Tag

Geospatial Model

All articles tagged with #geospatial model

technology1 year ago

Pokémon Go Players Unwittingly Train AI with Their Data

Niantic, the creator of Pokémon Go, has revealed plans to use data from the game to develop a Large Geospatial Model (LGM), an AI tool designed to help machines understand and navigate real-world environments. This has led to feelings of betrayal among some players, although Niantic maintains that the data usage was disclosed in the game's terms and privacy policy. The LGM aims to enhance future augmented reality and robotic applications, but concerns about potential misuse of AI data remain.

technology1 year ago

Pokémon Go Players Unknowingly Train AI with Their Data

Niantic, the company behind Pokémon Go, has been using data collected from the game to train a large geospatial AI model, which aims to improve navigation and interaction with physical spaces for robots and other devices. This model, built from over 10 million scanned locations, is intended to enhance technologies like AR glasses and autonomous systems. While many players are indifferent to their data being used in this way, critics express concern over potential military applications of the technology.

technology1 year ago

Pokémon Go Players Fuel AI's Real-World Navigation Skills

Niantic, the developer of Pokémon Go, is using data collected from players to train a "Large Geospatial Model" (LGM), an AI designed to auto-complete real-world locations with limited information. This model, similar to a Large Language Model but for physical spaces, aims to enhance technologies like AR glasses and robotics by predicting unseen areas based on common characteristics of known locations. Niantic's Visual Positioning System (VPS) uses images from phones to map locations with high accuracy, enabling persistent digital content in the real world. While currently focused on enhancing its products, the data collection raises concerns about future AI applications.

technology1 year ago

Pokémon Go Players Unknowingly Train AI for Real-World Navigation

Niantic is developing an AI navigation system using data from Pokémon Go and its Scaniverse app, creating a "large geospatial model" that processes physical spaces through geolocated images. This model, akin to large language models, is built from over 10 million scanned locations worldwide, with users contributing around 1 million new scans weekly. The system uses first-person scans to map areas inaccessible to traditional street-view cameras, training over 50 million neural networks to digitally represent physical spaces with 150 trillion parameters.