So I accidentally triggered an online shitstorm in the process of trying to craft a set of more restrictive LLM usage guidelines for KDE. Sorry about that. :/ I’m going to leave the comments …

Personal opinion: Nathan seems to be really side-stepping a lot of concerns brought of up in the comments there. And referring to people coming together collectively in an open-letter when virtually all other channels are closed as ‘brigading’ is… Not cool. Edit: he retracted the use of brigading.

EDIT 2: Nate has since closed the comments on his blog, and the one discussion thread on KDE Discuss about the AI Policy was closed suddenly without explanation after a KDE developer began to discuss the ethics of AI use. Nate then responded after the thread was closed with this:

  • Feyd@programming.dev
    377·
    4 days ago

    He can give 13 clearly articulated and factual reasons not to use LLMs and 1 nebulous opinion-based reason to use them yet gives them equal weight on the scale 🤔

    One would think the following would be reason for a project such as KDE to ban them outright at least until the dust settles.

    The license of source code they produce is legally unclear.

    • SigHunter@discuss.tchncs.de
      102·
      4 days ago

      literally the next sentence after the reasons, he calls LLMs “bad stuff”. That does not sound like “equal weight on the scale” to me.

      So yeah, bad stuff. I completely understand why a lot of people have problems with LLMs. I have these concerns as well.

      I agree that the license issue should be enough to not touch anything generated by AI, to not taint the codebase

      • Feyd@programming.dev
        72·
        4 days ago

        Did you stop reading there? Here is your quote + the next paragraph

        So yeah, bad stuff. I completely understand why a lot of people have problems with LLMs. I have these concerns as well.

        I also understand why a lot of experts in their field really like LLMs. These are very powerful and sophisticated tools. Capable people with decades of experience can successfully automate away a lot of tedious work, and still end up with a final product identical to what they would have done by hand.

        I don’t think that conclusion is even accurate and it’s certainly the subject of much debate. He then crunches all of both negative and positive into a 5 bullet list that makes it look much more balanced:

        • Very powerful
        • Genuinely useful for various things
        • Dangerous, de-skilling, and/or addictive in the wrong hands
        • Full of really bad externalities
        • Broadly available worldwide at effectively zero cost

        You’re apparently reading it differently but it looks to me like half the article is trying to funnel into justifying making this a difficult decision after making the case for it really not to be up front. Why do that?