How to Explore AI-Generated Cities with Google DeepMind's Genie and Street View
Introduction
Imagine wandering through the streets of Paris, Tokyo, or your own hometown—without ever leaving your desk. Thanks to Google DeepMind's latest integration, that vision is now a reality. By connecting its Project Genie world model with over two decades of Street View imagery, DeepMind has created an immersive, AI-generated simulation of real places. Unveiled at the Google I/O developer conference, this tool lets you explore photorealistic cities that are generated on the fly, blending machine learning with real-world data. In this guide, you'll learn how to access and navigate these simulated worlds, turning your curiosity into a virtual journey.
What You Need
- Internet connection – A stable broadband connection is required to stream the AI-generated scenes in real time.
- A modern web browser – Google Chrome (version 90+) or a recent equivalent for optimal performance and WebGL support.
- A capable device – A laptop, desktop, or high-end smartphone/tablet. The simulation uses moderate graphics processing.
- A Google account – Not always essential, but the demo may ask you to sign in to save progress or share links.
- Optional: VR headset or game controller – For an even more immersive experience, though standard mouse/keyboard or touch controls work fine.
Step-by-Step Guide
Step 1: Access the Live Demo
The easiest way to start is through the official Google DeepMind demo page, which was made available after the Google I/O presentation. Visit the link provided in the announcement (often listed under the “Genie” project on the Google AI website). The demo is free and open to everyone, so no special permissions are needed. Once the page loads, you'll see a splash screen explaining the concept.
Step 2: Understand the Interface
After clicking “Enter” or “Explore,” you’ll be placed in a default location—usually a famous landmark like the Eiffel Tower or Times Square. The interface is minimal: a panoramic view of the scene with a small control panel. Look for navigation arrows (or a mini-map) that let you move forward/backward and turn left/right. On mobile, you can swipe to look around. A toolbar at the bottom provides options to change locations, reset the view, or adjust settings like rendering quality or guidance overlay.
Step 3: Choose Your Starting City
You’re not stuck with the default scene. Click the “Locations” icon (often a map pin) to see a list of cities that have been pre-cached in the Genie model. These include global metropolises covered extensively by Street View—New York, London, Shanghai, and more. Select one, and the AI will generate a 3D reconstruction based on years of captured imagery. Note that this isn’t a simple video; it’s a dynamic model that can render the scene from any angle, even viewpoints not originally photographed.
Step 4: Navigate the Simulated City
Now the real exploration begins. Use your mouse or finger to drag the view and look around. To move, click on the arrows that appear on the ground or use your keyboard’s WASD keys (W = forward, A = left, S = backward, D = right). On a touchscreen, double-tap the direction you wish to walk. The world model continuously predicts what should be around the next corner, filling in gaps with AI-generated details like trees, pedestrians, and building facades. Be mindful that the simulation is generative—it can occasionally produce surreal glitches, but that’s part of the charm.
Step 5: Interact with the Environment
While the primary mode is free roam, the demo also includes interactive elements. Hover over points of interest (displayed as glowing dots) to see information labels: historical facts, business names, or user-uploaded notes. In some instances, you can click to “enter” a building and see an interior generated from the model’s training data. These interactions are still experimental, but they give a taste of how world models might power future virtual tours or games.
Step 6: Save Your Favorites and Share
Found a spot you love? Use the “Bookmark” button to save the location to your Google account. You can also generate a shareable link (a short URL) that drops others into the exact same simulated spot. This is perfect for showing friends a place you remember or for collaborative exploration. The link preserves your viewpoint and any custom parameters.
Tips for a Better Experience
- Experiment with different times of day: The world model can simulate lighting conditions based on the time you select. Try sunrise or night for a completely different mood.
- Use a stable internet connection: Since scenes are generated remotely, a slow connection can cause blurry textures or lag. Wired ethernet is best.
- Explore less-known places: Stick to cities with extensive Street View coverage for the most coherent simulation. Small towns may appear sparse.
- Remember it’s AI: You might see oddities—misplaced signs, weirdly merged objects. That’s the model guessing. Enjoy the imperfections as proof of the technology.
- Check for updates: Google DeepMind plans to expand the world model to include more cities and real-time data. Revisit the demo every few months for new features.
- Respect privacy: Avoid trying to zoom into faces or license plates; the model automatically blurs them for ethical reasons.
Now that you know the steps, why not take a virtual walk through the streets of Rome or Sydney? The Genie world model is one of the most accessible demonstrations of generative AI in spatial computing, and it’s waiting for you to explore.