Sign In

Communications of the ACM

ACM TechNews

The Machines Have Taught Themselves to Make Mario Levels


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
Super Mario screenshot

Research showed how AI could create games that automatically adapt to a player's skill level.

Credit: Emulator Online

Using a modern artificial intelligence technique called Generative Adversarial Networks, researchers at the University of California, Santa Cruz discovered a way to create new levels of Super Mario Bros by analyzing an existing game level. The project, called MarioGAN, demonstrates how AI could create games that automatically adapt to the player's skill level.

MarioGAN uses an AI technique called Generative Adversarial Networks, in which one neural network looks at a set of training data and tries to create new samples, while a second neural network tries to distinguish between the "real" training data and the new "fake" data. The first network learns to make better fakes, leading to more realistic Mario gaming levels. The researchers then developed a way to search the latent space of the neural network for certain characteristics, allowing them to create a level that gradually increased in difficulty.

The approach offers potential for creating an endless level that automatically grows more challenging over time or one that focuses on finding difficult-to-reach items.

From Fast Company
View Full Article

 

Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account