Google DeepMind introduces Genie 3, an advanced AI model that generates 3D environments for user interaction. This version promises longer engagement times, allowing continuous interaction for several minutes, a significant upgrade from Genie 2’s 10-20 second limit. Genie 3 retains visual memories for about a minute, ensuring elements do not change unexpectedly when looked away from. The environments operate at 720p resolution and 24 frames per second, enhancing user experiences. World models, a type of AI system, allow for interactive simulations for various applications like education and entertainment, emphasizing the potential of AI-generated worlds.
Genie 3 allows users to generate interactive worlds with a prompt, supporting extended periods of interaction, drastically improving from Genie 2's limitations, offering a more immersive experience.
Google states that Genie 3 can retain visual memory for around a minute, ensuring elements like wall paint or chalkboard writing stay in place when users look away.
These worlds will be rendered at 720p resolution and run at 24fps, enhancing the realism and fluidity of experience users will have in these generated environments.
World models are AI systems that create interactive 3D spaces autonomously, designed for applications like education and entertainment, showcasing the potential of AI-driven environments.
Collection
[
|
...
]