Google DeepMind previews Genie 4, an AI for live interactive worlds

Google DeepMind has introduced Genie 4, a research-preview AI system designed to generate a world that keeps responding as someone explores it, rather than a fixed video clip. Users can navigate with standard controls and alter a scene while it is running with plain-language instructions—changing weather, moving the camera, or adding people and objects—while the model also creates matching ambient sound, music, and dialogue. That makes the idea more useful than a visual demo: a convincing, controllable simulation could eventually provide inexpensive training grounds for robots and AI agents, as well as new tools for games and filmmaking. For now, those uses remain prospective, and Google is limiting access to trusted testers.