Computer Science Department
School of Computer Science, Carnegie Mellon University


Solving Hard AI Problems Using Computer Games

Luis von Ahn, M. Ian Graham, Laura Dabbish,
David Kitchin, Lenore Blum

November 2002

NOTE: This report is currently unavailable electronically.
Please contact the authors directly for access.


Keywords: Hard AI Problems, games, image classification

Automatically determining the contents of an image is a problem far from being solved by Arti cial Intelligence (AI) techniques. In this paper we introduce a simple game that is fun to play and has the following unique property: when humans play the game, they help computers determine the contents of images. We present a study providing evidence that the game is fun. If played by a large number of people, this game could provide a resource capable of classifying all images on the World Wide Web. This represents a novel form of interaction between humans and computers, a symbiosis in which humans playing the game are entertained and computers running the game obtain answers to problems they can't solve. Thus, we introduce a much more general idea: computer games can be used to solve hard AI problems.

9 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by