NVIDIA GameGAN is an AI model created by researchers at NVIDIA that was given a simple task – watch arcade classic Pac-Man being played over and over, study its design and look, and then go off an recreate it from scratch without an engine. Translation, NVIDIA is using AI for videogame emulation in the purest sense – creation. With the pellet chomper turning 40 this year NVIDIA has made sure Pac-Man is celebrating in style.

Here’s a look at Skynet NVIDIA AI’s Pac-Man – “a fully functional version of the classic without an underlying game engine”. And yes, it’s playable. Or will be later this year when it goes public.



How it all came together sounds simple enough, even if it’s sci-fi sorcery of the highest order. The AI was trained on 50,000 episodes of the game, understanding the rules and behaviour. From there the neural network mimics a computer game engine via generative adversarial networks, or GANs. Okay, maybe not so simple. Basically, an artificial agent then plays Pac-Man with the AI creating convincing frames of art and animation based on player interaction.

The trained GameGAN model then generates static elements of the environment, like a consistent maze shape, dots and Power Pellets — plus moving elements like the enemy ghosts and PAC-MAN itself.

It learns key rules of the game, both simple and complex. Just like in the original game, PAC-MAN can’t walk through the maze walls. He eats up dots as he moves around, and when he consumes a Power Pellet, the ghosts turn blue and flee. When PAC-MAN exits the maze from one side, he’s teleported to the opposite end. If he runs into a ghost, the screen flashes and the game ends.

“This is the first research to emulate a game engine using GAN-based neural networks,” said Seung-Wook Kim, an NVIDIA researcher and lead author on the project. “We wanted to see whether the AI could learn the rules of an environment just by looking at the screenplay of an agent moving through the game. And it did.”

Interestingly the AI can supposedly create layouts or generate scenes outside of what it has “seen” – which opens the door to fascinating real-world applications. From advances in science, automotive possibilities, to the creation of game levels. From scratch without the need for templates or rules or tile-sets or procedural techniques. Or, as this demo, err, demonstrates - without a game engine.

Theoretically in some not too distant future someone could design and create a new videogame character then ask an AI to show them how they might look in a Tomb Raider style setting or in a survival horror. And then play it.

In this scenario they may have to ask nicely because if the world of science fiction has taught us anything is that if an AI can create Pac-Man by absorbing the data from it being played, our days being on top of the digital food chain are numbered.

“We could eventually have an AI that can learn to mimic the rules of driving, the laws of physics, just by watching videos and seeing agents take actions in an environment,” said Sanja Fidler, director of NVIDIA’s Toronto research lab. “GameGAN is the first step toward that.”

Amazing stuff. Oh, and happy 40th Pac-Man! For more info - head here.