Re: GR -- Allow AI-Assisted Contributions (was Re: DFSG-new-queue is AI-generated)
"Theodore Tso" <[email protected]>
| Newsgroups | gmane.linux.debian.devel.vote |
|---|---|
| Message-ID | <[email protected]> |
> > > > https://lists.debian.org/debian-vote/2025/05/msg00145.html > > > > > > That includes the following section: > > > > > > +--- > > > > Any work resulting from generative use of a model can at most be > > > > as free as the model itself; e.g. programming with a model from > > > > contrib/non-free assisting prevents the result from entering main. > > > +--- > > > > > > which would ban projects like the Linux kernel, the Python interpreter, > > > LLVM, ... and their reverse dependencies from Debian, likely also from > > > non-free. > > > > > > What is the plan to deal with that? > > > > I don't think that is the only (or even intended) interpretation If you look at [1], it's *really* clear to me that this was the intended interpretation. Consider the text beyond that point: Any work resulting from generative use of a model can at most be ^^^^^^^^ as free as the model itself; e.g. programming with a model from contrib/non-free assisting prevents the result from entering main. The "/usr/share/doc/PACKAGE/copyright" file must include copyright notices from all training inputs as required by Policy for “any files which are compiled into the object code shipped in the binary package”, except for inputs already separately packaged (such as the training software, libraries, or inputs already available from packages such as word lists also used for spellchecking). Regarding availability of sources used for training, the normal rules of the non-free archive apply. If the intent is *not* to force the Linux kernel into non-free, the wording in [1] needs to be clarified. To the extent that "free" is defined by the DFSG, I'd argue that this is a legal question that should be answered by lawyers, and not something that should be subject to the GR. If the intent of this proposal is to define (or redefine) "free" to mean "no LLM generated work, regardless of the legal copyright impliciations in a particular legal regime", then sure, it should be something that would be fine to vote on in a GR. But then let's do that by changing the DFSG. > Given other mails, I think you are wrong and the intended > interpretation is to ban all LLM output from Debian. ... and I would agree. > The reason for a GR would be that the policy today doesn't say > anything. Otherwise no GR would be needed. Just as in last year, I don't think the project as a whole is ready for a vote on the GR. But if there are people who believe that we are ready to redefine "free" to mean "no AI, ever", then perhaps we should have a GR to make it clear whether or not that is the project consensus. The other reason I can think why we might want to have a GR is if you want to prohibit the use of LLM output in Debian packaging, or if we wanted to ban the use of non-DD's in the NM process from using LLM's. I personally think that is *also* a bad idea, but others might think that we should force that question to a vote. - Ted