GameNGen: Diffusion Model Powers Game Engine, Simulates DOOM at 20 FPS on Single TPU

GameNGen: Google's AI-driven game engine simulates DOOM at 20 FPS on a single TPU, marking a breakthrough in real-time neural model gaming.

GameNGen is the first game engine entirely powered by neural models.

Google has made another significant advancement in artificial intelligence.

In a recently published paper, researchers introduced a neural network that can generate real-time gameplay for the classic shooter game Doom without using a traditional game engine.

This system, named GameNGen, represents a major step forward in AI. It generates playable game scenes at 20 frames per second on a single chip, with each frame predicted by a diffusion model.

GameNGen is the first game engine driven entirely by neural models, capable of high-quality, real-time interactions with complex environments over long trajectories.

This achievement marks the first time AI has fully simulated a complex video game. GameNGen runs on a single Tensor Processing Unit (TPU), Google's custom AI accelerator chip, efficiently handling Doom's complex 3D environments and fast-paced action, all without the usual components of a game engine.

The research has sparked widespread discussion.