The developer behind Teach My Little Sister How to Drive, a driving instruction sim where you instruct a generative-AI-powered, dynamic NPC, has been spending $1,000 dollars a day to keep the demo working.

  • lepinkainen@lemmy.world
    link
    fedilink
    English
    arrow-up
    20
    arrow-down
    1
    ·
    1 day ago

    AI powered games don’t work.

    We did a super cool project for a game jam last year. Cost like 50c per game for us.

    Cool for the 100 or so players who tried it, but there’s no way to make a profit from it.

    Local models maybe.

    • Rhaedas@fedia.io
      link
      fedilink
      arrow-up
      9
      arrow-down
      1
      ·
      1 day ago

      For just simple and guided actions and dialogue, local would be fine for this. Or better, have AI design the large database of what to say or do when a player does things. No need to reinvent what players are going to duplicate. The token cost was worse when you consider all the same stuff being prompted over and over.

      • Ledivin@lemmy.world
        link
        fedilink
        English
        arrow-up
        6
        ·
        1 day ago

        If they’re actually hitting a cloud LLM, there’s no good reason not to wrap that and memoize the results… it would probably be hard to generalize the inputs (it could just be the whole game state for all we know 🤷‍♂️) but youd build up that database of all the most common input/outputs in real-time

        • Rhaedas@fedia.io
          link
          fedilink
          arrow-up
          4
          ·
          1 day ago

          Exactly, that’s just smart programming (maybe that’s one of the issues… but I can’t say that without evidence).

          I saw someone years ago develop that concept for a text to speech idea for NPC responses, and for the same reasons. Why repay for something when you can just capture the sound file for each new phrase? The cost is in the creation and addition of things, not in the actual running of every person’s game (or app in this case).

          • TrippaSnippa@aussie.zone
            link
            fedilink
            English
            arrow-up
            3
            ·
            23 hours ago

            Exactly, that’s just smart programming (maybe that’s one of the issues… but I can’t say that without evidence).

            That would require you to actually understand what your code is doing…

    • HeHoXa@lemmy.zip
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      1
      ·
      edit-2
      1 day ago

      In case it might be interesting info, maybe want to check out WebGPU, WebLLM, Wllama, or Transformers.js. You can write web apps that run an llm on the client’s browser.

      It requires permission and compatibility and isn’t… it might be tough to really make accessible to the average user.

      But it’s an option and kinda cool.