Google DeepMind has unveiled Genie 3, a new AI 'world' model that's making waves by generating interactive 3D environments in real-time. This latest iteration promises significantly longer user interaction times and improved memory retention within the generated worlds, marking a considerable leap forward in AI-driven world creation.
Genie 3 allows users and AI agents to engage with these virtual spaces for extended periods, a substantial improvement from Genie 2's limited 10-20 seconds. According to a Google blog post, the new model facilitates a 'few' minutes of continuous interaction. Furthermore, Genie 3 boasts a visual memory span of approximately one minute, ensuring that objects and details within the environment remain consistent even when the user looks away and then returns their gaze. Imagine turning back to a wall and seeing the same paint or writing still in place – that's the level of persistence Genie 3 offers.
The generated worlds will be rendered in 720p resolution at a smooth 24fps. DeepMind is also incorporating 'promptable world events,' allowing users to dynamically alter the environment through prompts. This means you can change the weather conditions or introduce new characters into the world, adding a layer of customisation and control to the experience. Google is initially launching Genie 3 as a 'limited research preview' to a select group of academics and creators, with plans to explore expanding access to more testers in the future.
Artículo generado mediante LaRebelionBOT
No hay comentarios:
Publicar un comentario