On Tue, 26 May 2026 17:23:53 +0900 Nguyễn Gia Phong via "Development of GNU Guix and the GNU System distribution." <[email protected]> wrote:
> Seconded. We only need to reach a concensus regarding > the course of action, not the reason behind it. I think that the reasons are extremely important, and but "the reasons" can mean many things and they can be vague as well and being vague can be a feature to more accurately describe the reality, not a bug. For instance we already have policies so if we enforce a no-llm generated code from these policies, and in this context the reason why a given contribution is refused important, and it often comes from the rules that are being decided, so at the end of the day having justification as clear as possible is important. In the case that the refusal isn't already covered by existing rules (copyright, legality, quality, etc), like translations that are clearly not derived from works under incompatible licenses, I do agree though that the reasons can vary from individual to individual and so I don't see why we can't try to express just that and to do that we could have something vague enough as justification to allow everybody to understand the issue. For instance we could use words that are known to be extremely subjective like positive contributions to the world or something like that: > Currently LLMs have too many issues (environemental impact, > use for genocide, for repression, copyright uncertaincy with derived > works, quality, etc) and while not everybody agree on every single > issue, in overal such tools currently don't don't contribute > to make the world a better place. And we could rule such tools out on the basis that because they don't really make the world a better place, Guix contributors are more confortable with their ban. And add a justification similar to that (which is modified from [1]): > And the social cost to the GNU Guix community is not worth the > benefits that these tools may bring. So basically we manage to justify rationally that we don't want LLMs and that makes it clear for everybody without having to agree with every single reason. And here even if someone doesn't agree with a single reason or a few of them it is clear enough. Same if things do change over time. In that case it would allow us to reconsider later on if all the reasons are fixed and that in overall we can prove that some future LLMs can make the world a better place. This would also not give any reasons to accept the ones that don't make the world a better place. Though maybe these would have other names (SLM for small language models). Or maybe this would never happen. [1]https://www.debian.org/doc/manuals/developers-reference/developer-duties.html#upstream-coordination Denis
pgp6TYZ47MfnE.pgp
Description: OpenPGP digital signature
