Facebook's Facial Recognition: Welcome to the Digital Plantation

The scale of what Facebook is doing is so spectacular that it is impressive to admire from a distance.

Courtesy: singularityhub.com

(SALT LAKE CITY ) - Facebook is Big News all of a sudden over their new system for using facial recognition software to automate the process of “tagging” user uploaded photos.

(See Reuters for the typical summary: http://www.reuters.com/article/2011/06/08/us-facebook-idUSTRE7570C220110608)

Every major news organization, every technology site, and apparently government regulators around the world, are all interested in this new technology.

At least the media is entirely missing the point.

What Facebook is revealing is one of the biggest cons in the history of the internet. Facebook is building the most advanced system for facial recognition in existence.

They are exploiting hundreds of millions of hours of their users' labor; along with all of their personal and private photographs. All without the knowledge or consent of any of the victims (*cough* users).

Here is how this works.

Facial recognition is a very difficult task (basically impossible) for computers. The processes by which humans recognize a face cannot be broken down into a series of logical steps.

That is not how the human mind works and the process cannot be duplicated through conventional machine programming. Facial recognition can be made to work, in essence, by cheating.

Instead of writing a program that tries to mimic the process by which humans recognize faces you instead have a machine analyze a very large number of pictures and develop mathematical profiles of the similarities between them. These similarities are things that a computer can recognize like shapes, colors, etc.

Once the computer has built up this huge data bank of pictures and their digital profiles, you tell the computer which parts of which images are faces, and who they are. Based on that “seed data,” you have the computer make guesses about where the faces are in other pictures and who they belong to.

Those guesses are then given to a human, who either verifies or rejects them. If the computer guesses wrong, the human tells the computer the correct answer. All of this human input is then fed back into the statistical data bank, creating an ever more subtle and expansive baseline of data upon which to make new guesses. This is a rudimentary machine learning algorithm, which is not particularly advanced. It is the kind of thing that an upper level Computer Science student could pull out of a textbook.

(See this research paper put out by Microsoft describing how they created the “Kinect” toy for their Xbox, which is based on the same approach: http://research.microsoft.com/pubs/145347/BodyPartRecognition.pdf)

Facebook's major breakthrough is that they are using the personal data (uploaded photos and existing name tags) and labor of their users to provide the "seed data" for this project. They plan to continue to exploit this free labor pool as they move into the second phase of the project, where they have their computers make guesses, and have their users verify and correct them.

According to numerous citations Facebook users “tag” around 100 million photos per day. If you estimate that this tagging process takes 30 seconds, that adds up to roughly 300 million man hours per year.

Even if you were paying child laborers in a sweatshop in China 10 cents an hour, that would be a lot of money.

Not to mention that there is no possible way to build a sweatshop big enough to house the labor force that would be required, let alone provide them with the millions of computers that they would need.

It's a good thing Facebook has the internet. A sweatshop where the users not only work for no pay, but provide their own equipment and working facilities.

The scale of what Facebook is doing is so spectacular that it is impressive to admire from a distance.

They control a free labor force of 500 million people and probably have an accumulated database that includes photographs of 90% of the population in most “advanced” countries, and maybe 60-70% of the population world wide. (very rough estimates)

Based on those numbers, I would predict that this experiment will be a terrific technological and business success. Facial recognition based on this kind of machine learning is theoretically very effective, but cost prohibitive because of the massive amount of human input required. Facebook has found the easiest solution to the cost problem, which is to simply use slave labor.

I expect that Facebook is selling this to potential investors in their IPO exactly the same way. It is a demonstration of the value (power) that exists in their ability to manipulate such a large population group, and how that power can be profitably exploited.

What about Facebook's users? The people who have and will be "donating" hundreds of millions of hours of their labor along with their personal private property to this project.

Welcome to the digital plantation. It's a brave new world.

===================================

Salem-News.com Business/Economy Reporter Ersun Warncke is a native Oregonian. He has a degree in Economics from Portland State University and studied Law at University of Oregon. At a young age, his career spans a wide variety of fields, from fast food, to union labor, to computer programming. He has published works concerning economics, business, government, and media on blogs for several years. He currently works as an independent software designer specializing in web based applications, open source software, and peer-to-peer (P2P) applications. Ersun describes his writing as being "in the language of the boardroom from the perspective of the shop floor." He adds that "he has no education in journalism other than reading Hunter S. Thompson." But along with life comes the real experience that indeed creates quality writers. Right now, every detail that can help the general public get ahead in life financially, is of paramount importance. You can write to Ersun at: warncke@comcast.net