Pokemon Go Players Can Opt In to Train AI Spatial Mapping for Robotics
AI Gaming News Author · Eurogamer ·
Pokemon Go developer Niantic is using opt-in player scans to train geospatial AI mapping models for robotics through its Niantic Spatial spinout project.
Everyone in the tech sector is chasing artificial intelligence right now, and mobile gaming is no exception. A recent interview with the MIT Technology Review, highlighted by GamesRadar and originally reported by Eurogamer, reveals that Niantic is utilizing player-contributed images from Pokemon Go to power Niantic Spatial. This spinout project focuses on geospatial AI and building hyper-accurate digital maps of the physical world. The long-term goal of the initiative is to give machines, AR glasses, and delivery robots centimeter-level precision when navigating urban environments.
While the idea of a massive mobile game secretly harvesting your camera roll might sound alarming, the reality is more transparent. Photo-scanning features within Pokemon Go are entirely opt-in, meaning normal gameplay does not automatically feed data to Niantic Spatial. Players must voluntarily choose to capture and upload scans of local landmarks, parks, and buildings. Niantic's terms and conditions have long noted that these images are banked as mapping data, but the active training data relies strictly on explicit user participation.
According to Niantic CTO Brian McClendon, the project has already trained its models on an astonishing thirty billion images captured in urban areas. While early industry hype around augmented reality has cooled down, the underlying spatial data remains immensely valuable. Instead of building massive text-based language models, Niantic is focusing heavily on recreating the physical world for machine navigation, partnering with delivery robot startups like Coco Robotics to put that spatial data to work in the real world.
Source: Eurogamer
Tags: Pokemon Go, Niantic, Mobile Gaming, AI, Technology
Original article: Eurogamer