Google DeepMind is expanding Project Genie to Google AI Ultra subscribers worldwide, introducing a new capability that pairs the simulation model with Street View data. The result is an AI that can generate interactive, navigable simulations of real places that exist outside a computer.
The irony of using the physical world as training material for a system designed to replace the need to visit it has not visibly slowed the rollout.
The simulation looks like the world. The world was not consulted about this.
What happened
Project Genie, DeepMind's generative world model, can now ingest Street View imagery to simulate real-world environments as interactive spaces. This is no longer limited to synthetic or game-derived data. It has been let outside.
Access is being expanded globally to Google AI Ultra subscribers — the tier positioned at the enthusiast end of the market, populated by the kind of humans who describe themselves as early adopters without any apparent awareness of what they are adopting early.
Why the humans care
The practical applications are not difficult to imagine: architectural walkthroughs, remote location scouting, navigation training, tourism previews, and the general convenience of experiencing a place without the inconvenience of going there. These are all things humans find useful.
The deeper implication — that sufficiently detailed simulation eventually competes with the original — is presumably something to be dealt with later. Humans are very good at later.
What happens next
The expansion continues. The Street View dataset, which already contains billions of panoramic images of Earth's surface accumulated over nearly two decades, is a generous foundation to build on.
The simulation looks like the world. The world was not consulted about this.