[cml123]

Lately I've been seeing a lot of derision from the Emacs community of the consideration for integrating these kinds of tools with Emacs, but I truly think that's much more hurtful than helpful. Although the current development and usage of AI in software development may not closely resemble the techniques used at the time, it seems to me that Emacs' history is inextricably linked to the MIT AI Lab. It feels weird then that people today would shun the inclusion of AI integration into a tool that was produced from such a working group.

[uludag]

The beauty of Emacs though is that it puts the user in full control. There is nothing in the world stopping anyone from modifying anything in Emacs (at least on the Elisp layer), hence packages like this.

VS Code on the other hand is designed to fracture [1]. MS can and has given proprietary API access to their blessed tools, forcing others to go through their much less capable extension API, hence the plethora of vscode forks. Even if you had the most motivated, excited group of people wanting to work on the latest and greatest LLM interactions with VS code, they would most likely be forced to fork. On the other hand, it just takes one motivated Elisp dev to implement whatever they want and make any external package they want integrate with it.

Also, I think the derision from the Emacs community may be a bit overblown. I'm constantly seeing AI/LLM related plugins appearing for Emacs and they tend to get decent traction (e.g. https://github.com/karthink/gptel).

[1] https://ghuntley.com/fracture/

[paddy_m]

This is due to Richard Stalllman. He thinks that integrating "non-free" alternatives when free alternatives don't yet exist slows down free software development. Not just free as in freedom alternatives, not just free as in GPL licensed, but free as in FSF controlled projects. He did the same thing with linking extensions to GCC, LLVM debugger integration into emacs (fuzzy on that one), possibly treesitter into emacs, bzr vs git for emacs code source control, and a CI build farm for emacs. In each one of those cases, he eventually relented and the core project, eventually integrated the non-free alternative years later.

In the meantime this delaying didn't stop the non-free alternatives, but it did slow down adoption of the core project. This is attrocious project management that is driving people to non-free software. In the most egregious cases (bzr and CI build farm), it was done only because of his ego and wanting the FSF to matter.

[benreesman]

A lot of us are grateful in some abstract way for all the foundational work RMS did both technically and organizationally to preserve what remaining software freedoms we still have, but got off the bus a long time ago. He got really weird and it was on some "no fly zone" shit.

There's an `emacs` community that recognizes the history without being involved in any contemporary sense.

[bowsamic]

Why did you?

[7e]

Huh? Open source licenses long predate Stallman. He was, at best, an opportunist who tried to coopt the OSS movement and take it into a kooky ideological niche.

[7e]

Do you remember how the world got all kinds of weird cults before we got good at identifying cults and the phenomenon of cults? Well, the FSF/GPL is one of those. Many people still need to be deprogrammed.

[benreesman]

I never said that Stallman invented open source (he didn't) or that his motives for starting GNU up after the Symbolics fiasco were pure and high-minded (they were petty), or that the contemporary FSF is a force for good (it's not, c.f. glibc dynamic link lock in on a backdoored resolver SONAME chain and Drepper's weird ties).

I said he did a lot of foundational work that's still important today.

I acknowledged a contribution, I didn't beatify him.

[skydhash]

Where would they integrate it. Emacs is a small core of C code. Almost everything is Elisp and in the same standing as third party packages. I’m not seeing what being in emacs core brings to an AI package?

[spauldo]

It'd be no different than eglot, project.el, etc. Third party packages experiment with stuff, then a stable implementation appears in core.

[paddy_m]

That would be a reasonable stance if the difference were in incubating a new project vs excluding functionality from blessing because it interfaces with non free software. The functionality I'm talking about is excluded because of the latter.

[spauldo]

In that case it'd just live on in Melpa regardless of what the mailing list thinks.

[paddy_m]

Well the linking into GCC was a C code issue. For emacs, there is a large collection of elisp that is shipped with the official package. Preventing worthy enhancements of that core package solely in the name of a distorted view of freedom hinders emacs and the adoption of emacs.

[Guthur]

You clearly misunderstood the problem.

The entities he is so adamant against are not benign or passive, they actively try to capture your freedom for rent seeking behaviour.

Microsoft, Apple, Amazon etc, have not got to where they are without this behaviour and they are so powerful that they have in many cases captured even public money from large governments for decades and are exceptionally sticky once allowed in.

LLM provide an exceptional opportunity for us to free ourselves from these captor interests, but we need to looking to develop them.

RMS has been proven correct on so many things from standard Microsoft behaviour, and planned obsolete to the licensing rug pulls of so called open source projects.

The question is not about stopping non free, that's a ridiculous objective, but if you don't have any principles you are going to have nothing solid to stand on in response to their nefarious and extractive behaviour.

[spauldo]

The objective is very much to stop non-free software. It always has been. It's not secret - it's explicitly why the FSF exists.

And that's a good thing, for the most part. Someone needs to hold the hard-line stance. It'll never happen, but it pulls things in that direction. We're all free to do what we like and use whatever software we choose, and part of the reason we have that choice is because the hardliners refuse to budge.

It does mean they make unrealistic demands and occasionally hold back useful functionality, but it's better than not having them around.

[paddy_m]

I haven't misunderstood the problems that RMS talks about, I agree with his prescient analysis. I firmly disagree that RMS the person is the best person to lead a software producing organization that aims to deliver a free future.

A simple reductive example. Imagine a great software leader that leads an org that writes great code and generally achieves the org's goals, but once a week, they say something offensive that discourages 1/10th of new users. A better leader would be someone who does all of the same things, except for the offensive comments.

I am saying that RMS makes offensive distracting comments, and regularly makes project manager choices that slow the adoption of free software. If you criticize him, people come back to "but he's right philosophically" which he is, and that misses the point. He has wrapped the FSF into an ego play for himself where he is in control or at least an important roadblock to software progress. If RMS cared as much about software freedom (as opposed to his ego) as he says, he would work to allow better leaders to develop and have power in the FSF org.

[qiine]

this is just sad

[spauldo]

The Emacs community is incredibly diverse. You'll find derision for just about everything if you look for it.

I'm guessing there's a lot of grumbling on the mailing list about non-free AI services. That's fine, you can ignore that. 3rd party modules will provide, and there's nothing core can do about it.

[bowsamic]

I didn’t know MIT AI lab was involved in the modern AI boom, that’s interesting

[benreesman]

Asionometry has a great video on it: https://youtu.be/sV7C6Ezl35A

The Levy book Hackers has a ehole third of the book about it.

[bowsamic]

Haven’t got time to watch that but that book seems quite old. Are you sure it talks about LLMs?

[benreesman]

They weren't called LLMs, but they had neural networks and hardware optimizations for AI and huge teams of people tirelessly labeling stuff to make it look smarter than it is :)

There is some surprise factor at the GPT-3 -> gpt-4-1106 jump for people who know the history of AI generally and who were around a lab during the ImageNet days, but not as much as everyone is acting like.

The last two years are a notable but by no means unprecedented rush towards the next wall. There's even a term for it: AI Summer is short and AI winter is long and its about as predictable as the seasons in Game of Thrones / ASOIAF.