The Bar For AI Gets Lowered

Once upon a time, we hoped for robots that could beat humans at cultured, highbrow games like chess. But as society falters and falls around us, it's another game that we hope our robot overlords will master: Super Mario Bros.

Julian Togelius and Sergey Karakovskiy of the IT University of Copenhagen, Denmark, have created a contest to create software that will learn how to play Mario successfully the same way that humans do - by playing it over, and over, and over. It sounds like a joke, but Togelius is convinced of its importance in comparing attitudes in software and artificial intelligence development, and also of his choice of test game:

As far as I'm concerned, Mario is the computer game, both as a gamer and as a good machine-learning challenge that requires a broad set of skills.

(The actual test game will be a recreation of the original game, rather than the real Super Mario, sadly.)

Winners will be named - and given cash prizes! - at London's Games Innovation Conference later this month, and Italy's IEEE Symposium on Computational Intelligence and Games in September.

Race is on to evolve the ultimate Mario [New Scientist]