Sign In

Communications of the ACM

ACM TechNews

Why Artificial Intelligence Researchers Love 'super Mario Bros.'

A scene from Super Mario Bros.

Artificial intelligence researchers like using the video game "Super Mario Bros." to test their work.

Credit: Georgia Institute of Technology

Artificial intelligence (AI) researchers find the video game "Super Mario Bros." especially amenable to testing their work, which includes AI that can learn the subtleties of game play well enough to beat the game and intelligently construct its own levels.

Games are ideal AI testbeds because they share with researchers the goals of developing systems that can acquire the same logic, creativity, situational awareness, and decision-making skills that players need to both play and create new levels.

Georgia Institute of Technology researchers Mark Riedl and Matthew Guzdial, who are using software to investigate how machines could help people design games, say the appeal of "Super Mario Bros." resides in the algorithmic challenges inherent in the game's mechanics. "The side-scrolling nature of the game means a lot of the game is unobservable to the AI, whereas many simpler arcade games have all information on screen at once," they note.

Moreover, the levels of the game are both patterned and abstract, which Riedl and Guzdial say can "push the envelope" of AI research. The game's nested-pattern structure also is useful for algorithms reliant on lexicographic ordering, according to Carnegie Mellon University researcher Tom Murphy.

New York University professor Julian Togelius says the game's popularity and influence is its biggest advantage for AI scientists, because their work garners a lot of attention as a result.

From Motherboard
View Full Article


Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA


No entries found