A massive artificial intelligence (AI) model for navigating the physical world is the creation of scans collected from Pokémon Go players.
Niantic Inc.’s plan to construct a “Large Geospatial Model” (LGM) draws from more than 10 million scanned locations worldwide, and captures about 1 million new scans weekly through mobile players of Pokémon Go and Scaniverse. It offers a uniquely pedestrian view inaccessible to cars and street-view cameras.
“Over the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games and Scaniverse [app],” Niantic said in a blog post.
The company said it has trained more than 50 million neural networks, each representing a specific location or viewing angle. The networks compress thousands of mapping images into digital representations of physical spaces. Together, they contain over 150 trillion parameters.
What Niantic is doing marks a major departure from traditional maps that are dependent on vehicles or airborne drones to take pictures from a pedestrian point of view. Ultimately, the company intends to weave and stitch together localized models into a global system, creating a constantly evolving 3D world map. The user-generated maps are then used to teach the LGM how to recognize objects, predict spatial layouts, and infer missing details about scenes — all skills critical to spatial intelligence.
“As we move toward more scalable models, Niantic’s goal remains to lead in the development of a large geospatial model that operates wherever we can deliver novel, fun, enriching experiences to our users. And, as noted, beyond gaming Large Geospatial Models will have widespread applications, including spatial planning and design, logistics, audience engagement, and remote collaboration,” Niantic wrote in a blog post.
“The path from LLMs (Large Language Models) to LGMs is another step in AI’s evolution. As wearable devices like AR glasses become more prevalent, the world’s future operating system will depend on the blending of physical and digital realities to create a system for spatial computing that will put people at the center,” Niantic wrote.
Pokémon Go is the ideal vessel to create an augmented reality map. Released in 2016, the AR game — which has been downloaded more than 600 million times — requires smartphone users to find and catch virtual Pokémon in the real world so they can collect items at PokéStops.
In an updated note, Niantic users need to visit a specific publicly-accessible location and click to scan. “This allows Niantic to deliver new types of AR experiences for people to enjoy. Merely walking around playing our games does not train an AI model,” the company added.
“Using the data our users upload when playing games like Ingress and Pokémon Go, we built high-fidelity 3D maps of the world, which include both 3D geometry (or the shape of things) and semantic understanding (what stuff in the map is, such as the ground, sky, trees, etc.),” Niantic Chief Scientist Victor Prisacariu said in a Q&A in 2022.
For those concerned about their personal information, Niantic’s privacy policy makes it clear what it does with the location data it collects, such as as name and email address. The privacy policy outlines what is and isn’t shared with third-party vendors, but not what the company does with the data.
Niantic’s privacy policy for children features a portal where parents can manage their kid’s profile; a company spokesperson said Niantic does not sell players’ personal information.
