Data Collected From Pokemon Go Players To Build ChatGPT-esque "Spatial Intelligence"

Remember Pokemon Go? It is back in the headlines, but for a very different reason this time - using your data.

For those who need a recap: Pokemon Go was released in 2016 and became an instant hit. It is a smartphone game based on virtual reality (VR) that uses a phone camera to scan the surrounding environment, using which the players can catch virtual Pokemon hiding in the real world. The game combines GPS tracking with real-time gameplay.

Niantic, the company behind Pokemon Go, posted in a blog earlier this week, revealing that they were building a "Large Geospatial Model (LGM)" and using data collected over the last eight years from millions of Pokemon Go users.

The game gathered a serious following at the time of its release, and it is now clear that millions of users played it.

LGM is an AI model based on Large Language Models (LLMs), the same fundamental architecture used by the popular chatbot ChatGPT. Just like ChatGPT processes large amounts of textual data, LGM aims to process original scans of real-world locations fed by users worldwide to design "Spatial Intelligence" as we have never seen before.

"Today we have 10 million scanned locations around the world, and over 1 million of those are activated and available for use with our VPS service," Niantic has said via the blog. "We receive about 1 million fresh scans each week, each containing hundreds of discrete images."

The company is betting on LGM to be the key technological enabler for augmented reality (AR) glasses, robotics, content creation, and autonomous systems. According to the company, it would be a tool that allows computers not only to perceive and understand their surroundings but also interact with them in new ways.

Pokemon Go creator is entering the already hotly contested AI product market. But Niantic has a considerable edge over the likes of Google Street View and autonomous vehicle GPS systems in the likes of Tesla models - the unique nature of its data given that the scans are pedestrian-driven and not roads mapped through cars. After all, cars can't reach all the tight corners and the nooks of the world, but you can while trying to catch that elusive Pikachu Pokemon.

The potential application of this AI system is not yet clearly known. Such a model can be used to teach robots to deliver food to your doorsteps, but could also end up in the hands of some bad actors. Contemplation aside, one thing is for sure- no one in 2016 could have imagined their fun little Pokemon game to have been a data collector for a gigantic AI model that seeks to capture the entirety of the earth.