Hi, Nguyễn Gia Phong <[email protected]> writes:
> Hi Maxim, > > On 2026-05-28 at 16:36+09:00, Maxim Cournoyer wrote: >> Of course, if it's ruled that X tool's output infringes >> on its trained material, and if software Y was written mostly with X, >> we'd have to conclude that Y's advertised license is invalid >> and thus nonfree. > > Similar to the current GCD draft, why does it matter if it is _mostly_? > I can be convinced if Y only takes outputs from X that are not > legally significant to be consider infringing though. My point was that I think it's preposterous to block adding packages that used LLM outputs to our collection; my opinion is that we should wait to see how legal understanding develops. For our own code base, I'm fine with being more careful and pledging to not commit LLM-authored output (that's also the guidance that is emerging from internal GNU discussions, to be published "soon" to the GNU website). > On 2026-05-28 at 16:36+09:00, Maxim Cournoyer wrote: >> About the first (1) pledge point: I disagree about blanket pledging >> not using any of the LLM tools. As others have found in our community, >> they can have a use. I'd suggest dropping this point of the pledge. > > Isn't this _the point_ of the pledge? > I think other found uses are not within the stated actions: There are many points in the pledge; one (1) is not using or recommending LLM in the artifacts-producing activities surrounding Guix; two (2) is not installing any LLM-generated output in the Guix source tree. I think 2. mostly covers 1., so I'd drop the first one, which I find overreaching. -- Thanks, Maxim
