It has been revealed that various metadata, including AR data from Niantic’s flagship title, Pokémon GO, is being used to enhance the precision of delivery robots.

This news was reported by the U.S. technology media outlet MIT Technology Review. The report details how Niantic is refining its AI technology based on data accumulated through its games, thereby building more accurate location information. In particular, it was noted that the vast amount of images and location data collected during the gameplay of Niantic’s core AR titles are being utilized for AI training.

At the core of this learning process are the images captured by game users and their associated metadata. As players scan PokéStops, gyms, and various landmarks in Pokémon GO to earn rewards, the GPS coordinates, camera angles, timestamps, weather conditions, and movement patterns collected via AR features are recorded on the server. Based on this massive amount of crowd-sourced data, the system can identify the status of specific areas or locations with centimeter-level precision.

This accumulated data is being expanded into actual industrial sectors through Niantic Spatial, an AI company under Niantic. A prime example is the collaboration with the delivery robot firm Coco Robotics. Niantic Spatial stated on its official blog, "Cities are essentially obstacle courses that never stop changing. GPS signals falter amidst skyscrapers, environmental conditions shift constantly, and street closures or new pick-up points render maps obsolete in seconds." They added, "Humans can navigate this chaos with ease, but robots simply cannot."

In essence, Niantic is using the data collected through Niantic Spatial to supplement the limitations of GPS in urban environments where location accuracy is constantly fluctuating, helping robots reach their destinations more precisely. This suggests that data collected by users while playing games could be utilized for various location-based technologies beyond logistics and robotics.

However, as this news spreads, concerns regarding this method of utilization are emerging. Critics point out that while data collection is disclosed in the Terms of Service, it was not sufficiently communicated that this data would be used for commercial AI model training or robot operations. Discussions regarding user privacy are ongoing, given that precise data capable of estimating an individual's location, line of sight, and movement paths is being accumulated. Furthermore, concerns have been raised about the potential for such data to be used for military purposes amidst the increasingly volatile global climate.