• schipelblorp@sh.itjust.works
    link
    fedilink
    arrow-up
    6
    ·
    2 days ago

    I’ll eat the downvotes: AI is most useful early on when your speaking skills would burden a native speaker. Having conversations, even with AI, where you are on-the-spot generating language and figuring out how to say novel things is really valuable.

    And I don’t really see what the problem is, exactly, within the bounds of learning a language (and not making it your companion). AI hallucinates facts, but I don’t think it hallucinates grammar or vocabulary.

    Open question of whether it is good at corrections or not. After conversations, I would ask it review what I’ve written and give me corrections. I would try to understand what my error was and evaluate the correction. There’s lots of “how would a native speaker say this” type questions that grammar guides aren’t great with. But that’s really where I think you might get into trouble: relying on completely on AI to tell you how to speak the language, but I understand the temptation!

    But as far as absorbing content, you’re better off with the amazing variety of human-made fare that’s available on the interwebsand actually finding some real people to talk to.

    • rowansprig@leminal.space
      link
      fedilink
      arrow-up
      2
      ·
      5 hours ago

      Language is a thing that large language models are pretty good at in the scheme of things. My only concern is that the tone of an LLM might be bad to adopt. But if you’re watching and absorbing enough human made content, it should be okay.