It is rare that one sees a game that features adaptive difficulty, which is certainly true when it comes to classic games, which almost never adapted their difficulties.
But what if games like Doom or Super Mario Bros. featured evolving difficulty? As reported by The Register, this is something AI researchers are testing out with a system called “general adversarial networks” or GANs.
GANs is a system of two networks called a generator and a discriminator. The generator builds false training data samples while the discriminator discerns whether the samples are real or fake. These two compete with one another. As the process continues, the generator creates increasingly realistic samples in an attempt to deceive the discriminator.
AI researchers used GANs to create new levels for both Super Mario Bros. and Doom. The aim of this research topic is that this research will help game creators develop games with better level designs.
"Level design usually heavily relies on domain expertise, good practices, and an extensive playtesting," said the Doom paper. "To deal with these issues, several game researchers are spending 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 human designer."
GANs isn’t a full-proof system by any means. The researchers ran into some problems while creating Super Mario levels. "MarioGAN generates new levels very quickly but occasionally makes structural mistakes such as incompletely assembling pipe tiles," says Adam Smith, co-author of the paper and an assistant professor at the University of California, Santa Cruz. "This project benefits AI research by helping us to map out the space of which techniques work where and when they fall apart when a hidden assumption they make is broken."
At the time of this writing, both of these AI-created games are merely prototypes. However, the code for the Super Mario GAN is available. The team wants to eventually create levels that constantly adapt their difficulty to the way the player plays the game. If this research pans out, perhaps we’ll eventually see more games with adaptive AI. It certainly would make titles interesting.
Make sure to check out the article over on The Register for a more technical breakdown of how GANs works.