• MrScottyTay@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    6
    ·
    4 hours ago

    Usually with decomps you want people to still understand the resulting code, it’s not just about it “just working”. Recomps ideally should be an exploratory and learning experience and not just a case of “but I want X to run on Y”

    • brsrklf@jlai.lu
      link
      fedilink
      English
      arrow-up
      1
      ·
      7 minutes ago

      That’s almost the same argument from mathematicians complaining about AI “solving” maths challenges.

      AI is shitting out a result they now have to check, but it’s depriving them from the only actually useful part of those problems, which is creating and developing new tools to understand how to even get to that solution. Tools that are added to their field’s toolbox and may be used elsewhere.

      Without that process, cool, you’ve been served an answer to a completely abstract puzzle that never had any purpose. No surprise that they don’t even care about doing the tedious part of checking the answer.

      Sometimes, it is about the journey.