NO GAME ENGINE

This is a fully playable video game with no game engine underneath it. Every single frame is imagined by a neural network in real time. It's called Mira, an open source AI world model running a live 2v2 Rocket League for four players at once. It looks like a 3D world, but there isn't one inside. No physics code, no graphics code, no coordinates stored anywhere. Just a network predicting what the next frame should look like. The model learned the entire game by watching 10,000 hours of AI bots playing other with zero human data. It works. The model manages to simulate many aspects of Rocket League. The cars move, they can kick the ball, you can score goals. The four cars move in sync on the four screens. The model tracks how much boost you have and doesn't let you boost when you're out. Demolitions work. It shows event messages like "Shot on goal" or "Save". Getting rammed, saves and demolished and goal all on a single GPU. Limitations. Replays. After a goal, Rocket League cuts to a replay of the shot. But the model only has a roughly four-second context window, so by the time the replay plays it has no memory of what actually happened: it simply makes something up. The replay it generates looks plausible but doesn't match the goal that was really scored. The strange memory only lasts about four seconds. So when a goal triggers a replay, the model has already forgotten what happened and invents a completely fake highlight of a moment that never occurred. Why simulate video games? The real goal isn't teaching AI to simulate video games in the first place. Why play Mira when you can just play Rocket League? To which we say: You should play Rocket League! Our work is not meant to replace video games with AI models. The reason researchers are excited about world models is rather physical AI: think robotics and self-driving. In AI's expensive to collect data and testing the models is dangerous, since you don't want to crash a car or break a robot. If we train a world model that predicts with high accuracy how a car would behave in the real world driven by a model we're testing, we can evaluate it by running it in the world model, which is safer and quicker. If the world model is accurate enough, you could even directly train in the world model. Simulating video games is therefore a stepping stone to physical AI: it's a setting where data is abundant and it's cleaner than real-world videos since it's generated by a game engine. But learning the mechanics of an AAA video game purely from observing video is still no easy feat, so the hope is that what we've been studying by studying video game world models will be useful in real-world settings later. So robots and self-driving cars can train safely in a virtual one instead of crashing in reality.