• expr@programming.dev
      2·
      2 days ago

      I get it’s part of the joke, but just so people are aware (since this seems to be lost on a lot of people): there is no process, because it’s not actually thinking. It’s merely generating text that resembles natural language, with no semantic understanding. It doesn’t know what a location is, nor how to figure out where the closest McDonalds is.

      • pelespirit@sh.itjust.works
        1·
        2 days ago

        I wonder if they’re using 2 AI systems, one for location and the other for information. That would work for this scenario.

    • expr@programming.dev
      1·
      2 days ago

      Man, people really don’t know how LLMs work, do they?

      No, it doesn’t have any information whatsoever. It is merely generating text that resembles natural language, with no semantic understanding of what location is or the ability to compute such information on the fly.

      The location given by the LLM is almost certainly nowhere near where the user actually is. It’s merely spitting out a location that happened to exist in its training data.

    • unwarlikeExtortion@lemmy.ml
      4·
      2 days ago

      No one mentioned, say, GPS or a similar technology. Therefore, if you can narrow down someone’s location in any way, you have location information.

    • Zerush@lemmy.ml
      1·
      2 days ago

      Not with the public IP, this point to the used ISP server, which can be between 50-350 km away. The user IP normally isn’t visible by third party. In mobile they can localize you iv you have activated the GPS or localisation in the settings, than they can use the GPS data or the nearby transponder location (200-300 m). If it is an AI from Google, MS or Apple, they already know your direction, eg. using their maps, or by logging you.

      Simply DON’T use AI from big US corporations.