Here’s an interesting video that is quite long about how we can use human cycles, currently being wasted on billions of hours of playing solitaire, to perhaps make artificial intelligence smarter.
By turning visual and semantic relationships into a match-making game, we can categorize all the photos from Google Images in little time at all, if we were to employ the cognitive surplus out there in a way that people would find it fun to tag images.
The example the PhD uses is the ESP game, which shows two players the same image. They have no knowledge of what the other is guessing. However, if they both end up guessing the same word in the allotted time, they win points. Another game tests semantic relationships: if the narrator’s word is “milk” and he gives the other person clues like “is near to cereal” and “is a liquid” then hopefully the other person will guess it correctly. In these games, the links between words and images are verified by two different parties per test, aggregated over multiple tests.
Could it be possible to use crowdsourcing/the cognitive surplus to rate people accurately and fill out their biographical information in a trusted manner? What other ways can we use this to solve large messy problems using simple, fun games?