Pokémon Go Players Are Unknowingly Training AI for Reality

Niantic, the developer behind Pokémon Go, is developing a "Large Geospatial Model" that trains AI to auto-complete real-world locations based on limited information. This initiative utilizes data collected from Pokémon Go players, creating a system that understands physical space by analyzing numerous real-world characteristics. By gathering extensive location data, Niantic aims to enhance augmented reality (AR) applications, including a Visual Positioning System (VPS) that allows users to interact with digital content integrated into real-world environments. The project currently has over 10 million scanned locations, with significant concerns regarding data privacy and the future applications of trained AI models.

Niantic's Large Geospatial Model uses extensive data collected from players to train AI on recognizing and completing real-world locations. It processes common characteristics of various locations, allowing for accurate representations, even of places it has never seen, which can enhance AR experiences.

Pokémon Go, released in July 2016, revolutionized mobile gaming by merging the digital and real worlds. Players use their smartphones to catch virtual creatures, known as Pokémon, while exploring their surroundings. The game's popularity has contributed to Niantic's collection of diverse geospatial data, making it a crucial part of their new AI model development.