Skip to main navigation Skip to search Skip to main content

The AI model built on billions of Pokemon GO images

Press/Media: Expert Comment

Description

In this podcast, I discuss how Pokémon Go has contributed to large-scale spatial data collection through gamified user interaction. I explain how player activities, including scanning landmarks and navigating real environments, have generated extensive visual datasets that are now being used to train artificial intelligence systems.

I situate this within a broader shift in AI, moving from language-based models toward systems that interpret and navigate the physical world, with applications in robotics, augmented reality, and spatial computing. I also reflect on the role of gamification in enabling data generation at scale and consider emerging issues around data ownership, transparency, and the repurposing of user-generated data.

Period23 Mar 2026

Media contributions

2

Media contributions

  • TitleThe AI model built on billions of Pokemon GO images
    Degree of recognitionNational
    Media name/outletThe Daily Aus
    Media typeWeb
    Country/TerritoryAustralia
    Date23/03/26
    DescriptionIn this podcast, I discuss the emerging relationship between gamification, large-scale spatial data collection, and the development of artificial intelligence, using Pokémon Go as a contemporary case study. I reflect on how location-based games have generated extensive visual datasets of real-world environments through sustained user participation over time.

    I highlight how player interactions, including the scanning of landmarks and navigation of urban spaces, have contributed to the creation of large-scale spatial datasets that are now being applied beyond entertainment contexts. In the discussion, I explain how these datasets are being used to train artificial intelligence systems capable of understanding and navigating physical environments, with emerging applications in robotics, augmented reality, and spatial computing.

    I situate this within a broader shift in artificial intelligence, moving from language-based models trained on internet text toward systems that interpret and interact with the physical world. I also explore the role of gamification as a mechanism for large-scale data generation, demonstrating how user engagement and play can underpin the creation of AI training infrastructure.

    The podcast further considers key issues related to data ownership, transparency, and the implications of repurposing user-generated data for applications beyond their original intent. Through this contribution, I position Pokémon Go not only as a cultural and technological phenomenon, but as an early example of how everyday digital interactions can support the development of next-generation artificial intelligence systems.
    Producer/AuthorOrla Maher
    URLhttps://open.spotify.com/episode/2SpRqzKqufxkQy3RAe154h?si=jwR3UtTIS1i7DhVAKsZzeA
    PersonsJames Birt
  • TitleThe AI model built on billions of Pokemon Go images
    Degree of recognitionNational
    Media name/outletThe Daily Aus
    Media typeWeb
    Country/TerritoryAustralia
    Date23/03/26
    DescriptionIn this podcast, I discuss the emerging relationship between gamification, large-scale spatial data collection, and the development of artificial intelligence, using Pokémon Go as a contemporary case study. I reflect on how location-based games have generated extensive visual datasets of real-world environments through sustained user participation over time. I highlight how player interactions, including the scanning of landmarks and navigation of urban spaces, have contributed to the creation of large-scale spatial datasets that are now being applied beyond entertainment contexts. In the discussion, I explain how these datasets are being used to train artificial intelligence systems capable of understanding and navigating physical environments, with emerging applications in robotics, augmented reality, and spatial computing. I situate this within a broader shift in artificial intelligence, moving from language-based models trained on internet text toward systems that interpret and interact with the physical world. I also explore the role of gamification as a mechanism for large-scale data generation, demonstrating how user engagement and play can underpin the creation of AI training infrastructure. The podcast further considers key issues related to data ownership, transparency, and the implications of repurposing user-generated data for applications beyond their original intent. Through this contribution, I position Pokémon Go not only as a cultural and technological phenomenon, but as an early example of how everyday digital interactions can support the development of next-generation artificial intelligence systems.
    URLhttps://youtu.be/S5C-4SLg8yA?si=c8ggJtZCFq0mEzCW
    PersonsJames Birt