Comment on As A.I. Makes Strides in Mathematics, Mathematicians Urge Caution

• XLE@piefed.social ⁨1⁩ ⁨week⁩ ago

  It’s apparently able to disassemble words to run through its own parser now, but… Wow that twist ending. Trillions of dollars

  Image

  source

  • Korhaka@sopuli.xyz ⁨1⁩ ⁨week⁩ ago

    I was asking it about growing stuff and it told me to plant seeds for a plant that is sterile and has to be propagated by taking cuttings. So I decided to disregard all information it told me on caring for plants because fuck knows it just made it up. Who knows how much is accurate?

    Even worse is a lot of websites are AI generated now too.

    source

    • Tolookah@discuss.tchncs.de ⁨1⁩ ⁨week⁩ ago

      The ai generated websites are the absolute worst if you want information, and at this point, they are everywhere.

      source

      • Korhaka@sopuli.xyz ⁨1⁩ ⁨week⁩ ago

        Sadly SEO seems to be one thing that the LLMs are moderately good at. That or they just churn out so much crap that it floods results with some being good enough to get to the top.

        source
  • orca@orcas.enjoying.yachts ⁨1⁩ ⁨week⁩ ago

    I made Claude work around this by having it write a Python script to use instead that split the letters and provided a count. A script I could’ve written myself, for a task that’s completely pointless.

    source

    • i_am_not_a_robot@discuss.tchncs.de ⁨1⁩ ⁨week⁩ ago

      This is the reason why conventional AI can’t make strides in mathematics. The LLM is a statistical model that generates human text. The training data is the output text divorced from the process that generates it in humans. If you feed an LLM lots of mathematical text, it will confidently produce mathematical texts of its own, but the LLM can’t add two numbers together, and it doesn’t know that it can’t add two numbers together. It can be trained to use a calculator, but humans don’t interrupt their writing to say “And now I’m using my calculator to determine the value of 1 + 1” so the LLM is just going to draw upon its training data to predict that "1 + 1 = " is followed by “2” or maybe “3 (for large values of 1).” Maybe someday it will learn that “9 + 10 = 21.”

      source