Reports or announcements from educational institutions or government education departments

Google DeepMind's Genie 2: AI Foundation Model Generates Action-Controllable 3D Worlds from a Single Prompt

Dec 4, 2024, 03:20 PM

Google DeepMind has announced Genie 2, their most advanced AI model capable of generating endless, playable 3D environments from a single image or text prompt. Genie 2 creates diverse and consistent worlds in stunning 720p resolution, supporting both first-person and third-person perspectives. The virtual environments are action-controllable and interactive, playable in real-time with just a keyboard and mouse for up to a minute. This large-scale foundation world model supports both human and AI agents, simulating different outcomes from the same starting point and remembering previously unseen parts of the environment. Genie 2 represents a significant advancement in prompt-to-game technology, potentially unlocking new capabilities for embodied agents and transforming the future of video game development.