LLM policy

[Regina Obe]

> From: Greg Troxel 


> > Ultimately the purpose of a code of conduct is to help increase the
> > value of our work for the world in the most efficient way and to
> > welcome valuable resources into our community.
> 
> Some think that, and some see it as a codification of a group norm of
> courteous and non-discriminatory behavior because that courtesy is morally
> correct independent of whether it helps the project's value or it doesn't.
(The
> reality is uglier, but I'll leave it as theory.)

Morality is such a subjective thing.  If I can't explain why I think
something is immoral I'd rather not bring morality into the equation.
Just saying something is immoral seems to be a common way for people to shut
down conversation of dissenting opinions.

Is capital punishment immoral?  I really don't care cause it's not part of
my equation as to why or why not it should be done.

Most things I think of as immoral are because I can qualify the damage they
do for my community or the world, not because I have a list of what is moral
and immoral.

Since my project is supposed to be a common good to better the world,
anything that is bad for the world is bad for my project too.  Ergo no point
in bringing morality into the conversation.  Yes postgis can be used for
immoral things like sending drones to kill people, but sadly that is out of
my scope of influence. 

> 
> >> LLMs are only being talked about specifically because essentially
> >> everything else you are calling about as bad is near-universally
> >> considered not ok. 
Most things in Coc should be fairly obvious to everyone, but we still state
it 
because what is obvious to us is not necessarily obvious to someone else.

> LLM use is under discussion because some think it is outside existing
norms or
> that it should be outside norms, and some think it's ok.
> 
Agree and I think it should be treated exactly like all existing norms.
It's no different than when we discussed is it okay for users to use
computers or calculators during exams.
It's okay if all have access to the same tools, but when there is unfair
advantage that is bad for the community at large.
That is something we should focus on even if we can't quite level the
playing field.

By unfair I mean a large company co-opting a project just cause they have
immense processing power and then using it for their own needs.


> > Agree, except we don't allow any stolen work period whether it be LLM
> > or someone who stole code from another project that doesn't allow code
> copy.
> 
> Then, I think we agree that under *the project's existing rules*
> 
>   1) any code which was not written by the submitter must be identified
>      as such, with a stated author/authors, and submitted only if all
>      those authors have consented to licensing the contribution under
>      the project's license.
> 
Agree and if an LLM helped you author it you should say so in your
submission, just as you would credit any author.
Treating an LLM as a proxy human works perfectly here.

>   2) LLM-generated code is not acceptable, unless it was generated
> 
>      a) by a model using only training data that is licensed for LLM use
>         *which includes permission to reuse without attribution or any
>         licensing text*, and includes only data for which there is true
>         consent, not via a fraudulent claim of expected-not-to-be-read
>         clickthrough license as part of some larger service
> 
Disagree.  There is such a thing as fair use, and I think if you look at
legal docs

Eg. The Anthropic case which I am a plaintiff of you will see that.

For example if I read a book on C that I bought and I learn how to write C
from it. Then I apply my new knowledge to provide a patch for PostGIS.
How is this different from an AI reading the same book that its owner
purchased and gleaning the same knowledge. Neither is using code verbatim
from the book.

There is nothing to my knowledge that says licensed to LLM.  If you have
such a reference I would like to see that.

Yes if an LLM takes verbatim code from a GPL project or a private repository
whatever and provides it then that is wrong, as it would be wrong for a
human to do so.
Most coding LLMs do provide a note that yes you can use code generated by
LLM and give provisions under what conditions.

If a project explicitly says, no LLMs can read my code, that too is a clear
violation of an LLM if it doesn't abide by that requirement.
Just as if you called out my name "All are free to use my code except Regina
Obe".


>      b) (probably, but arguably the existing rules do not prohibit
>         antisocial behavior by contributors) not produced by an
>         organization that engages in abusive scraping.  Arguably abusive
>         scraping is a CoC violation as it is harassment of humans who
>         manage other web systems, and/or vandalism.
> 
> Because there are not any code-generating LLMs that meet 2(a) it follows
that
> LLM contributions to the project are not allowed.
> 
> This should then be straightforwardly clarified, as those that like using
LLMs
> seem to be willing to adopt differing interpretations of existing rules.

As mentioned I don't agree with your 2 (a) 
I do agree that if your code is not fully your own, you should state the
additional authors, an LLM is an author like any other and deserves to be
attributed.
And that you have proof you have right to use such code -- again the LLM use
contract.