Google’s DeepMind has unveiled Genie, an artificial intelligence model capable of converting images into playable video games effortlessly.
Genie’s Capabilities:
The Genie model, comprising 11 billion parameters, has undergone training on over 200,000 hours of two-dimensional platformer gameplay footage. Remarkably, Genie comprehends the mechanics and physics of these formulaic games without explicit information on button or control inputs. It streamlines the game creation process, allowing users to fashion platformer game worlds in just a few straightforward steps.
Interactive Transformation:
Genie takes a single image, be it a photograph, sketch, or AI-generated image, and promptly transforms it into a rudimentary interactive environment. This dynamic conversion responds to the user’s controls, offering a glimpse into the potential of AI in gaming.
Current Limitations:
Despite its innovative approach, Genie is still in the research phase and not a final product. The model’s training utilized ultra-low-resolution videos (160×90 pixels, 10 frames per second), resulting in brief “games” of low quality that run for only 16 seconds at 1 fps. However, these limitations are anticipated to be overcome through scaling, notes NIX Solutions.
Future Outlook:
Genie’s basic concept has been successfully tested, paving the way for substantial improvements with scaling. By employing longer videos with higher resolutions and increased processing power, Genie is expected to evolve into a more refined and sophisticated gaming AI.