Login
You're viewing the functional.cafe public feed.
  • Oct 29, 2026, 1:32 PM

    Public Domain code is neither copyrightable, nor copylefteable.

    Isn't this the most free possible software at the moment?

    💬 3🔄 0⭐ 1

Replies

  • 💬 1🔄 0⭐ 2
  • Oct 29, 2026, 1:48 PM

    @ryanprior Yeah, I'm more about public domain in general.

    According to referenced post: I don't think it's correct to mix the technology and companies that produce one or a few particular variation of it in questionable ways.

    💬 2🔄 0⭐ 0
  • Oct 29, 2026, 1:55 PM

    @abcdw when approximately 100% of "LLM usage in open source" is from models dependent on slave labor from training, and the harm is ongoing and getting progressively worse, it's not appropriate to talk about separating the technology from the means of producing it.

    💬 1🔄 2⭐ 4
  • Oct 29, 2026, 2:21 PM

    @ryanprior IMHO it's more efficient to name organisations or people doing the wrong things rather than a technology/ideology or whatever they may have some association with.

    💬 1🔄 1⭐ 1
  • Oct 29, 2026, 2:34 PM

    @abcdw the technology itself is compromised by the misdeeds-at-scale of the industry that's trying to capitalize it. People's understanding of what an LLM model is, and can do, is overwhelmingly shaped by that industry's framing. Using language that creates ambiguity between the real, harmful current application and a hypothetical future application is not neutral.

    💬 0🔄 1⭐ 1
  • Oct 29, 2026, 2:16 PM

    @abcdw @ryanprior can you share any examples of llms of non-"questionable" provenance that are actually being used to develop software, rather than just theoretically usable for this purpose? if all you can find is a possible future theoretical use, what makes you certain that future will materialize? and why talk about it now?

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Oct 29, 2026, 2:26 PM

    @abcdw @ryanprior models to write code?

    trained at home or on some cloud, i wonder?

    any consent obtained from usenet folks and such or just kinda stolen?

    (comparatively though, smaller models like that, which don't require all the exploitations of the larger ones so much less evil)

    💬 0🔄 0⭐ 0
  • Oct 29, 2026, 1:42 PM

    @abcdw
    I think it depends in what legal regime you work. I think its possible to take a public domain code, add some features, and then copyright the whole thing in some legal systems ... you can't keep others from using the original public domain code. but you can close off the new code as proprietary. not 100% sure about this

    💬 1🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • Oct 29, 2026, 2:24 PM

    @abcdw@fosstodon.org llm output is in the same legal regime as intermediate build files, like .elc files or .go files for guile. It's free in the sense that it has no value except for the creator, no possible community can be built just sharing elc files with each other and hoping we trust each other.

    The most valuable part of llm generation is the input along with a model specification, that's the 'code' part of using a tool like an llm.

    Everyone should be disclosing their prompts and model every time an llm is used make a change. I am amazed at how uncommon that is. It would allow others to see your contribution along with the output; it's the only copyrightable thing.

    I don't think it would make it ethical, but it would sure be prosocial, which I feel is also severely lacking in the ai community.

    💬 1🔄 0⭐ 0
  • Oct 30, 2026, 7:00 AM

    @amy The question was about public domain code, not LLM output. I guess the reply to the previous message about llm output made it misleading.

    💬 0🔄 0⭐ 0