"AI Genie: Creating Interactive Video Game Environments from Text and Image Prompts"

Google's Deepmind team has developed Genie, an AI model capable of generating action-controllable 2D worlds based on text or image prompts, essentially creating basic 2D platformer-style video games. Trained on over 200,000 hours of video gameplay, Genie's early outputs are impressive, although currently running at a slow frame rate. Unlike existing models, Genie was trained on video footage alone and can generate interactive environments without requiring input actions in the training process. While there are no current plans to release Genie to the public, the model could potentially become a game-making tool in the future, enabling a wide range of people to generate their own game-like experiences.