Saturday, April 23, 2011

Commoditization of Face Recognition Technology – Is there a dark cloud lurking behind?

I continue to be intrigued by the notion of a public camera looking at an individual and matching up his or her personal information for whatever purpose. So I wanted to do a little follow up on my last blog. Just like the Brazilian world cup soccer initiative, facial recognition algorithms have already been applied successfully to many public security applications. .

The question is what happens when it becomes part of everyday technology, commoditized into apps that are available universally to all.

Think of the movie Minority Report, the 2002 Steven Spielberg movie, where cameras captured and read Tom Cruise's face, and then customized ads for his character pop up. Immersive Labs, a New York startup, recently introduced its smart billboard technology that combines video analytics with multiple data sources such as Twitter or Foursquare information to select the most suitable ad to display to the consumer nearby. The software seemingly understands the geometry of faces enough to determine the gender and approximate age range of the face looking at the webcam of the billboard – not exactly the more prying technology shown in Minority Report, but certainly a step in that direction.

The advertising billboards of the future will promote a product by analyzing  the audience to display items that the viewer is more likely to buy. The concept is not new. Last year NEC Corp. demonstrated  its  interactive digital signage solution, now popular in Japan. A built-in camera captures an image of anyone looking at the signage. The system then compares it to more than 10,000 stored patterns to determine the gender and approximate age.  Finally it displays an image that is most likely to appeal to the audience.

Digital Signal Corp.  seemingly wants to move further in that direction using 3D long-range facial recognition. The Virginia based start up’s laser radar system can provide a range of biometric data which, among other things, can tell you whether someone appears to be particularly nervous or tense.

A recent article nicely summarized my emotions - the technology is both “exciting and creepy”. 

10% of digital signage by 2020 is expected to have facial identification technology to provide personalized advertisement.   The day may be close when you’ll walk into a restaurant, the host will greet you by name and suggest a menu that will appeal to your palate; or you may be greeted with a sign announcing the price and location of items you may be most interested in as you walk into a department store.

I don’t want to add an Orwellian twist to such free market innovations, but I certainly bristle at the thought of a prying billboard camera that scans  my face every time I stand in front of it and searches  through a database of my personal information to create an ad that I simply won’t be able to ignore. The nagging thought in my mind is not about the technology but about a scenario when the technology is available to the masses.  

A recent start-up Viewdle is bringing a cool little addition to your Android phone. You shoot your friends and family, it recognizes faces (after you teach it who's who), then automatically tags photos for Facebook. Perfect for Facebook users!
Your smartphones will now be able to do these simple face recognition calculations in real time, compare faces to images previously stored and identified. The company’s website features a video of five women walking towards your camera when labels pop up to identify them and post their Facebook comments in real time.

Google Goggles is an exciting little app  for object recognition.  What would be the implications of a possible mash-up facial recognition technology with Goggles?  There are some interesting use cases as reported here . You meet someone, say in a conference, that you can’t remember the name of. You simply take a quiet flash-free photo, upload it into your device and let an app find out any tags associated with the photo, essentially launching a search engine to find name and any other details available on the web. Really cool!

But what about that person in the bar shooting your photo incognito to search your profile in a web database with less honorable motives? What about that stalker who can secretly take a photo in the street and then pull up all your information tagged to your pictures in cyberspace?

I am sure a company like Google, with its demonstrated effort to do public good, intends to make sure there is no abuse of people’s privacy from the use of its tools. But things go wrong, as evidenced in Google’s settlement with FTC over released email contacts associated with the launch of Google Buzz.  In a similar vein, Pandora internet radio was served with a subpoena as part of a Federal grand jury investigation of personal data collection through its popular Android and Iphone app. Google also got into trouble for seemingly collecting private information with its Street View data collection.

Clearly leaks happen when it comes to personal information – with or without malicious intent.
Recent reports that iPhone and iPad are regularly recording the device position and storing it in a hidden unencrypted file will do little to assuage consumers’ concerns about their personal data being used or potentially abused.  A similar situation exists for Android phones as well. Technically, these are all opt-in, yet the story followed in WSJ indicates that the data contained a unique identifier tied to an individual’s phone (but not to the user’s name) and is not totally anonymous.

Don’t get me wrong. I am sure all the kinks will be ironed out and  all these possibilities will eventually be more about convenience and quality of life and less about surveillance or loss of personal data.  As I indicated in an earlier blog, technology after all has overwhelmingly been the biggest driver of human progress.  Despite the nuclear disasters in Three Mile Island (US -1979), Chernobyl (Ukraine - 1986) and Fukushima Daiichi (March 2011),  today over 440 nuclear power plants worldwide are running safely to generate 14% of global  electricity. The benefits of nuclear fission technology decisively out-weighs the pitfalls. 

So, I hope, will be the case with all the commercial recognition techniques. 

No comments:

Post a Comment