What can users do inside these worlds?
Users can explore the scenarios, give text commands, and see how new actions are caused in real time. However, virtual characters can't modify the environment; they can only move within it.
Google's AI can generate 3D worlds in real time for playing, training robots, or simulating environments
Users can explore the scenarios, give text commands, and see how new actions are caused in real time. However, virtual characters can't modify the environment; they can only move within it.
Google introduced its new artificial intelligence Genie 3, a tool capable of generating complete 3D environments that can be instantly explored and modified.
Developed by DeepMind, this technology goes far beyond entertainment. It could also be used to train robots and autonomous systems in artificially caused virtual environments.
Genie 3 allows users to create scenarios from scratch by simply uploading an image or writing a prompt. In just a few seconds, it generates a world that can be explored as if it were a video game.
The most remarkable feature is that users can interact with that environment. They can move freely, causing changes that persist over time thanks to its visual memory.
Additionally, it uses a technique called autoregressive generation, which allows it to build each image based on the previous ones, maintaining visual coherence.
For example, if users destroy something in the environment and move away, when they return, the change will still be there. This persistence remains for at least one minute.
For now, it is also not possible for multiple users to interact simultaneously or for real places to be reproduced with precision.
Google explains that Genie 3 also has enormous potential in the field of applied artificial intelligence. By allowing realistic situations to be simulated, it can be useful for training autonomous agents without physical risk.
This could include everything from drones to industrial robots or autonomous vehicles that need to practice in safe environments before doing so in real life.
For now, it will not be open to the public. Google announced that Genie 3 will be offered to researchers and developers to continue testing it and expanding its possibilities.
The idea is that eventually it could reach more users, although there is no specific date. For now, it will be an experimental tool for academic and technological circles.