Pac-Man was released on May 22, 1980, the legendary comecocos video game that sparked a real revolution, first in arcades, then in computers and consoles.
Its legacy has been enormous, and now NVIDIA wanted to pay special tribute to the Bandai Namco title. For this, they created an artificial intelligence system that just play 50,000 games Pac-Man has recreated the game from scratch, in itself, respecting the mechanics and the rules of the game.
No code is generated, images are generated
The system created by NVIDIA was named GameGAN, and as its name suggests, uses networks of antagonistic generators (GAN, Contradictory generative networks) a type of algorithm that takes advantage of the unsupervised learning philosophy.
This approach to this type of problem is already familiar in NVIDIA, which in the past used it in other areas such as artistic with GauGAN or the generation of animal images from stuffed animals with GANimal. The idea is that the system learns the rules of the game simply by observing it over and over again.
According to NVIDIA GameGAN, this is the first neural network model that emulates the behavior of a video game engine. It uses two competing neural networks (one generator and the other discriminator), and ends up creating new content that emulates the original content they looked.
The idea is relatively simple: when the artificial agent plays the game generated by this unique engine, GameGAN reacts to these actions by generating images in real time. This engine does not generate code which is then executed, but it essentially generates frames consistently because the different actions that occur in the game are recognized and evaluated.
GameGAN wants to go beyond Pac-Man
Observing the development of these games, the general model of the static elements of the environment, but also abides by its rules. for example Pac-man can’t break through the walls or by passing a ball, he eats it.
With this system, it would therefore be possible to offer developers a tool to automatically create new levels for their games, but it also allows AI researchers to have a tool to develop simulation systems to train autonomous machines.
The creators of this system underlined that with GameGAN, they could “have an artificial intelligence learn to imitate the rules of conduct and the laws of physics just by watching videos and seeing how agents make certain decisions in their environment. ”
The Pac-Man game created by this artificial intelligence Will be available soon on the NVIDIA AI Playground site thanks to the agreement with Bandai Namco for the pleasure of all.
More information | NVIDIA