AI recreates Pac-Man from scratch by just looking at the screenplay

To celebrate 40-years of Pac-Man, NVIDIA GameGAN, a new artificial intelligence model created a fully playable version of the old classic. The AI model was trained on 50,000 episodes of the game, it learned and created all this without the underlying game engine. This is amazing! 

Generative adversarial networks or GANs, are made up of two competing neural networks, a generator and a discriminator. NVIDIA Research has the same as GameGAN which can mimic a computer game engine and create new content. 

In a blog post about the project NVIDIA Research team member and lead author Seung-Wook Kim said, "This is the first research to emulate a game engine using GAN-based neural networks. 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."

With GAN, AI model learns from game play episodes and recreates the screen elements such as maze shape, dots with Pac-Man himself. This also creates the game rules by just watching the element movements in the game. 

This can be really helpful for the developers to auto generate layouts for new game levels. This can also be really game changer for simulating environments and recreate the virtual replica of the same. AI can suggest what can be changed in business environments to make processes much simpler or improve the overall output.