• Canaconda@lemmy.ca
    link
    fedilink
    arrow-up
    56
    ·
    edit-2
    29 天前

    As someone who grew up using telephones… the last part is the most human chat bot interaction I’ve ever seen.

  • groats_survivor@lemmy.world
    link
    fedilink
    arrow-up
    15
    ·
    edit-2
    28 天前

    Me to AI: alright, I’m about to send you a two part message. Do not respond to the first message.

    AI: Gotcha! I won’t respond

    • davidgro@lemmy.world
      link
      fedilink
      arrow-up
      9
      ·
      28 天前

      That could reasonably be interpreted as you haven’t sent the first part yet.

      But I assume it still responds like that when you do.

    • Victor@lemmy.world
      link
      fedilink
      arrow-up
      1
      arrow-down
      1
      ·
      28 天前

      How would it know not to respond to the first part without processing it first? The request makes no sense.

      Like telling a human, hey, don’t listen to this first part! Also don’t think about elephants!

  • thebestaquaman@lemmy.world
    link
    fedilink
    arrow-up
    12
    ·
    29 天前

    I would think that, since it’s been recognised that these messages are costing a lot of energy (== money) to process, companies would at some point add a simple <if input == “thanks”> type filter to catch a solid portion of them. Why haven’t they?

    • Hoimo@ani.social
      link
      fedilink
      arrow-up
      1
      ·
      28 天前

      It won’t be as simple as that and the engineers who work on these systems can only think in terms of LLM and text classification, so they’d run your message through a classifier and end the conversation if it returns a “goodbye or thanks” score above 0.8, saving exactly 0 compute power.

      • thebestaquaman@lemmy.world
        link
        fedilink
        arrow-up
        2
        arrow-down
        1
        ·
        28 天前

        I mean, even if we resort to using a neural network for checking “is the conversation finished?” That hyper-specialised NN would likely be orders of magnitude cheaper to run than your standard LLM, so you could likely save quite a bit of power/money by using it to filter for the actual LLM, no?

  • Pratai@lemmy.world
    link
    fedilink
    arrow-up
    7
    arrow-down
    3
    ·
    edit-2
    28 天前

    Who in their right mind would want to talk to AI begin with? I don’t think I’ll ever understand why someone would want to do this outside of a single instance of curiosity.

  • SUPER SAIYAN@lemmy.world
    link
    fedilink
    arrow-up
    1
    arrow-down
    2
    ·
    edit-2
    28 天前

    I dont understand. People argued a book with me here for using AI quoting usage of large amounts of water and noise around the data server areas. But upvoted this? And how is this a meme? Its a long screenshot.

  • bleistift2@sopuli.xyz
    link
    fedilink
    English
    arrow-up
    8
    arrow-down
    14
    ·
    29 天前

    People on lemmy bitching about how bad AI is for the environment, yet this gets upvotes.

    Doublethink 10/10

      • NihilsineNefas@slrpnk.net
        link
        fedilink
        arrow-up
        1
        ·
        28 天前

        Wouldn’t that be the gpt user that posted this screenshot? Someone making some kind of “I’m getting back at those pesky AI companies by costing them money” … by pressing the "push to run a machine that burns the planet’ button?

        Or worse yet, it’s an accelerationist willingly pushing the button.