Computer scientists from Carnegie Mellon University are developing and using online games to train computers visual system according to this article from the Pittsburgh Post-Gazette.

An example game is Peekaboom, which is used online by teams of two players. The first one, designated as “Peek,” sees on his screen an image — initially empty — and a word that describes the image or one element of the image. The second one, named “Boom,” gradually reveals the image or gives hints to “Peek” until he correctly guesses the word associated to the image. A computer program can use this information to learn how to better categorize and identify objects visually. Essentially, humans are helping to train computer vision systems.

This reminds me of the method people use to break CAPTCHAs. A Completely Automated Public Turing Test to Tell Computers and Humans Apart is often used to protect web forms from automated use. To break a CAPTCHA, one simply needs to redirect the CAPTCHA image that the script comes across - to a human of course. The most common way to do this, is to ask a human to evaluate a CAPTCHA image before giving them access to free pornography. Clever, but extremely simply, and I would guess it is quite effective.

Why I blog this? I believe we will see many more ‘teaching systems’ where computer software ‘learns’ by working with humans. Why? Because humans already know how to learn. We can categorize incoming sensory information and manipulate it abstractly with symbols. And, we know instinctively how to teach learning when we raise children. I quite expect to one day raise a personal A.I. using human training and attention.

via FutureWire via Unmediated via SmartMobs