
Google DeepMind has unveiled a new version of its AI game engine, Genie 3, that can build entire video game worlds in real time. This development helps users seamlessly create 3D environments simply by writing a prompt for the AI.
These spaces can be explored, creating a gaming experience without the traditional game development toolset.
AI-generated video games are becoming increasingly sophisticated, and Genie 3 is another evolution of the progress AI has made recently. This AI can create interactive, 3D scenes that can react to what a user does within the space in real time.
Unlike its predecessors, Genie 3 can make it possible for users to spend longer amounts of time within a virtual world and now retains visual features, such as text on a chalkboard or a paint pattern on a wall, even when someone looks away and back again.
Google's Genie 3 has improved memory and efficiency that let it run at 720p with 24 frames per second, which is a major upgrade from Genie 2 that allowed only 10-20 seconds of interaction before crashing.
With the new model allowing a few minutes of playing time to understand the environments, Genie 3 can be useful for educational, entertainment, and testing purposes for how robots or AI agents function in digital environments.
The tool also includes a new feature called “promptable world events”. This means users can type in changes they want, like changing the weather or adding a new character, and the AI will adjust the world instantly.