位zyd

I asked emacs-devel: Are LLM contributions welcomed in GNU Emacs by the maintainers and contributors?

Eli Zaretskii:

We are awaiting the decision by the GNU Project on these matters, which will define the policy for all the GNU packages, and in the meantime we don't accept LLM-generated code, as a precaution.

For now, Emacs won't accept LLM contributions. However, given this:

Obviously, the right thing to do is protect computing freedom: share complete training inputs with every user of the LLM, together with the complete model, training configuration settings, and the accompanying software source code. Therefore, we urge Anthropic and other LLM developers that train models using huge datasets downloaded from the Internet to provide these LLMs to their users in freedom.

I sort of expect FSF, and therefore GNU along with Emacs, to be somewhat accepting of LLMs eventually, if they are "open". Too copyright-brained for their own good. Unfortunate.

#emacs

What is the policy/attitude towards LLM contributions in GNU Emacs?

Mar 14 at 2:41pmWeb
Show threadRobert Pluim 馃嚜馃嚭Mar 14
@zyd Given that in the US LLM generated code cannot be copyrighted, how do you legally integrate such code in a GPL-licensed project?
Show thread位zydMar 14
@rpluim Dunno but laws can change and they usually benefit holders of capital. Being opposed to LLMs merely on copyright grounds is a weak position I wouldn't want to be standing from. Any strong opposition has to be based primarily on the social element of LLMs.
Show threadRobert Pluim 馃嚜馃嚭Mar 14
@zyd I'm opposed to current LLMs on the ground that they've scraped vast quantities of stuff without permission nor compensation, not the status of the results.
Show thread位zydMar 14
@rpluim That's a good start.
Show threadbenMar 14

@rpluim @zyd My understanding is that although that鈥檚 true for something completely machine-generated, it doesn鈥檛 take much human input to make it copyrightable again. So although a completely vibecoded work probably wouldn鈥檛 be copyrightable, an LLM-assisted contribution might well be.

And besides, there is nothing against integrating public-domain code into free software as far as I know. It wouldn鈥檛 receive copyright protection but would be available to use by anybody.

I don鈥檛 think this means LLM contributions should be allowed, but I don鈥檛 think the copyright argument alone is strong enough to prevent it.

Show threadlhpMar 14

@zyd This is a pretty weak-ass response, quite disappointing.

Although rms calling LLMs "bullshit generators" in one the linked mails does give a slither of hope that he uses his powers in GNU[1] to stop LLM contributions. Although that conflicts with another of his mails to emacs-devel where he states he's looking into the copyright situation with a lawyer before making a decision...

Thank you for your efforts!

[1] which he really shouldn't have, based on *gestures wildly* everything

Show threadDekks HertonMar 14

@zyd

seeing how the tech bros scraped llm trainng data off the net regardless of licences ultimately any vide code is potentially tainted so will it ever be accepted?

Show threadbig hungryMar 17
@zyd vommit in Emacs scratch branch. Let's see how this goes
https://yhetil.org/emacs-devel/CALDnm5[email protected]/T/#t
Re: branch scratch/less-dubious-intern created (now 72596d17985)

Show thread位zydMar 17
@viz Jo茫o T谩vora also maintains several important Emacs projects like eglot and sly (alternative to slime). I wonder to what extent he has contributed LLM generated code into either...
Show threadbig hungryMar 17

@zyd his multiple LSP middleman thingy in python has a fair bit of vommit

I'm slowly losing respect for a lot of these Emacs people...

Show thread位zydMar 17
@viz Same friend, same.