."

5 of 187 comments (clear)

  1. Slow news day? by gavron · · Score: 5, Insightful
    How did this make it onto /.? This isn't news, it's not new, and it's not technically on the cusp of anything except *yawn* sleep.

    When you choose to run your photos through facial recognition software (or give them to others who may do the same) you should expect .. ta da.. that they will run them through that software.

    The criteria for success includes Facial Identification (figuring out where the face is), Facial Recognition (figuring out if the face matches one on file), and some method of Facial Labeling ("tagging" that face with an identifier).

    Calling google "creepy" (pejorative nontechnical evaluation) doesn't give it the credit for doing all three parts correctly. Not liking that google's choice of identifier is more unique than "LAST, FIRST" or "FIRST LAST" is a personal foible, not a problem with the technology.

    Was this a slow "news" day?

    E

    1. Re:Slow news day? by mccalli · · Score: 5, Insightful

      "Calling google "creepy" (pejorative nontechnical evaluation) doesn't give it the credit for doing all three parts correctly. Not liking that google's choice of identifier is more unique than "LAST, FIRST" or "FIRST LAST" is a personal foible, not a problem with the technology."

      No, that's shortsighted. There are criteria used to evaluate success that aren't technical - that a superb technical job has been done to get an unwanted result is neither here no there, the result is still unwanted.

      That 'personal foible' - Google are perfectly capable of understanding unique ids, therefore they have chosen email for a reason. It's not too hard to extrapolate a scenario where they have location information and email addresses, and are therefore able to sell location-based marketing information about people who have been entered into their system without them even knowing. All it takes is one friend who doesn't realise the implications, or one business using services for free, and you're on whether you wanted to be or not.

      Of course, that's already the case the moment you've been entered into someone's online service-based address book. But combined with your image and location information...I find that disturbing. I don't know exactly why I do, but it's something that I feel disquiet about.

      Cheers,
      Ian

    2. Re:Slow news day? by Improv · · Score: 5, Insightful

      Unwanted to whom? Not all geeks care that much about privacy. There may be a loud portion that's always talking about PGP and privacy plugins for pidgin, including a (much smaller) contingent that hides away from cameras in real life and tries to obscure their features there. There's also a fair set of just-as-clued-and-geeky folk who are resigned to privacy being not worth the pain, as well as those who value radical openness and push for far less privacy than tradition has given humanity in the past.

      --
      For every problem, there is at least one solution that is simple, neat, and wrong.
  2. The fundamental difference by wickerprints · · Score: 5, Insightful

    ...which was pointed out in the article as well as the summary, but so far has failed to gain any notice in the comments, is that one implementation is purely local to the owner's physical machine, whereas the other is hosted on a corporate server, with no provision that the data of interest is solely under the author's control.

    That's the crux of the entire matter. Talking about unique identifiers or linking to other metadata is secondary. The real issue is that anything you submit to Google, Facebook, etc. is no longer really yours. The companies who host and mine this data have a vested interest in allaying such fears. They will say and do anything to give the appearance of trustworthiness. Whether they actually follow through is simultaneously independent and irrelevant, because the fact remains: once you put data online, or have it hosted remotely, someone else has it. Data is infinitely copyable, modifiable, crackable.

    When you use a program like iPhoto to tag images you took on a camera, nobody else has access to that information, provided you don't share or publish it in some manner. The recognition technology is programmed into the application, and the application runs locally. Google's service does not. The trend toward server-side computing to be alarming. The price of convenience and robustness is security and privacy. I am becoming increasingly convinced that the former is not worth the loss of the latter.

    (I do not have the latest version of iPhoto. And I'm not an Apple apologist by any means--for instance, I despise MobileMe for the exact same reasons I find Google's practices to be problematic. We live in a time when avoiding the harvesting of personal user data by powerful, ethically questionable governments and global corporations is virtually impossible, and it is getting more difficult by the day.)

  3. The reality by Anonymous Coward · · Score: 4, Insightful

    The title should be "Face Recognition - Works or is Bullshit" ?

    After working in computer vision for a few years, I learned how this stuff works, and I find it a lot of baloney. Sure, you can get something that works for a bit of the time for a restricted amount of data, but as we found out with the MIT debacle after 9/11, this stuff isn't robust. It can't handle simple changes in lighting, obstruction of portion of faces etc., which makes sense, since it isn't magic. Unfortunately it is a hot media topic, and computer vision researchers and others keep hammering away at it. I am here posting mainly to ask slashdotters to be more critical of the performance of such systems. Remember, when someone shows one of these systems, and this goes for other stuff like in-painting, tracking, edge detection and other standard computer vision problems, always ask yourself if the person presenting the solution has demonstrated it is a very wide variety of situations. You can always get something to work for one photo. (Usually of Lena...)