Google DeepMind announced Genie 3 is a new version of the universal artificial intelligence model that is capable of creating full-fledged interactive virtual environments simply based on a textual description. Unlike of the previous version The new model generates the world at high speed — in real time at 24 frames per second and 720p resolution.
What Genie 3 can do
The model allows you to create dynamic environments that look and behave as close to reality as possible. The features include realistic water behavior, lighting, interaction of objects in space, and ecosystem modeling with detailed animal habits and plant growth.
In addition, Genie 3 supports fictional world creation. You can create animated characters, fantasy locations, or historical eras with high visual fidelity and immersion.
DeepMind engineers note that the model takes into account previous frames when generating new ones, forming a kind of “visual memory”. This allows the environment to remain stable and logically consistent for several minutes, and changes in space look natural and not accidental. The Genie 3’s memory covers up to a minute of visual history, which is a significant breakthrough for this type of model.
Despite its amazing features, Genie 3 is not without its drawbacks:
- Limited freedom of action — although the world responds to requests, not all events are performed by the “agent” itself, i.e. the virtual participant.
- The complexity of the interaction of several agents — modeling the behavior of several independent objects in the same environment is still a problem.
- Insufficient geographic accuracy — created locations do not match real locations from geodata.
- Problems with text rendering — written interface elements or labels often look fuzzy if not explicitly defined.
- Interaction time — so far, the model supports only a few minutes of continuous interaction.
Currently, the Genie 3 model is only available to a limited number of researchers and creators. Google plans to gradually expand access for testing in the future.
Genie 3 — is another step towards interactive, visual artificial intelligence that can become the basis for new formats of games, training, design, and simulations. The model combines generative AI with physics, biology, and storytelling elements, opening up a new level of creative freedom. If the technology is able to scale and overcome current limitations, it will potentially change the approach to creating virtual content — from entertainment to scientific modeling.
Source: neowin
Spelling error report
The following text will be sent to our editors: