A neural network called Game Engine can recreate Doom from scratch by predicting the next frame based on gameplay recordings. It runs at 20 FPS without the original code, producing visuals and game state that look identical to the real thing.
Researchers just turn an entire video game into a neural network. It's called Game Engine and it can play the classic game Doom in real time at 20 frames per second. But here's the crazy part. There's no actual game code running. It's all being generated by AI. Here's how it works. First, they trained an AI to play Doom and recorded tons of gameplay. Then they use that data to train a special kind of AI called a diffusion model. This model learns predict what the next frame of the game should look like based on previous frames and player actions. The result, a neural network that can simulate Doom so well that human players have a hard time time telling it apart from the real game. It keeps track of health, ammo, enemies, everything the original game does. It's all happening inside the AI. This is a huge deal because it opens up a whole new way to create and run video games. Imagine being able to design a game just by describing it to an AI or instantly modifying existing games without ever having to change any code. Game Engine is just the beginning. We're entering an era where the line between human made and AI generated content is blurring. More every day follow for more AI advancements and cool future tech.
AI News: Dots, GPT-6.1 Sol, Sonnet 5.5, Gemini 4, and everything you need to know
