Hello Guix, On 2026-05-03 at 22:18+02:00, Noé Lopez wrote: > Untrusem <[email protected]> writes: > > So that's why I wanted a way to know when a guix package > > updates or gets added [...] that LLMs is being used > > That’s an interesting idea, I also agree that I would like to know what > is made with LLM or not. Maybe we could start tracking this kind of > things. > > I was also thinking that it would be good to have something like > F-Droid’s anti-features. For things like using nonfree APIs or nonfree > graphics.
I think package-properties is a good fit for this. Our neighbor was discussing this matter a couple months ago [0] and one option came up was overriding the license field. I would like to be able to query a list of transitive dependencies that include LLM output, which should be similar to other graph traversals like refresh --list-dependent or graph --path. [0]: https://public-inbox.gentoo.org/gentoo-dev/[email protected]/t On 2026-05-03 at 13:05-07:00, Vagrant Cascadian wrote: > As a member of the guix kernel team, I regret to inform you that the > linux kernel upstream uses AI/LLM to backport patches for all the > versions of the linux-libre kernels we provide with guix. > > I think even the GNU Hurd development has started using some of these > technologies, if you are looking for another kernel... that would > (eventually?) have the same issues... Sad news aside, let not make perfect the enemy of good. I use non-free firmware and drivers, but I still want to avoid using other proprietary software, and would love some data and tooling to make that easier. On 2026-05-03 at 13:09-07:00, Ian Eure wrote: > I also have some reservations about it being a package contributor's > responsibility to do this; it can be difficult to know if an LLM > is used in the first place[1]. Are folks contributing version bumps > responsible for auditing the changes in the new version > for the presence of LLM use? > > [1]: Some maintaners deliberately obscure presence and extent of > LLM tools in their projects. Packagers are also responsible for auditing the changes in the new version for the presence of bugs, so I think it's okay if we miss things from time to time. On 2026-05-03 at 13:09-07:00, Ian Eure wrote: > I completely agree with you: users should have the freedom to run > software which doesn't contain LLM output. However, the subject > is complicated, and I think requires a GCD. Now that the discussion periods for GCD 6-7 are over, I'm open to proposing a common stance regarding LLM. Linking back to an earlier thread [2], I think it should cover incorporating LLM ouput into Guix itself as well. I like OpenJDK interim policy [3]: > Oracle, as the corporate sponsor of the OpenJDK Community, > is working to draft a full policy governing the use > of generative AI tools in OpenJDK contributions. > Oracle will propose that policy to the OpenJDK Governing Board > in due course. Until that policy is in place, the Governing Board > has approved this interim policy: > > > Contributions in the OpenJDK Community must not include content > > generated, in part or in full, by large language models, > > diffusion models, or similar deep-learning systems. [...] > > > > Contributors in the OpenJDK Community may use generative AI tools > > privately to help comprehend, debug, and review OpenJDK code > > and other content, and to do research related to OpenJDK Projects, > > so long as they do not contribute content generated by such tools. > This interim policy aims to [...] limit their risks > while we gain further experience that will inform the full policy. If Oracle, the corporation memed as having more lawyers than engineers, and a major stakeholder of the big statistics market, is voicing concern over incorporating deep-learning models' output in their codebase, I think we should listen. As mentioned in the linked thread, while we wait for legal advisories from the FSF or FSFE, a similar temporary policy should be in place. (Though I do not deny wishing that such temporary policy become effectively permanent, I believe my line of reasoning is objective.) [2]: https://yhetil.org/guix-devel/[email protected]/t/ [3]: https://openjdk.org/legal/ai Best wishes, Phong
signature.asc
Description: PGP signature
