Google DeepMind has begun releasing a new version of the Genie 3, an AI “world” model. This AI model can generate a 3D environment that allows both users and other AI agents to interact in real time. The company promises that users can interact with these worlds for much longer than their previous models. Demon 3 It is also designed to “remember” the object's position even if the object is not in the user's gaze. This update takes a step forward in creating a more interactive and persistent virtual environment.
What is the AI world model and what's new in Genie3
The World Model is a type of AI system that simulates the digital environment of apps needed for education, entertainment, or training AI agents. These systems generate space based on prompts, allowing users to move around like video games. Unlike traditional games that rely on hand-crafted 3D assets, these environments are built using AI. Google is researching this area Demon 2 A model introduced last year. This AI model can create an interactive world from images. The company also established a dedicated World Models team led by Tim Brooks, former co-leader of Openai's SORA video tools.However, the current capabilities of such models remain limited. For example, Genie 2 only allows for approximately 1 minute of interaction. Another interactive video tool developed by a startup backed by the co-founder of Pixar, provided an experience similar to the blurry version of Google Street View, using unexpectedly shifted visuals.The Genie 3 offers some technical improvements. According to the blog post, users will be able to interact with the generated environment for 10-20 seconds to several minutes in Genie 2 for several minutes. The model remembers visual elements for about a minute, so if the user turns away and returns to the scene, features like wall markings and writing on the blackboard may remain intact. These environments are rendered at 720p resolution and 24 frames per 24 frames.DeepMind is also built in “A rapid world event“ Genie 3 allows users to adjust weather and other factors, as well as add text via written prompts.Access to Genie 3 will be restricted at launch. Google is releasing it as a limited research preview of a small group of scholars and creators to study risks and improve safety measures. Current limitations include narrow interaction options and difficulty in generating easy-to-read text unless included in the initial prompt. Google has shown that it could expand access in the coming days.
