Google DeepMind announced at its I/O developer conference that it is connecting Street View imagery to Project Genie, its general-purpose world model capable of generating diverse, interactive environments. The integration allows users to simulate real streetscapes with unprecedented detail, including the ability to adjust weather conditions and visualize scenarios like a snow-covered New York City block or the rare glint of sunlight on London buildings. Jack Parker-Holder, a research scientist on DeepMind's open-endedness team, highlighted how the technology could help train robots deployed in cities like London, where sunny days are uncommon, so they won't be caught off guard when sunlight actually appears.
The company brings two decades of data collection to this project. Google's fleet of camera-equipped vehicles and backpack-mounted collectors have amassed over 280 billion images spanning 110 countries across all seven continents. Parker-Holder noted that this vast repository of real-world imagery combined with Genie's simulation capabilities represents a powerful tool for both research and consumer applications.
Genie 3 launched for research preview in August 2025 and became available to Google AI Ultra subscribers in the United States this past January, enabling users to create interactive game environments from text prompts or images. Google has positioned the technology for educational experiences, gaming, and robotics training applications.
The system is already proving valuable in autonomous vehicle development. Waymo, Google's self-driving car subsidiary, has been using Genie 3 to power its simulators, training vehicles on extraordinarily rare events such as tornadoes or unexpected animal encounters. Incorporating Street View data into these training scenarios could significantly expand the range of conditions Waymo's vehicles can prepare for before encountering them on actual roads.