Hi,
On Mon, 23 Feb 2026 at 16:40, Ludovic Courtès <[email protected]> wrote:
> As this discussion shows, we’ll probably need a policy as to what we
> accept in Guix: what packages we allow (is “open weight” good enough?
> what’s the Corresponding Source?), and how we deal with contributions.
I think it soaps a slippery slope. How would you detect this
contribution using some inputs from LLMs?
Many thanks to Hugo for their moral value and their transparency!
For me, the list in [1] are good examples that shows contributions using
LLMs are impossible to detect. Look, this is the contribution [2], line
560-572:
(add-before 'check 'patch-potfiles.skip
;; Fix for https://github.com/lxde/lxsession/issues/42
(lambda _
(let ((file "po/POTFILES.skip"))
(with-output-to-file file
(lambda ()
(display "lxsession-default-apps/combobox.c\n")
(display "lxsession-default-apps/main.c\n")
(display "data/lxpolkit.desktop.in\n")
(display "lxpolkit/main.c\n")
(display "lxsession/app.c\n"))))
#t)))))
and this is the ChatGPT output [3]:
(add-before 'check 'patch-potfiles.skip
(lambda _
(let ((file "POTFILES.skip")))
(call-with-output-file file
#:append #t
(lambda (port)
(display "file1.c\n" port)
(display "file2.c\n" port)
(display "file3.c\n" port)))
#t)))))
Therefore, what would be the policy? And how do we apply it?
IMHO, such policy is a dead-end and/or inapplicable.
If I stretch, consider that I’m using non-free drivers, running non-free
OS, I’m asking help to some evil person who never respects any items of
our Code of Conduct, in summary, I’m living the worst unethical* life.
Does the project reject my contributions?
Including LLMs in a policy as to what we accept in Guix paves the way
for a kind of “police of thought”, IMHO.
Maybe I overlook something, well, from my point of view, we cannot
encode ethical behavior at the level of the PRs.
*unethical life: the opposite of what I’m trying to live for real. :-)
> One thing I wonder is whether the project should have an “opinion” on
> generative AI in general and its relation to computer programming and
> human creativity. I do have an opinion :-) and the question is whether
> taking a stance as a project is worth it and feasible.
I think we could have a kind of “Social Contract“. More than one year
ago, I’ve written this [4]. As a project, we would draw some ethical
principles (moral values) that are encoded as ethical obligations (moral
duties).
In the light of my ethical principles / moral values, do I fulfill my
ethical obligations / moral duties?
By contributing to the project, I contribute to build a world where
the community is benevolent and autonomous, and thus the community acts
in this direction by relying on concrete measures.
Each people do their best considering their own constraints, history,
skills, knowledge, lives, etc. The community trusts people to act in
good faith, although each of us are not perfect. For many ethical
questions, it’s difficult to draw beforehand a clear line that fits all
the grey cases, IMHO.
Well, if we all are here and are spending countless hours, for sure it’s
because some technical stuff, and it’s also for the humanist project.
For what it is worth, I do not feel qualified – using my own personal
history, skills, knowledge, way of life, etc. – to emit a strong
judgement of “badness” when someone would contribute using LLMs inputs.
Well, I have a strong personal opinion on the topic of LLMs but I
refrain myself to push this opinion on the behavior of peers (although I
sometimes fail).
In all in all, the stance as a project appears to me to draw what as a
project we value, i.e., what “we should do” without entering in some “we
must not do”.
> and the question is whether
> taking a stance as a project is worth it and feasible.
Concretely, my proposal reads:
• two ethical principles / moral values:
• Being benevolent.
• Being autonomous.
• many ethical obligations / moral duties:
• Have an actionable Code of Conduct.
• Have transparent decision-making process.
• Documentation and manual are more important than features.
• Translate as much as possible.
• Emphasize equity by helping other to make their own progress.
• Rely only on software that I run as I wish for any purpose and
that I study and/or change as I wish.
• Refuse to negociate with software that I cannot (or not allowed
to) study and/or modify and/or share copies (modified or not).
• Accept to negociate with software that I cannot study and/or
modify and/or share copies only if I can then distribute software
that I can study, modify and share copies (modified or not).
• The bootstrap of the program is more important than features.
• Craft my own solution if I do not feel autonomous with the others.
• Using LLM doesn’t help in the long term
Bah it is not because the ideal is unreachable that one stops to try to
live with it; it is perhaps by trying again and again to live with this
ideal that it might make this ideal possible. :-)
Cheers,
simon
1: Re: Can a project accept LLMs' code contribution and remain free software?
Hugo Buddelmeijer via "Development of GNU Guix and the GNU System
distribution." <[email protected]>
Thu, 05 Feb 2026 09:52:46 +0100
id:[email protected]
https://lists.gnu.org/archive/html/guix-devel/2026-02
https://yhetil.org/guix/[email protected]
2: https://codeberg.org/guix/guix/pulls/1623/files
3: https://chatgpt.com/share/69844b1c-0a10-8011-8308-333ed22e3e2d
4: https://simon.tournier.info/posts/2024-11-01-visiting-future-gnu.html
