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:

  • Eldritch@piefed.worldEnglish
    2·
    2 days ago

    Currently this does not exist.

    False. There are a number of models trained on data they were explicitly given permission to train on. Are these the models they’re using? Well I know as much as you, which is to say I don’t know. Though I think we’d probably agree they aren’t largely. But aren’t != can’t.

    The AI policy did not restrict AI usage to non-destructive self-hosted models.

    Brody’s videos that I linked go into this rationally and reasonably.

    1. These are unpaid volunteer developers.
    2. With what tools and funds will they police this?
    3. The Black Arts of control C control V are also widely known. Better that the code be attributable and traceable than not.

    There’s no reasonable way they can actually ban this. And they don’t have the time or money to police it. The only real solution is to just shut the projects down at this point. Is that what you’re aiming for?

    The people who are most impacted by the negative externalities of AI are not taken care of. The people losing their jobs are also not taken care of. That is why people care so much. It’s not viable to handwave away the negatives AI are causing in the here and now with the assurance that in some far off day we’ll live in utopia where the negatives won’t matter.

    No one is hand waving away the negative impacts. You are simply misattributing them to artificial intelligence and not wealthy authoritarians seeking to profit. By all means, fuck zuck, musk, altman, etc and their companies. They’ve always been and always will be destructive in every way possible. AI or not.

    • A_norny_mousse@piefed.zipEnglish
      2·
      2 days ago

      Not that I’m agreeing with OP and their ilk, but imo “ethical” must include broader environmental and societal and geopolitical impact.

      You do not address this.

      But afaik LLMs that are ethical in this respect as well, also exist. Although maybe they aren’t “Large”.

    • ProdigalFrog@slrpnk.netOPEnglish
      3·
      2 days ago

      False. There are a number of models trained on data they were explicitly given permission to train on.

      So far from what I’ve seen when I investigated self-labeled ethical models, I found that they weren’t actually trained ethically, and not on code that is compatible with GPL or code that doesn’t need to be attributed. (if you can find one that has proven otherwise, I’d be open to taking a look at it).

      That said, I don’t think there would have been as much push back if the KDE proposal had explicitly limited AI use to self-hosted models.

      There’s no reasonable way they can actually ban this. And they don’t have the time or money to police it.

      According to the KDE devs, the proposal was intended to allow them to easily reject clearly slop PRs. A policy that limits AI use to self-hosted models would have achieved that goal just as well as one that allows any model, and would essentially be an honor system, which I suspect most contributors would be willing to follow (obviously not all, and there is nothing stopping a contributor from lying that their corporate AI code was generated on a local LLM, but then again, there are also rules asking contributors to not knowingly submit copyrighted code from proprietary software, which is also an honor system).

      You are simply misattributing them to artificial intelligence and not wealthy authoritarians seeking to profit. By all means, fuck zuck, musk, altman, etc and their companies. They’ve always been and always will be destructive in every way possible. AI or not.

      To be honest, I’m struggling to understand why you would be against a policy that at least attempts to limit the benefit/gain to those wealthy authoritarians. A policy that did so would not require any further resources from the dev team compared to the one they attempted to implement, it would basically just be an official request to contributors to not use the explicitly fascist owned tools that are most contributing to global warming and our subjugation.