Google Deep Mind is releasing a new version of its AI “World” model, called Jenny 3, capable of creating a 3D environment in which users and AI agents can communicate in real time. The company is also promising that consumers will be able to communicate with the world for a long time and that the model will really remember where things happen when you look away from them.
The global model is a type of AI system that can imitate the environment to help train education, entertainment, or robots or AI agents. With global models, you give them a hint and create a place that you can rotate in a video game like this, but instead of crafting the world with 3D assets, all of this is being manufactured with AI. It is an area in which Google is trying a lot. The company demonstrated Jenny 2 in December, which could create interactive worlds on the basis of an image, and has formed a world model team headed by the former co -leadership of the Openi Surah Video Generation Tool.
But the models currently have many defects. For example, Jenny 2 the world was just a minute. I recently tested the “interactive video” from Pixar’s Coofunder -backed company, and it felt as if the Google Street View was running in the blurred version of the blonde, where things changed and changed in ways that I did not expect when I saw around.
Jenny 3 looks like it can be a notable step. According to a blog post, users will be able to create worlds with a indicator that supports the “few” minutes of permanent interaction, which is more than 10-20 seconds of interaction with Geni 2, according to a blog post. Google says that Geni can keep spaces in visual memory for 3 minutes, which means that if you turn away from something in a world and then turn towards it, things like painting on the wall or writing on the chalk board will be in one place. The world will also have 720p resolution and they will run on 24 FPS.
Deep Mind is also adding what he calls the “quick world events” in Jenny 3. Using a gesture, you will be able to work like changing weather conditions in the world or adding new roles.
However, this is probably not a model you will be able to try for yourself. It is launching as a “a limited research preview” that will be available for “a small group of experts and creators” so that its developers can better understand the risks and how to properly reduce them, according to Google. There are also many restrictions, such as limited methods, such as users can communicate with the created worlds, and this text is “often created in the description of the input world.” Google says it is “searching” how Jinn should be brought to “additional testers”.
