Re: withdrawing Proposal A -- Interpretation of DFSG on Artificial Intelligence (AI) Models

On Thu, 8 May 2025 at 22:58, G. Branden Robinson
<[email protected]> wrote:
> I'm not sure our Constitution contemplates a "conditional withdrawal" of
> a GR ballot option.  An option is either eligible for inclusion or it is
> not, and the less discretion we demand of the Project Secretary to judge
> that fact, the more predictable our processes will be.

To make the process more clear - I am withdrawing my option for this
ballot. Let's reintroduce the GR with all appropriate choices once
more research and discussion is completed outside the formal GR
process.

> > * People holding different opinions have too short time to prepare
> >   (although I already signaled everyone long time ago that I'll press
> >   the start button).  The lack of other options can make the result
> >   less convincing.
>
> To put my own spin on Russ's response to you, developers, being human,
> love to procrastinate.  Reasoning carefully about the impact of LLMs on
> software development, software freedom, copyright law, and Debian's
> mission is a significant demand and to the serious-minded, likely sounds
> suspiciously like "hard work".

Guilty! I did not even really notice this topic percolating on other lists.

> The rest of the world is not going to wait for us, and very likely will
> make decisions with less careful consideration, as we've seen (IMO) with
> OSI.  In many places, the deepest pockets make the biggest decisions.
> One of the nice things about our Project is how that's less true here.

Yes, but also if we wait long enough, we might actually see more clear
court decisions on some key issues here. It would be good to have a
*very explicit* statement from a court saying that a developer
downloading a copyrighted work and then using that to train an AI
model is *not* committing copyright infringement. And that the model
and its outputs are not derived works of the training data in the
sense of copyright law.

> One of the few less-than-salutary aspects of DebConfs, mini and major,
> is the opportunity they present for further off-the-books decision
> making.

When mentioning DebConf I was more thing about beer-lubricated
philosophical discussions about "what even is learning?" or "what even
is the difference between autocompletion statistical suggestions and
LLM statistical suggestions?" and not a shadowy cabal meeting in a
secret underground room and deciding to rewrite sudo in Rust :D