Summary
[UPDATE: Niantic has provided clarification in an update to its blog post. The update is below.]
“We use player-contributed scans of public real-world locations to help build our Large Geospatial Model. This scanning feature is completely optional – people have 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.”
Pokemon GO developer Niantic has revealed that data scraped from its popular AR games is being used to build a new geospatial AI model that will allow neural-network-powered machines and interfaces to better interact with physical spaces across the globe.Pokemon GO, an AR spin-off of the popular franchise, uses players' smartphone cameras coupled with neural networks to place virtual objects in real spaces.
Niantic’s founder, John Hanke, led Google’s Geo division, which created Google Maps, Earth, and Street View, before forming Niantic in 2010. Prior to that, his company Keyhole created an interactive interface for live mapping conflicts in the Iraq War. Niantic is best known for itssuite of popular augmented reality games, includingPokemon GO,Ingress Prime,Pikmin Bloom, and more.
As stated ina company blog post, Niantic’s newest venture is the construction of a “Large Geospatial Model” or LGM, which would combine the more than 50 million neural networks the company has trained into a single model of geographic locations and their spatial configuration. The company cites “AR glasses…, robotics, content creation and autonomous systems” as key applications of this technology.
Niantic is betting onthe rise of wearable XR technologyand autonomous machines to make spatial computing the predominant operating system of the near future. The company’s current geospatial models are informed by millions of scans generated by users of its AR games and Scaniverse app. These assets have been combined to create its Visual Positioning System, which allows for the placement of persistent virtual objects in real space with a high degree of accuracy. The Large Geospatial Model is the next evolution of that system, combining the millions of neural networks generated by the VPS into a single entity capable of extrapolating unscanned location information out of local spatial and positioning data.
WhilePokemon GOusers will likely not be surprised that thecorporation behind a popular game is scraping their datain pursuit of its larger business goals, the implications of a large geospatial model are far-reaching.
As large geospatial models continue to develop, it seems inevitable that Niantic or some other entity will explore the entire spectrum of the new tech’s potential integrations. It remains to be seen what regulatory or ethical boundaries are established for the deployment of the striking advancement, and whatpros and cons consumers might encounter with this latest AI evolution.
