Updated · The Keyword | Google Product and Technology News · May 19
Project Genie Adds Street View Grounding, Rolling Out to $200 AI Ultra Subscribers Globally
Updated
Updated · The Keyword | Google Product and Technology News · May 19
Project Genie Adds Street View Grounding, Rolling Out to $200 AI Ultra Subscribers Globally
9 articles · Updated · The Keyword | Google Product and Technology News · May 19
Google launched a new Street View grounding feature in Project Genie, letting users build interactive AI worlds anchored to real places and starting now with U.S. locations.
The upgrade ties Genie’s generative world model to Street View imagery through Maps Imagery Grounding, aiming to help AI agents and robots navigate more realistic virtual environments.
Eligible Google AI Ultra subscribers — priced at $200 and aged 18+ — are getting the feature in a gradual global rollout, while Google says the Labs prototype still has accuracy limitations.
Project Genie, already used in research and Waymo road simulations, is being positioned beyond entertainment toward robotics and other real-world AI training use cases.
After major AI failures in 2025, can virtual training make robots truly safe for our world?
Is Google building a new reality for everyone, or is it building a world only it can control?
How will AI learn real-world physics when its simulated worlds don't obey them?
Genie 3’s Street View Breakthrough: Google DeepMind’s Leap Toward Realistic, Interactive World Models
Overview
On May 19, 2026, Google DeepMind announced the integration of Google Maps Street View imagery into its Genie 3 world model. This breakthrough enhances Genie 3’s ability to generate interactive, physically consistent environments, allowing for more sustained play and interaction within its simulated worlds—lasting several minutes, which is a major improvement over earlier models. Genie 3 is designed to create dynamic environments, opening new possibilities for simulation and AI training. Google is already using Genie 3-generated worlds to train its SIMA agent, helping AI systems practice tasks that can be transferred to real-world scenarios.