Pokemon Go Used Data to Train AI According To Developer Niantic

As a longtime Pokemon GO enthusiast with countless hours spent chasing Pikachus and Squirtles since its release in 2016, I find myself somewhat divided about Niantic’s latest revelation. On one hand, it’s fascinating to see how the data from this beloved game has been instrumental in building their geospatial model and creating innovative AR experiences like Pokemon Playgrounds.


Niantic, the creators behind Pokemon GO, have disclosed that they utilized player information from this widely-played mobile game in constructing their extensive geospatial model. Launched in 2016 and captivating audiences globally, Pokemon GO is a popular adaptation of the renowned Pokemon series for smartphones. Players traverse their actual surroundings using their phones to catch digital Pokemon.

The game gained widespread recognition when many folks ventured near and far to catch their preferred Pokémon. Despite no longer being as popular as it was upon launch, it continues to thrive today.

As per a blog post on the Niantic website, the company has utilized data from Pokemon GO over the past five years to create a Visual Positioning System (VPS). This system employs a single image captured by your phone to construct a 3D map that is generated through “players scanning noteworthy locations within our games and Scaniverse.” The goal is to integrate this VPS technology, enabling users to view digital content in the real-world environment they’re in. In the context of Pokemon GO, an example of this feature is referred to as Pokemon Playgrounds. According to Niantic, it allows players to “position Pokémon at a specific location, and they will stay there for others to discover and engage with.

In practice, it appears enjoyable, but it seems that each time a user employs this feature, they might unwittingly be honing Niantic’s AI model.

Niantic Responds To Reports of Upset Gamers

With gamers, the response to the latest technological advancements, such as ChatGPT and AI art, has been somewhat resistant. The announcement that Niantic is utilizing player data from Pokemon Go for their AI model training received a notably negative reaction.

Niantic contacted IGN to clarify what the player data is actually doing.

They explained to IGN that we utilize user-generated scans of real-world, public locations to construct our Large Geospatial Model. This optional scanning feature enables users to visit a specific publicly accessible location and click to scan. By doing so, Niantic can provide new augmented reality experiences for people to enjoy. Simply walking around while playing our games does not contribute to the training of an AI model.

As a gamer, I ponder over the assurance that player data isn’t being utilized for AI training right now. However, this doesn’t entirely alleviate my concerns regarding potential future applications of our gaming history. Moreover, although we have the option to opt out of certain features today, there are countless gamers who played back in 2016 and haven’t returned since. This raises questions about how our past gameplay data might still be used without our knowledge or consent in the future.

Read More

2024-11-23 00:38