SAN FRANCISCO, May 26 ― After watching 50,000 episodes of Pac-Man being played, Nvidia's AI was able to recreate a convincing replica of the game from nothing.
On the 40-year anniversary of the beloved game Pac-Man, Nvidia, known for its graphic processing unit technology, announced that it has trained an AI to create an operational replica of the game simply from watching it being played 50,000 times ― no human intervention required.
The artificial intelligence used to accomplish this is called NVIDIA GameGAN, a neural network model which “mimics a computer game engine by harnessing generative adversarial networks, or GANs.” These GANS use a pair of competing neural networks which work together to create new content that looks impressively like the real deal it was designed to copy.
Trained on 50,000 episodes of PAC-MAN, a new #AI model created by NVIDIA Research called GameGAN re-created the game environment in just days. Celebrate the #pacman40th anniversary and see how GameGAN puts an AI spin on this classic game. https://t.co/zOoioy0MU9 @BandaiNamcoUS pic.twitter.com/nyrXp9xhdA
— NVIDIA (@nvidia) May 22, 2020
And thus, without a game engine, GameGAN recreated Pac-Man without anyone programming it with code.
In 2019, the company used this same GAN technology to create two other apps: GauGAN and GANimal. While the former could take user doodles and convert them into photorealistic landscapes, the latter could put the expression of one animal ― like that of your pet, for example ― realistically onto the face of another.
As for GameGAN, Nvidia states that this tech could possibly be eventually used by game developers to “automatically generate layouts for new game levels,” or by AI experts to efficiently generate “develop simulator systems for training autonomous machines.”
The AI-generated version of Pac-Man will be available for the public to play later this year. ― AFP-Relaxnews
