Google DeepMind has introduced Project Genie, an AI tool designed to construct interactive game worlds from text prompts or images. This technology is now available to Google AI Ultra subscribers in the U.S., offering them the opportunity to engage with this experimental research prototype. Project Genie leverages Google’s latest world model, Genie 3, along with the Nano Banana Pro image generation model and Gemini.
The launch of Project Genie follows the research preview of Genie 3 by five months and reflects DeepMind’s commitment to gathering user feedback and training data as part of its ongoing efforts to advance more sophisticated world models. These AI systems create internal representations of environments, enabling the prediction of future outcomes and the planning of actions. DeepMind and other industry leaders view world models as pivotal in progressing towards artificial general intelligence (AGI).
As the competition in the world model space intensifies, with initiatives like Fei-Fei Li’s Marble and Runway’s recent world model release, Project Genie stands out as a significant development. The experimental nature of the tool is acknowledged by DeepMind researchers, who highlight its variable performance in generating playable worlds. This technology opens doors to a realm where user engagement and feedback play a crucial role in shaping the future of AI-driven interactive experiences.
Source: TechCrunch