Login
You're viewing the discuss.systems public feed.
  • Oct 12, 2026, 8:24 PM

    "AI will solve the simple problems so humans can focus on the hard ones" is such a ridiculous fail at understanding anything about how humans acquire skills and knowledge.

    Humans HAVE TO solve simple problems and then advanced problems in order to learn how to solve the really hard ones. This is not optional.

    Good luck solving any difficult problems in the future with a generation of people who spent their entire education typing every problem into a chatbot's prompt box.

    💬 6🔄 107⭐ 103

Replies

  • Oct 12, 2026, 8:27 PM

    The only way someone can seriously make such a nonsense claim is, if they don't have any idea what they're talking about, probably because they never had to learn a difficult skill in their lives, or if they're deliberately trying to create an uneducated underclass for the near future that is easy to control because it has no means of figuring out anything that isn't served to them by a (corporate or state controlled) chatbot.

    💬 1🔄 3⭐ 15
  • Oct 12, 2026, 9:19 PM

    @aesthr Plus the goal of AI from the techbro is not to do the simples tasks but do to any tasks.

    So it will become hard to learn hard skills AND not profitable financially to do so.

    💬 0🔄 0⭐ 0
  • Oct 12, 2026, 9:31 PM

    @aesthr the "simple problem" is a rich person asking the machine for something. The "hard problem" is a human completing said task while pretending the machine did it.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 13, 2026, 2:21 AM

    @aesthr this has already been happening before LLMs and will continue to happen after LLMs. Abstraction-forgetting is a common feature of human progress. Example: making a pencil from scratch, compilers compiling compilers, smelting iron, making plastic, and so on to make a toaster. This doesn't even touch stuff like ASML. It's normal to abstract and delegate problems so you can solve the hard parts.

    💬 0🔄 0⭐ 0
  • Oct 13, 2026, 2:23 AM

    @aesthr I'm infinitely grateful I never have to write boilerplate again and I can use my coding time thinking deeply about what matters.

    💬 0🔄 0⭐ 0
  • Oct 13, 2026, 3:01 AM

    @budududuroiu @aesthr yes, *because you already have those skills*. What this conversation is about is the effect on all those people who *do not have* those skills yet - all of your examples are of 'abstracted' technological steps which aren't "simple versions" of the things we use the resulting products for - LLMs tend to be pushed as (and used by students for) automation for the simple(?) versions of tasks that we do to develop the skills to do those hard things.

    💬 1🔄 0⭐ 0
  • Oct 13, 2026, 3:04 AM

    @aoanla I think future office workers will be just fine delegating busy work to LLMs. No engineer should have to know the GitHub Actions syntax

    @aesthr

    💬 1🔄 0⭐ 0
  • Oct 14, 2026, 7:59 PM

    @budududuroiu @aoanla @aesthr no engineer have to think that github actions are the best thing there is. All this boilerplate just happens cause noone thinks how to abstract it properly. Why do we had code generators to generate boilerplate for us but why though if it can be generated why is there no system that does it on the fly so you dont have to geneeate the code in the first place. Llms will lead to a bazillion lines of code that doesnt have to be there

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 14, 2026, 8:02 PM

    @aesthr its always not about how to do hard things its about how to solve hard thibgs in a simple way by shifting someones perspective. The obvious solution is seldomly the esiest one. Which is mostly embraced by llms. They solve it bruteforce. Rolling a dice all the time and get a good aproximation whats most of the time not the solution. Like pi is not just aproximatly 3.14

    💬 0🔄 0⭐ 0