Google’s Genie 2 artificial intelligence to generate 3D worlds + video

Google DeepMind has just released a new version of Genie artificial intelligence. The previous version of this model could generate two-dimensional worlds for users and artificial intelligence Genius 2 able to produce 3D worlds It is fully interactive.

Although Genie 2’s features are apparently built for game development, Google says it’s not a game development engine. In fact, Google has developed a kind of Diffusion Model that can create images that will display a completely simulated world when the user interacts with them.

Models built with Genie 2 artificial intelligence

Genie 2 can generate ideas about the surrounding environment while generating the required frames and has the ability to model water, smoke and physical effects well. However, some of these models are very similar to elements of games.

Also, this model has no limitations in rendering different scenes and can display landscapes in first person, third person and isometric (top down view).

All you need to create a 3D model with Genie 2 is an image as a primer, which can be generated by Google’s Imagen 3 AI or from any Used another image.

Another thing worth noting is that Genie 2 can remember parts of a previously simulated world as soon as you see them again and restore certain elements. This feature makes Google’s artificial intelligence different from its competitors like Oasis.

Regarding Genie 2’s limitations, DeepMind says that the model can create fully interactive worlds for 60 seconds. Also, in the samples exhibited by the technology giant, models with a much shorter time are also seen.

Currently, Google DeepMind has developed the Genie 2 artificial intelligence model as a tool for training and evaluating other artificial intelligence agents, including its SIMA algorithm, so that artists and designers can use it. Therefore, we cannot expect this model to be released to the public soon.