First, the machinesat Go. Now, they're making games harder for us.
In two separate papers, researchers have developed artificially intelligent models -- namely generative adversarial networks -- to build new game levels for Super Mario Bros and Doom. And with the addition of another tool, researchers found level generation for Super Mario is "further improved."
GANS are a form of AI used in unsupervised machine learning. Two neural networks -- the "generator" and the "discriminator" -- are pitted against each other so the generator learns better ways to fool the discriminator, which (of course) tries not to be tricked.
"Level design usually heavily relies on domain expertise, good practices, and an extensive play testing," researchers wrote in the Doom paper.
"To deal with these issues, several game researchers [spent] considerable effort on studying and designing procedural content generation systems that, exploiting machine learning and search algorithms, can model the level design process and assist [the] human designer," they added.
GAN is the same technology a researcher used to generate what CNET contributor Bonnie Burton called portraits of "recreate partially erased images last month as well as create last October." in March, and what Nvidia used to
It's not the first time people have tried applying GAN to games either. Programmers at Maverick Games published results on news site Gamasutra in February pertaining to their experience using GAN to create new game levels for an RPG called Fantasy Raiders.
Despite limitations -- researchers found GANs had occasional problems that resulted in broken structures -- the programmers agree with the researchers writing in the Doom paper that GANs are a "promising approach to level generation."
The AI-generated levels aren't yet available on Doom and Super Mario as the projects are still prototypes. There's no telling when they could become available though, so enjoy the games while they're still easy enough to play.
Tech Enabled: CNET chronicles tech's role in providing new kinds of accessibility.