• FatVegan@leminal.space
    0·
    7 hours ago

    I have only ised chatbots a handful of times, and only if i needed to 3d print something and let it write a simple code. If it works, great, but the fucking thing doesn’t understand when it doesn’t. Suddenly you have floating parts, or parts that otherwise make no sense. “Yes, you are right, i fixed it, now it should work.” No it doesn’t, you fucked it up more. “The part should just fit on your print bed if you angle it right.” What do you mean? It’s 3 times the size. “Yes, you are right, it’s too big to fit on the print bed.”

    Anf people call it actually intelligent. It absolutely blows me away. I’m convinced these people’s brains got fried in the process, or the just enjoy the fact that someone or something keeps telling them that they are right over and over again.

    • MBech@feddit.dk
      0·
      5 hours ago

      I think it’s a matter of worktype. If people only use it for something where they don’t/can’t verify the result, like a search engine, they’ll assume it’s always correct, but if you use it for something where you don’t have a choice verifying it, like your example which is very easily visually checked, you get to see everything it fucks up.

      It’s also about workload. Give it something small that any idiot could do with a few google searches, like putting together a formula in Excel, it’ll likely do it without fucking it up 99% of the time. But something more complex with more parametres, will have a higher chance to fail spectacularly.

    • iocase@lemmy.zip
      0·
      17 hours ago

      “Yes, i was wrong. You don’t actually need to remove the engine to get the battery out. Let’s put the engine back in and i’ll show you how to remove the battery”

      Although i wouldn’t blame AI for a mistake like this… by now blindly trusting them is asking for trouble…

      For those that do, All quiet on the frontal lobe