2 pointsby aspittel5 hours ago2 comments
  • don_r5 hours ago
    I hadn't heard the RLHF feedback loop argument before, but it seems like a solid explanation for all the bad ticks LLMs have when writing and why it's so hard to break them out of it.

    All of that stuff... "That's not ____; it's _________" and the like... you can imagine human evaluators going "Wow, that's profound! Great output!" and now we all get to suffer through reading that for the rest of our lives.

  • lich_king5 hours ago
    An article about the AI writing style, written using AI.