A game developed at CMU by researchers to tag images has become popular in image search community. This game was presented at CHI conference by Ahn and Dabbish.
First Yahoo started using this philosophy and now Google has licensed this game (see here). After one of our discussions on image search, Michael Parekh sent this to me along with a link to an experience (report) on this.
Intriguing idea here. Yahoo used tags for a different environment — in Flickr search is useful but the emphasis is more on social networking. Using tags for primarily for searching a large collection is definitely intriguing. How will tags be combined with other information sources? Or will it be the tags only? How reliable will the tags generated from such games will be — more reliable than the current text oriented techniques used?
I am definitely eager to see how this experiment will affect image search by Google. Of particular interest is the fact that Google recently bought Neven Vision and now this. Does this mean that Google is serious about images? Are they thinking of combining human assistance (thru the ESP game) and computer algorithms?