Home / Technology / Google's Genie: AI Builds Game Worlds From Your Words
Google's Genie: AI Builds Game Worlds From Your Words
29 Jan
Summary
- Google DeepMind now allows US Google AI Ultra subscribers to use Project Genie.
- Project Genie creates interactive game worlds from text prompts or images.
- The tool is an experimental research prototype with limitations.
Google DeepMind has opened access to Project Genie, an AI tool capable of generating interactive game worlds from user-provided text prompts or images. Starting this week, Google AI Ultra subscribers in the United States can access this research prototype. Project Genie utilizes a combination of Google's Genie 3 world model, Nano Banana Pro image generator, and Gemini.
This release marks a significant step in DeepMind's pursuit of advanced world models, which are considered crucial for achieving artificial general intelligence (AGI). While currently focused on entertainment applications like video games, the long-term vision includes training embodied agents for real-world tasks. Researchers acknowledge the experimental nature of Project Genie, noting its occasional inconsistencies in generating playable environments.
The tool allows users to input text prompts for environments and characters, with an AI-generated image serving as a starting point. Users can also leverage existing photos as a basis for world creation. Currently, world generation and navigation are limited to 60 seconds per session due to computational resources. DeepMind is actively seeking user feedback to improve realism and interaction capabilities in future iterations.
