• agamemnonymous@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    17
    arrow-down
    15
    ·
    1 day ago

    Unpopular opinion, but I can see the utility of well-trained LLMs in legal preparation.

    Not for writing arguments, obviously, but preliminary searches are one of the best uses I’ve found for AI. It could be pretty effective to sift through mountains of case law for relevant precedent, granted that you actually evaluate your results manually.

    • bizarroland@lemmy.world
      link
      fedilink
      English
      arrow-up
      10
      ·
      edit-2
      1 day ago

      There are explicit published guidelines for using AI in the legal field.

      No court is actually against the use of LLM-generated text.

      The problem is lazy lawyers. They just accept what is written on the screen in front of them. They do not rewrite it. They do not verify any quotes or legal briefs or opinions that it spits out at them.

      And that is expressly forbidden. Everything that is submitted under a lawyer’s name has to be either written by or directly reviewed and approved by that lawyer.

      Even lawyers know 80% of any legal brief is just stuff put in there to meet the text requirements for review.

      If you use an LLM to fill in that 80% with some manual guidance, then it can allow you to focus on the 20% of the legal brief that actually matters.

    • SparroHawc@piefed.world
      link
      fedilink
      English
      arrow-up
      11
      arrow-down
      3
      ·
      1 day ago

      Less unpopular than you might think.

      LLMs are an interesting and potentially useful tool; however, as it does with most things, capitalism is ruining it for everyone.

    • zr0@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      4
      arrow-down
      4
      ·
      1 day ago

      I also use it for writing arguments, as it can come up with some creative ideas. However, they do not just blindly end up in the final document. In every letter, there is at least one major mistake, probably due to the token limit.