Google DeepMind just unveiled Genie 3, a “world model” that generates interactive virtual worlds from simple text prompts.
Unlike static video generation, Genie 3 creates navigable environments in real-time at 24 frames per second with 720p resolution, maintaining visual consistency for several minutes.
The system features "world memory" that preserves user actions and "promptable events" allowing on-the-fly modifications like weather changes or new objects.
Google has successfully tested Genie 3 with their SIMA agent for AI training scenarios.
Currently available as a limited research preview, the technology has applications in gaming, robotics training, disaster preparedness, and education, marking a significant step toward artificial general intelligence.
Unlike static video generation, Genie 3 creates navigable environments in real-time at 24 frames per second with 720p resolution, maintaining visual consistency for several minutes.
The system features "world memory" that preserves user actions and "promptable events" allowing on-the-fly modifications like weather changes or new objects.
Google has successfully tested Genie 3 with their SIMA agent for AI training scenarios.
Currently available as a limited research preview, the technology has applications in gaming, robotics training, disaster preparedness, and education, marking a significant step toward artificial general intelligence.
- Catégories
- prompts ia
Commentaires