What to know

  • Niantic uses Pokémon Go player data to build a Large Geospatial Model for real-world AI navigation.
  • Players submit 1 million new scans weekly, contributing to over 10 million scanned locations worldwide.
  • The AI model helps computers understand and navigate physical spaces, useful for AR glasses and robots.

Niantic now uses gaming data from Pokémon Go players to create an AI navigation system. The company develops a “Large Geospatial Model” that processes physical spaces similar to how ChatGPT processes language.

Players have scanned over 10 million locations worldwide, adding 1 million new scans every week. Each scan includes hundreds of images showing locations from a walking perspective.

The data differs from Google Street View because it shows places from a pedestrian view rather than from cars. This gives a better picture of areas vehicles cannot reach.

The system expands on Niantic’s Visual Positioning System that places virtual items in exact real-world spots. Players test this through Pokémon Playgrounds, placing Pokémon for others to find.

The technology opens new ways to interact with surroundings. Architects could see designs in real locations, robots could move through spaces better, and AR glasses could show digital information in the real world.

Players generate valuable data while playing the game. Their scans and movements help train AI systems to understand physical spaces better.