Tech Desk 
Aug 29, 2024
Like other major tech giants, such as Microsoft, OpenAI, and more, Google has been stepping up its game in the realm of artificial intelligence (AI) when it comes to achieving something trend-setter.
Bringing about a significant milestone in AI, one which supplements Google's visually potent portfolio in AI, Google researchers have created a neural network called GameNGen, which is capable of developing real-time gameplay for the classic shooter Doom without using a game engine.
The measure of Google's game-generating AI system's potential can be gauged by its ability to generate playable gameplay at 20 frames per second (fps) on a single chip, wherein each frame is an estimation of a different diffusion model.
“We present GameNGen, the first game engine powered entirely by a neural model that enables real-time interaction with a complex environment over long trajectories at high quality,” stated Google researchers in a paper published on the preprint server arXiv.
The development marks Google's first-ever venture towards an AI-backed system that simulates a complex video game with precision and high-quality graphics and interactivity.
Google's custom-built AI accelerator chip called Tensor Processing Unit (TPU) is at the helm of GameNGen, enabling it to imitate Doom’s intricate 3D environments and chaos efficiently, all without utilising the traditional elements of a gaming engine.
