Mechanical Turk 2.0
Google has recently released their answer to Amazon's Mechanical Turk.
If you don't remember the Mechanical Turk, it was Amazon's attempt to get people to provide solutions to problems that computers are bad at solving. This data can then be directly used to alter systems behavior, or used to train a larger neural-net.
But the new idea? Don't give workers money, learn from Tom Sawyer and let them know just how fun it is to whitewash that fence. The new Google Image Labeler
is designed to make labeling more like a game by pairing individuals together and concocting a scoring rubric. Get points and esteem from your peers by applying the same label to an image as your randomly selected partner.
Doesn't this remind you of the prizes you got as a kid in school? Do well, get a piece of candy; everyone in the room with see you eating it and know. There's been a lot of resistance to competition fueled learning in recent years, but I think Image Labeler reminds us: games and competition may or may not be conducive to learning, but it certainly gets people moving.