Google’s new AI model creates video game worlds in real time

Google DeepMind verses a new version of the “World” model, called Genie 3, is able to generate 3D environments that users and artificial intelligence agents can interact with it in real time. The company is also that users are able to interact with the world for a long longer period before and that the model will actually remember the place of things when you look away from them.

Global models are a type of artificial intelligence system that can simulate environments for purposes such as education or entertainment or to help train robots or artificial intelligence agents. Through global models, you can give them a router and create a space in which you can move as you do in a video game, but instead of the world manually made with 3D origins, all of this is created with AI. It is an area that makes Google a lot of effort; The company Genie 2 was shown in December, which can create interactive worlds based on a picture, and it builds a team of the world models led by a former leader of the Openai video generation tool.

But the models currently have a lot of defects. Genie 2 worlds were playing up to only one minute, for example. I recently tried a “interactive video” from a company supported by Pixar founder, and I felt as if he was going through a blurry version of Google Street View where things turned and changed in ways I did not expect when I looked around.

Genie 3 seems to be a noticeable step forward. Users will be able to generate worlds by demanding a few minutes of continuous interaction, which rises from 10-20 seconds of interaction with Genie 2, according to the blog’s publication. Google says that Genie 3 can keep the spaces in the visual memory for about a minute, which means that if you move away from something in a world, then return to it, it will be things like painting on the wall or writing on the blackboard in the same place. The world will also have a resolution of 720 pixels with 24 frames per second.

Deepmind adds what you call “world -rejected events” to Genie 3 as well. Using a wave, you will be able to do things like changing weather conditions in a world or adding new characters.

However, this may not be a model that you will be able to try yourself. It is launched as a “limited research inspection” that will be available for a “small group of academics and creators” so that its developers can better understand the risks and how to reduce them appropriately, according to Google. There are also many restrictions, such as limited ways that users can interact with with the created worlds, and the readable text is created “often only when it is provided in describing the global input.” Google says it is “exploring” how to bring Genie 3 to an “additional test”.