I'm not an emacs user, but it is heartening to see firm declarations like this pledging to protect fundamental development tools from slop and its enthusiasts.
The first thing is, the correct course of action here is very easy. The fact that they haven't already done the very easy thing is the first red flag.
If you look at other GNU projects, there is a lot of pushback to a policy that is grounded in ethical behavior primarily. Even guix, which I would say has the project leadership I trust most among all GNU projects, is still deliberating what to do about it: https://codeberg.org/guix/guix-consensus-documents/pulls/13
GNU historically has been myopically focused on license and copyright at the expense of every other consideration. Part of the concern is that some LLM enthusiasts will come up with an argument that the licensing situation is actually fine, and then since GNU can't see beyond licensing, there will be no further discussion of the myriad other problems.
Of course, it's certainly possible that GNU will do the right thing and all this will be for nothing. That is the ideal outcome we are hoping for.
There's some people in the community that think that if it's an open weight model then it's good enough and we should use it. I was very surprised to see that one of the most important news sites about Linux in Spanish had a piece talking about "why do you hate AI so much?"0 while trying to argue (badly) why the critics are wrong and that "open source AI (sic)" should be welcomed.
PS: I'm a no one in the Emacs community, just a user. Can I sign the petition?
Yes; this is for users and contributors both. The site explains how to get added but if you message me a name and optional URL I can add you that way too.
Most likely no. FSF is super cautious with this from what I hear from some FSF people. The difference here is that this is not an ethical problem according to GNU's philosophy (at least explicitly it is not, many free software devs are against LLMs for ethical reasons.) But concern about copyright-ability of such code, which could nullify the copyleft.
Still, we have seen too many projects doing this. I think it is necessary to prepare before it happens.
I think you're confusing it with another project? I couldn't find anything like that in the recent stories. I did however find confirmation of LLM use.
... their effectiveness is not at all relevant to the question of whether their use can be part of a principled software movement dedicated to user empowerment.
To paraphrase: I want to stay the gatekeeper because the muggles have dirty hands, disguised as user empowerment.
In my experience users who have access to LLM do feel empowered. They finally can finish up toy projects, get their 3D Printer working by fixing some weird lua bug in automation scripts, clean up their Emacs config, ect...
I'm not seeing an actual ethical argument. It seems to assume the conclusions without an explicating ethics. Which is fine! People are free to use this, if it fits their ethical framework.
LLMs will improve Emacs in areas where it hasn't been possible before. They will enhance it, as has already happened with very complex bugs and abstraction layers that required a lot of work.
We are not here to discuss whether LLMs are effective at what they are claimed to be able to do; their effectiveness is not at all relevant to the question of whether their use can be part of a principled software movement dedicated to user empowerment.
What complex features are you thinking that couldn't be solved using elisp and require a change to the core?
Emacs is very permissive and elisp very capable.
Users, including non-devs, will remain free to use as much LLMs as they wish to build any features out of elisp, on top of human-emacs. They simply won't be able to submit LLM generated code to repository, for ethical reasons.
I don't see why would someone use human-emacs over emacs, and then use LLMs to write elisp, but hey, each to their own.
Beside, nothing stops anyone from modifying emacs however they wish. That's the whole point of free software. Get the code, patch it using LLM or don't. You choose. It's free.
Your argument isn't very convincing (with respect!)
I think that deep down there's more human fear than a real problem. The licensing argument seems like an excuse to me: today you can work with a free, open-source LLM, running it on your own machine.
The project's stated position is explicitly presented as ethical, not practical.
It's all about what you believe in, where your values lies.
You may not share those values, and that's fine; that's the beauty of free software. You can use as much LLM as you want, or even start an "llm-emacs" and rally people to your cause. No judgement here. Whatever float your boat.
But calling them an "excuse" assumes bad faith, which isn't nice. You're entitled to your opinion; so are we.
Besides, the "fear" argument is weak as it cuts both ways: one could equally say the use of LLM is driven by fear of missing out or "falling behind". It doesn't really get us anywhere.
I dislike the "us vs. them" mentality LLMs have brought to the industry. What is so peculiar about this technology that we must all be classified as pro or anti, and relentlessly attack the other "side"?
If anything, what we should celebrate here is the idea of free software itself. We don't share the same values and opinions, and yet we can share the same software. How great is that?
As we say in French: "La liberté des uns s'arrête là où celle des autres commence."
You're right: "excuse" was a poor choice of word, no bad faith intended. "Legitimate but easy to work around" is what I meant. Fair point too on fear cutting both ways: fear of LLMs degrading what we love, and fear of falling behind. Neither side is free of it. My actual disagreement is narrower: I don't share the premise that LLMs are incompatible with user empowerment (for someone who can't program, adding a complex feature to their editor is precisely that), and a policy that can only rest on voluntary disclosure ends up punishing honesty, as the recent emacs-devel patch rejection showed. Longer, calmer version here: https://en.andros.dev/blog/b70b058a/human-fear-in-emacs/
Internet_Janitor | 15 hours ago
I'm not an emacs user, but it is heartening to see firm declarations like this pledging to protect fundamental development tools from slop and its enthusiasts.
hugoarnal | 17 hours ago
Why’s that? Is GNU going to eventually allow LLMs (very mostly non-free) near any of their software?
technomancy | 16 hours ago
It's a fair question.
The first thing is, the correct course of action here is very easy. The fact that they haven't already done the very easy thing is the first red flag.
If you look at other GNU projects, there is a lot of pushback to a policy that is grounded in ethical behavior primarily. Even guix, which I would say has the project leadership I trust most among all GNU projects, is still deliberating what to do about it: https://codeberg.org/guix/guix-consensus-documents/pulls/13
GNU historically has been myopically focused on license and copyright at the expense of every other consideration. Part of the concern is that some LLM enthusiasts will come up with an argument that the licensing situation is actually fine, and then since GNU can't see beyond licensing, there will be no further discussion of the myriad other problems.
Of course, it's certainly possible that GNU will do the right thing and all this will be for nothing. That is the ideal outcome we are hoping for.
aarroyoc | 14 hours ago
There's some people in the community that think that if it's an open weight model then it's good enough and we should use it. I was very surprised to see that one of the most important news sites about Linux in Spanish had a piece talking about "why do you hate AI so much?"0 while trying to argue (badly) why the critics are wrong and that "open source AI (sic)" should be welcomed.
PS: I'm a no one in the Emacs community, just a user. Can I sign the petition?
algernon | 13 hours ago
I'm a mere user, and have signed it. I think it is important that we sign it too.
technomancy | 12 hours ago
Yes; this is for users and contributors both. The site explains how to get added but if you message me a name and optional URL I can add you that way too.
orib | 7 hours ago
The earlier and more loudly people say "if you do, we're leaving", the less likely it is to happen.
[OP] tusharhero | 16 hours ago
Most likely no. FSF is super cautious with this from what I hear from some FSF people. The difference here is that this is not an ethical problem according to GNU's philosophy (at least explicitly it is not, many free software devs are against LLMs for ethical reasons.) But concern about copyright-ability of such code, which could nullify the copyleft.
Still, we have seen too many projects doing this. I think it is necessary to prepare before it happens.
Garbi | 14 hours ago
rsync have aggressively adopted LLM
I thought we just read about rsync contribs that were rejected due to LLM smell.
jmiven | 14 hours ago
I think you're confusing it with another project? I couldn't find anything like that in the recent stories. I did however find confirmation of LLM use.
kalkin | an hour ago
To paraphrase: I want to stay the gatekeeper because the muggles have dirty hands, disguised as user empowerment.
In my experience users who have access to LLM do feel empowered. They finally can finish up toy projects, get their 3D Printer working by fixing some weird lua bug in automation scripts, clean up their Emacs config, ect...
jfb | 3 hours ago
I'm not seeing an actual ethical argument. It seems to assume the conclusions without an explicating ethics. Which is fine! People are free to use this, if it fits their ethical framework.
aloys | 3 hours ago
See https://codeberg.org/guix/guix-consensus-documents/pulls/13/files for details.
jfb | 3 hours ago
Thanks, this helps.
jlarocco | 10 hours ago
This is a bit presumptuous - Emacs isn't accepting LLM-generated contributions until GNU decides on it.
kreeft | 9 hours ago
The page is written with that in mind, and they bring it up in the third(?) paragraph.
andros | 12 hours ago
LLMs will improve Emacs in areas where it hasn't been possible before. They will enhance it, as has already happened with very complex bugs and abstraction layers that required a lot of work.
pmbauer | 11 hours ago
The motivation is all about ethics.
quoting from the Not Under Discussion section
rau | 8 hours ago
Seems to me that it’s pretty empowering for a non-developer to be empowered to add more complex features to Emacs.
aloys | 8 hours ago
What complex features are you thinking that couldn't be solved using elisp and require a change to the core?
Emacs is very permissive and elisp very capable.
Users, including non-devs, will remain free to use as much LLMs as they wish to build any features out of elisp, on top of human-emacs. They simply won't be able to submit LLM generated code to repository, for ethical reasons.
I don't see why would someone use human-emacs over emacs, and then use LLMs to write elisp, but hey, each to their own.
Beside, nothing stops anyone from modifying emacs however they wish. That's the whole point of free software. Get the code, patch it using LLM or don't. You choose. It's free.
Your argument isn't very convincing (with respect!)
andros | 3 hours ago
I think that deep down there's more human fear than a real problem. The licensing argument seems like an excuse to me: today you can work with a free, open-source LLM, running it on your own machine.
aloys | 3 hours ago
Andros, I think this misses the point completely.
The project's stated position is explicitly presented as ethical, not practical. It's all about what you believe in, where your values lies.
You may not share those values, and that's fine; that's the beauty of free software. You can use as much LLM as you want, or even start an "llm-emacs" and rally people to your cause. No judgement here. Whatever float your boat.
But calling them an "excuse" assumes bad faith, which isn't nice. You're entitled to your opinion; so are we.
Besides, the "fear" argument is weak as it cuts both ways: one could equally say the use of LLM is driven by fear of missing out or "falling behind". It doesn't really get us anywhere.
I dislike the "us vs. them" mentality LLMs have brought to the industry. What is so peculiar about this technology that we must all be classified as pro or anti, and relentlessly attack the other "side"?
If anything, what we should celebrate here is the idea of free software itself. We don't share the same values and opinions, and yet we can share the same software. How great is that?
As we say in French: "La liberté des uns s'arrête là où celle des autres commence."
andros | 3 hours ago
You're right: "excuse" was a poor choice of word, no bad faith intended. "Legitimate but easy to work around" is what I meant. Fair point too on fear cutting both ways: fear of LLMs degrading what we love, and fear of falling behind. Neither side is free of it. My actual disagreement is narrower: I don't share the premise that LLMs are incompatible with user empowerment (for someone who can't program, adding a complex feature to their editor is precisely that), and a policy that can only rest on voluntary disclosure ends up punishing honesty, as the recent emacs-devel patch rejection showed. Longer, calmer version here: https://en.andros.dev/blog/b70b058a/human-fear-in-emacs/
aloys | 2 hours ago
Nice article, I like your website, I came across it a few times.
I'll leave a comment on your blog directly, as to not go off topic here!
andros | 2 hours ago
Thank you! You're always welcome 😊