Competition to design NES-beating AI

New contest is looking to develop artificial intelligence that can play the perfect game of 'Super Mario Bros'
New contest is looking to develop artificial intelligence that can play the perfect game of 'Super Mario Bros'

A new contest at the University of Copenhagen in Denmark is encouraging students to design an artificial intelligence that can play the perfect game of Nintendo's classic Super Mario Bros.

University Prof Julian Togelius also hopes that, in addition to generating some healthy competition amongst his students, the results could be hugely influential within the games design and development community, helping devs make better games.

"When Deep Blue beat Garry Kasparov at chess in 1997, the world got a sharp taste of the power and potential of artificial intelligence," notes a New Scientist report on the project.

Bah! Forget chess, we are far more interested in a robot that can master that other "complex game deeply rooted in human culture" – Super Mario!

Making better games easier to make

"I'm interested in creating learning or evolving software that can make games more fun, or easier to produce," says Togelius.

Togelius' contest charges his students to make software that can teach itself how to improve at Mario, with the hope that the contest will "bring together advocates of different ideas about how to make software learn; those who mutate and evolve their code to improve it and those who prefer more established, mathematical techniques to spot and learn relationships."

"Having a direct shoot-out like this is a good way to compare them," says Togelius.

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

Head right here for more details of the contest - the winners will be announced at the Games Innovation Conference in London at the end of August.

Via New Scientist

Adam Hartley