Pac-Man’s ghost

GameGAN. © Nvidia.

GameGAN is a powerful new neural network model created by a Toronto-based Nvidia artificial intelligence research lab. This model is able to generate a functional simulation of a video game just by watching it being played on-screen and observing keyboard or joystick interactions. It is based on generative adversarial networks, or GAN. As a tribute to Namco’s timeless game launched forty years ago this year, in 1980, researchers decided to test GameGAN with Pac-Man, and the result is surprisingly good. Nvidia says that its technology could help developers create games, for example by developing new levels. This said, Pac-Man is a straightforward game, and GameGAN still needed 50,000 hours of game time to learn it (conveniently generated by another artificial intelligence!) AI-developed games might not be in our immediate future.

⇨ YouTube, “NVIDIA GameGAN: Celebrating 40 years of Pac-Man with game-changing AI.”

Speaking of Pac-Man, we recommend this excellent video that explains why it was so revolutionary in its time, how each ghost has its own personality, and how the game changed the future of video games: 

⇨ YouTube, “Pac-Man | Design Icons.”

⇨ Ars Technica, Sam Machkovech, “After watching 50,000 hours of Pac-Man, Nvidia’s AI generated a playable clone.”

⇨ Nvidia, Isha Salian, “40 Years on, Pac-Man recreated with AI by Nvidia researchers.”