In this first article of this series, I discuss face recognition technology in advertising as one current step on the path towards the era of ubiquitous computing.
I will then revisit a discussion of data security and encryption.
Faces of Big Data Tomorrow
"John Anderton, you could use a Guinness right about now!" a friendly voice hails the hero of Steven Spielberg's Minority Report. Many viewers laugh at the ad's cheerful tone in addressing a man who warily walks through a luxury shopping area in fear for his life.
Other ads in the same space identify Anderton as a Lexus owner who is on a “road less traveled,” an “American Express Member since 2037,” and ask him if he is stressed out and might need to “get away" and "forget [his] troubles”.
With cross-cuts the movie depicts a face recognition system whose sensors, cameras, and software apparently recognize John Anderton's face, identify his expression, track his gaze to specific ads, and deliver personalized messages.
Ironically--or perhaps of course?--the system’s real-time access to personal data through what one presumes is a ubiquitous computing grid inspires in Anderton a creeping paranoia instead of the soothing companionship the ads seem to offer.
Faces of Big Data Today
Currently contemporary site-specific advertising with face recognition senses the gender and age of a viewer, serves an ad, provides touch interaction, and records what you do on the touch pad and how long you look at the ad. This kind of system creates data that is basically anonymous.
One can expect this implementation of face recognition to become common in airports, commercial spaces, casinos, really anywhere ads go except perhaps roadside billboards. And it’s only a matter of time before webcam sensors become advanced enough to obtain face-related information on the viewer’s perusal of elements in web pages. Linked with account information, this kind of face recognition would be quite personal though it would also be elective; adjust settings on the webcam or disable it, as you like.
Facebook is the site where you would expect commercialized face recognition innovations eventually to show up. And indeed they have: Facebook Camera, for example, lets you manage the photos on your iPhone, helping you tag, group, and push them into your Timeline. Facebook's Instagram, like Twitter, lets you send mobile photos into specific areas of a shared social web space according to filters you specify. And through its acquisition of Face.com, Facebook takes a step further in using face recognition software to correlate tags with faces in photos not just on your own Timeline but on anyone's. Though users can disable the relevant feature—called 'photo tag suggest'—--it's active by default.
With these mobile photo technologies and its Timeline application, and after training its user with the intentional interactivity of the ‘Like’ button, the company seems to be working on the big data piece of personal ad serving in a way that even Amazon cannot match.
Just considered in terms of data quantity, even before acquiring Instagram and Face.com, and launching the Timeline and Facebook Camera features, Facebook was daily adding over 500 TB of compressed data. Though official numbers aren’t available, with the acquisitions and feature launches, we can safety assume that the company’s daily storage number is most certainly much higher. As the virtual space of unlimited expansion, Timeline holds all user content and prompts each uploading user to add various kinds of metadata for an event.
Any data the user enters into the Facebook Timeline has a shadow effect in the form of enabling applications running in the cloud to correlate the personal information with demographic categories (gender, income bracket, educational level, profession, etc.) that would make marketers squeal in their Guinness.
Monitoring Big Data Arrays
Facebook currently has thousands of servers handling the flow of their user created data. It would be site management suicide for them to use a single vendor or storage technology to store a data set that will easily scale into the zettabytes over the next 5-10 years. Pragmatically, even a company much smaller than Facebook would not use a single source for its storage array components; multiplying the number of vendors as part of a procurement strategy means having some insurance against vendor failures. Simultaneously, this pragmatic outlook also means finding a storage monitoring solution that unifies status views across different vendor systems.