Google’s Genie world model can now simulate real streets with Street View

Google DeepMind has integrated Street View with its Genie world model, allowing for realistic simulations of real-world environments. This new feature, launched during the Google I/O developer conference, enables users to interact with simulated streets, adjust conditions like weather, and even explore hypothetical scenarios. This advancement aims to be beneficial for both AI agent and robotics applications, as well as for human users.

One of the key applications of this integration is in robotics training. For example, Genie can simulate rare weather events or environmental conditions, preparing robots for unpredictable real-world scenarios. Similarly, users can explore different locations under varied conditions, such as seeing New York City streets in the snow, offering a dynamic and immersive experience beyond static imagery.

Google has accumulated an extensive Street View dataset over two decades, comprising over 280 billion images across 110 countries. This vast collection of real-world data, combined with Genie's ability to simulate worlds, offers powerful potential for diverse applications, from educational experiences and gaming to advanced robotics training.

While impressive, the technology is still in its experimental phase. The simulations currently offer video game-quality visuals rather than photorealistic ones, and the models are not yet fully physics-aware. However, researchers are actively working on improving accuracy and quality, anticipating significant advancements in the coming months.