[gdal-dev] Proposed RFC 8 amendment regarding (prohibited use of) generative AI tools

[Jan Heckman]

Sorry to butt in.
I have no say in this whatsoever.
A few mayhaps/perby useful points:
- not using GAI is not a lasting option, so rather focus on precautions and
be aware of dangers.
Prime danger, to me, seams copyright trolling ala Oracle. They have
bought/adopted stuff in order to sue people.
Of course, GDAL is not rich enough to profit, but reputation damage is also
to be avoided.
Precaution would be rewrite with gdal naming convention, see whether the
usually general code can be more specific to the context.
Another weird idea: perhaps you can ask your copilot or whatever GAI
whether copyrighted material is involved. Experimental, wishful thinking,
Don't know yet.
Suppose you get a positive answer, then do a strong rewrite or do not use
at all.
Sometimes framing your questions in a slightly more general way may help
you without too much copy paste.
- SO. identifying each successful SO inquiry can be quite obnoxious.
Can you persuade SO into a simpler agreement?
I am thinking of decorating all modules with an SO mention and
adding an option to each (ogrinfo, ogr2ogr etc) executable: --credits.
Hardly have to explain more, I guess. If you do, mention --credits in
--help.
I have no suggestion for RFC8, except the general tone: don't use unless,
which is already there.
And the final advice to anyone would be: do not just copy-paste.
I'll show myself out 😉

On Tue, Oct 8, 2024 at 6:21 PM Howard Butler via gdal-dev <
gdal-dev at lists.osgeo.org> wrote:

>
>
> > On Oct 7, 2024, at 11:38 AM, Even Rouault via gdal-dev <
> gdal-dev at lists.osgeo.org> wrote:
> >
> > Hi,
> >
> > I propose we update RFC 8 with the following changes:
> https://github.com/OSGeo/gdal/pull/10958
> >
> > Can we enforce that? No (similarly we can't prevent someone from
> stealing copyrighted material from source code of proprietary software they
> have access to, or "re-license" or "forget" copyright attribution of open
> source code without authorization), but at least we set the policy.
> >
> > This is admittedly a highly controversial topic. Debian for example has
> had related discussions but didn't come with a position for now, whereas
> Gentoo banned use of generative AI: https://lwn.net/Articles/972331/
> >
> > It seems more prudent to me to ban for now, and potentially relax in the
> future, when/if the legal status of this becomes clearer. (not even
> mentioning the ethical or environmental impact sides of such tools)
> >
> > Even
>
> I wonder if the policy should be to require any contributions that use
> generative AI *be labeled as such*. This would serve two purposes – it
> would allow us to react to contributions that would be problematic to
> include and it would allow us to preserve those labels in the codebase for
> any of the contributions for future users of the software.
>
> It is always the responsibility of the users of the software to be aware
> of the copyright implications of their use of the software. The project
> should do its best to give users the tools to know the status of the
> software in regard to where code that is included originated.
>
> As to how to functionally implement it, maybe a simple tag on PRs that
> users must check ("did you use copilot or similar when making this PR?" and
> then some kind of automation that push that tag into the changeset history
> when it is merged.
>
> Howard
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20241008/8066946b/attachment.htm>