Niantic builds AI navigation sysem using Pokémon Go player data

The gaming company Niantic is leveraging player-generated data from Pokémon Go and other apps to develop an artificial intelligence system for real-world navigation, marking a novel approach to AI training data collection through mobile gaming.

Project overview and scope: Niantic announced its development of a “large geospatial model” (LGM) that will process physical spaces using geolocated images collected through its gaming applications.

• The system builds upon Niantic’s Visual Positioning System (VPS), which uses phone camera images to determine position and orientation within 3D mapped environments
• The company has accumulated data from over 10 million scanned locations globally
• Users contribute approximately 1 million new scans weekly through Pokémon Go and the Scaniverse app

Technical infrastructure: The AI system represents a massive computational undertaking, with extensive neural network deployment and parameter optimization.

• Niantic has trained more than 50 million neural networks, each dedicated to specific locations or viewing angles
• The combined networks contain over 150 trillion parameters that help recognize and interpret physical locations
• Multiple networks can collaborate to map single locations, with plans to integrate knowledge into a comprehensive model

Unique data collection approach: The system’s strength lies in its ground-level perspective of spaces typically inaccessible to traditional mapping methods.

• Data comes from pedestrian-level scans, capturing areas that vehicles and street-view cameras cannot access
• The system can recognize common characteristics across similar structures, such as churches, even from previously unseen angles
• Recent features like Pokémon Playgrounds demonstrate practical applications, allowing precise virtual object placement

Potential applications: Niantic envisions broad implementation possibilities beyond gaming.

• The technology could support various augmented reality products
• Additional applications include robotics, autonomous systems, spatial planning, and logistics
• Remote collaboration tools could benefit from the spatial understanding capabilities

Privacy and user awareness: The data collection has sparked discussions about transparency and user consent.

• Reddit discussions suggest many players were aware of potential data usage beyond gaming
• The collection process is covered by Pokémon Go’s terms of service
• Player reactions have been mixed but generally unsurprised by the revelation

Forward-looking implications: This initiative represents a significant shift in how companies might leverage gaming platforms for AI development, while raising important questions about data collection transparency and user consent in gaming environments.