Google is rolling out Project Genie, an AI that generates interactive 3D worlds from text prompts or images. Describe what you want to see, and Genie creates an environment you can explore in real time.
Project Genie runs on Genie 3, a world model developed by DeepMind. World models simulate how environments behave and how actions affect them. Unlike a static image or video, Genie generates the path ahead as you move, simulating physics and interactions at 24 frames per second in 720p resolution.
Google positions this as a creative tool for game designers and artists, useful for ideation and rapid prototyping. A spokesperson noted that Genie “is not a game engine and can’t create a full game experience.” This is a beta test. Generated worlds disappear after 60 seconds, and to use it you have to be in the U.S. and have a $250/month Google AI Ultra subscription.
This kind of announcement is easy to take for granted or simply ignore. Simulated world generators have been around for decades. Why would anyone care about this?
Project Genie’s ability to generate entertaining fantasy landscapes will help train AI systems that will eventually control robots, autonomous vehicles, and industrial automation. However, DeepMind’s mission is AGI (artificial general intelligence), and world models are essential infrastructure for that goal. An AI that can predict how environments change and how actions produce consequences is an AI that can reason about the physical world.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.