To celebrate PAC-MAN’s 40th Anniversary, Nvidia researchers recreated this iconic game using Artificial Intelligence. For this, they used GameGAN, the first neural network that is capable of imitating a graphic engine of a computer game taking advantage of Adversary Generative Neural Networks (GAN in English).
Composed of two competing neural networks, a generator, and a discriminator, GAN-based models learn to create new content convincing enough to impersonate the original, and this means that the AI version of this game is completely functional.
“This is the first research to emulate a graphics engine using GAN-based neural networks,” says Seung-Wook Kim, Nvidia researcher and lead author of the project. “We wanted to see if AI could learn the rules of an environment just by looking at the script of an agent moving through the game. And so it was.”
While an artificial agent plays the game generated by GAN, GameGAN responds to the agent’s actions, generating new frames from the game environment in real-time. GameGAN can even generate never-before-seen game layouts if you train in-game scripts with multiple levels or versions.
This ability could be used by game developers to automatically generate layouts for new levels of a game, as well as by AI researchers to more easily develop simulator systems for autonomous machine training.
“We were amazed when we saw the results, in disbelief that AI could recreate the iconic PAC-MAN experience without a game engine,” said Koichiro Tsutsumi of BANDAI NAMCO Research Inc., who provided the PAC-MAN data for train GameGAN. “This research presents exciting possibilities to help game developers accelerate the creative process of developing new level designs, characters, and even games.”