in

40 Years After Pac-Man Took Over The World NVIDIA Has Recreated The Game Using AI

40 Years After Pac-Man Took Over The World NVIDIA Has Recreated The Game Using AI
Pac-Man Has Been Recreated Using AI. NVIDIA

After Pac-Man’s release 40 years ago, the game became so popular it spawned everything from children’s cartoons, to hit pop songs, and a never-ending string of sequels and clones that persist to this day. Hop onto the Google Play Store, search for Pac-Man, and you’ll be greeted by numerous entries for official titles and unlicensed knock-offs alike. As it turns out, NVIDIA has a new version of Pac-Man in the works as well, that should be released later this year. What makes this particular version special though, is that is was created using Artificial Intelligence.

NVIDIA and Bandai Namco (the original publishers of Pac-Man) worked together to commemorate the 40th Anniversary of Pac-Man’s release by reverse engineering the game. They used new AI models, dubbed NVIDIA GameGAN, which processed 50,000 episodes of the game, and the end results was a fully-functional version, produced without an underlying game engine.

NVIDIA claims GameGAN is the first neural network model that can mimic a video game engine by leveraging the capabilities of generative adversarial networks, or GANs. If you recall, NVIDIA previously released something called GauGAN, which is a deep learning model that’s capable of transforming the simplest of doodles into near-photorealistic images. GauGAN used generative adversarial networks as well, but with GameGAN NVIDIA’s research team has taken things a significant step forward, to not only recreate the game’s imagery, but the mechanics as well.

A GAN is comprised of two competing neural networks, a generator and a discriminator. And the GAN-based models NVIDIA’s researchers have created are able to learn to create new content that’s nearly indistinguishable from the original. Ultimately, GameGAN could help speed up and ease game development, by automatically generating layouts for new game levels, for example. However, there are many potential other uses as well. According to NVIDIA, GameGAN could also help simplify development of simulator systems for training autonomous machines, where the AI can learn the rules of an environment before actually interacting with any objects in the real world.

“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.”

NVIDIA plans to make the AI-generated version of Pac-Man available later this year on AI Playground, where the company posts demos of many of its AI research projects for the public to experience first-hand.

Source: www.forbes.com

What do you think?

48 points
Upvote Downvote

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

5 Deep Learning Challenges To Watch Out For

5 Deep Learning Challenges To Watch Out For

How Neural Network Can Be Trained To Play The Snake Game

How Neural Network Can Be Trained To Play The Snake Game