This initiative marks a significant shift in AI training data collection, as it primarily relies on user-generated data from a mobile gaming platform rather than traditional sources.

Over the past five years, Niantic has refined its VPS, which utilizes a single image from a smartphone to determine a user's position and orientation based on a 3D map created from user-generated scans.

The LGM aims to advance various fields, including augmented reality glasses, robotics, and autonomous systems, potentially creating new revenue streams for Niantic beyond in-game purchases.

Niantic's recent blog post highlighted the connection between the LGM and the VPS, emphasizing its potential to enhance understanding of complex three-dimensional spaces for future applications in augmented reality and robotics.

Critics point to a lack of transparency in Niantic's data policies, particularly concerning how location data is utilized, which has led to heightened concerns about user privacy.

However, the data collected from millions of users raises ethical concerns, as much of it was gathered without clear disclosure or consent, particularly regarding its use for training AI models.

Since its inception, Niantic has amassed a database of 10 million scanned locations globally, with one million of these processed for its Visual Positioning System (VPS), allowing for the integration of persistent digital content into real-world environments.

As excitement builds around these technologies, there are increasing worries about data privacy and the implications of using such information to train AI, especially as applications become more complex.

While some players express dissatisfaction with Niantic's data collection practices, others acknowledge the company's transparency in using player data to improve their services.

Niantic's model is designed to synthesize local data to create comprehensive 3D images of locations, allowing it to extrapolate missing details based on existing knowledge of architectural patterns.

Niantic, the company behind popular augmented reality games like Pokémon Go, is developing an innovative AI model known as the Large Geospatial Model (LGM) to enhance navigation in the physical world using data collected from its players.

The announcement of the LGM, made in November 2024, represents Niantic's first clear communication regarding the use of data collected from its games, aiming to improve how computers perceive and navigate the physical world.