Over just a few months, ChatGPT went from accurately answering a simple math problem 98% of the time to just 2%, study finds::ChatGPT went from answering a simple math correctly 98% of the time to just 2%, over the course of a few months.

  • xantoxis@lemmy.one
    link
    fedilink
    English
    arrow-up
    59
    arrow-down
    8
    ·
    1 year ago

    Why is “98%” supposed to sound good? We made a computer that can’t do math good

    • Dojan@lemmy.world
      link
      fedilink
      English
      arrow-up
      38
      ·
      edit-2
      1 year ago

      It’s a language model, text prediction. It doesn’t do any counting or reasoning about the preceding text, just completes it with what seems like the most logical conclusion.

      So if enough of the internet had said 1+1=12 it would repeat in kind.

      • kromem@lemmy.world
        link
        fedilink
        English
        arrow-up
        2
        ·
        1 year ago

        Not quite.

        Legal Othello board moves by themselves don’t say anything about the board size or rules.

        And yet when Harvard/MIT researchers fed them into a toy GPT model, they found that the neural network best able to predict outputting legal moves had built an internal representation of the board state and rules.

        Too many people commenting on this topic as armchair experts are confusing training with what results from the training.

        Training on completing text doesn’t mean the end result can’t understand aspects that feed into the original generation of that text, and given a fair bit of research so far, the opposite is almost certainly the case to some degree.

    • themeatbridge@lemmy.world
      link
      fedilink
      English
      arrow-up
      15
      ·
      1 year ago

      Reminds me of that West Wing moment when the President and Leo are talking about literacy.

      President Josiah Bartlet: Sweden has a 100% literacy rate, Leo. 100%! How do they do that?

      Leo McGarry: Well, maybe they don’t and they also can’t count.

    • WackyTabbacy42069@reddthat.com
      link
      fedilink
      English
      arrow-up
      9
      arrow-down
      15
      ·
      edit-2
      1 year ago

      This program was designed to emulate the biological neural net of your brain. Oftentimes we’re nowhere near that good at math just off the top of our heads (we need tools like paper and simplifying formulas). Don’t judge it too harshly for being bad at math, that wasn’t it’s purpose.

      This lil robot was trained to know facts and communicate via natural language. As far as I’ve interacted with it, it has excelled at this intended task. I think it’s a good bot

      • jocanib@lemmy.world
        link
        fedilink
        English
        arrow-up
        6
        ·
        1 year ago

        This lil robot was trained to know facts and communicate via natural language.

        Oh stop it. It does not know what a fact is. It does not understand the question you ask it nor the answer it gives you. It’s a very expensive magic 8ball. It’s worse at maths than a 1980s calculator because it does not know what maths is let alone how to do it, not because it’s somehow emulating how bad the average person is at maths. Get a grip.

      • xantoxis@lemmy.one
        link
        fedilink
        English
        arrow-up
        2
        arrow-down
        17
        ·
        1 year ago

        Bro I wasn’t looking for a technical explanation. I know how they work. We made computers worse. The thing isn’t even smart enough to say “I wasn’t designed to do math problems, perhaps we should focus on something where I can make up a bunch of research papers out of thin air?”