Teaching Computers to Be More Like Humans

While search engines have come a long way at effectively locating what we’re looking for, they generally lack the intuition and subjectivity of a human mind.  For instance, a computer should have no problem finding scores of pictures of women, but distinguishing which ones are beautiful is a whole other thing.

In an attempt to give computers an aesthetic sensibility, Luis von Ahn, a professor at Carnegie Mellon has developed a technique for ‘teaching’ computers to be more like humans. Based on the same principles as a Captcha, those distored words you are asked to type that ward off spammers (a system also developed by von Ahn), the young computer science professor has developed a series of casual games that extract human subjectivity by pairing individuals together and asking them to classify certain images or sounds.  In the game Matchin’, pairs of images sourced from the internet are flashed on the screen and each players is asked to identify the picture the they think the other will find attractive – the more matches made, the more points earned.

This technique is referred to as “human computation,” the art of using massive groups of networked human minds to solve problems that computers cannot.  As more and more people play the games, the images displayed begin to carry with them the subjectivity of humans.  Theoretically, this process could be applied to every singe image on the web, enabling search engines to not just search for women, but beautiful women.

gwap

[via Wired]