Mastodawn

The ongoing discussions about whether to ban #LLM code contributions to #Drupal core can be summed up as:
Antis: Using the child-stabbing machines is bad and we shouldn't do it.
Pros: It's fine to stab children in limited and considered ways. Lots of other things are bad for kids. And how would you even know if people have stabbed children? #FuckAI

Jürgen Haas 1d ago

@joachim
I need to disagree on this. What I do agree with is that there are far too many technologies, not just LLMs, that are threatening, and we as human beings urgently need to deal with the issues.

But making people assume that Drupal banning the use of LLMs would solve any of these issues isn't really responsible.

joachim 21h ago

@jurgenhaas It's not about solving these issues, and I think it's disingenuous and derailing to say that. It's about not having blood on our hands.

@jurgenhaas Actually, it's worse than that. It's pessimistic and cowardly. Boycotts work.

Jürgen Haas 14h ago

@joachim
If we applied that threshold, we would have to do the same to cars, meet, big tech, banking, religions, and so many more threatening and/or unethical products, services, or businesses.

The result would be nothing else than back to the island, and nothing noticeable would happen or change.

I find that unrealistic and pessimistic. Instead, let's be better than those rather then banning or ignoring them. E.g. learn from today's LLMs and push their open source alternatives.

jpoesen | 🇪🇺 | 🏳️‍🌈12h ago

@jurgenhaas @joachim

Banning baby grinders will not solve the baby killing problem. Let's learn from today's baby grinders and push their open source alternatives.

#EthicalBabyGrinding

jpoesen | 🇪🇺 | 🏳️‍🌈12h ago

@jurgenhaas @joachim

> "we would have to do the same to cars, meet, big tech, banking, religions, and so many more threatening and/or unethical products, services, or businesses"

You make that sound like a bad thing.

So if we can't stop all evil at once we shouldn't try to stop any? Sorry, but that's textbook whataboutism.

https://en.wikipedia.org/wiki/Whataboutism

Whataboutism - Wikipedia

Jürgen Haas 7h ago

@jpoesen
@joachim
I would agree if my proposal were to do nothing about it. However, I not only share the anger and the concerns about the emerging technologies, I also advocate for doing whatever we can to eliminate them.

It matters to me, though, that we chose carefully which counter measures to take. And when it comes to LLMs or big tech, calling for a ban will achieve nothing other than us finding ourselves on an island and nobody cares.

Building alternatives will have an impact.

jpoesen | 🇪🇺 | 🏳️‍🌈6h ago

@jurgenhaas @joachim
I believe that the tech we call AI / LLM / GenAI:

1) is built on stolen content & hidden exploitative labour

2) is fundamentally flawed tech

3) has far ranging detrimental impact on societal cohesion & the environment

As such, this tech is *poison*. I consider *any* software that uses it to be fruit of the poisonous tree, similar to the "virality" of licenses like the GPL.

No amount of sourcewashing can change this. The number of babies I'm willing to grind is zero.

jpoesen | 🇪🇺 | 🏳️‍🌈6h ago

@jurgenhaas @joachim
Jurgen, that said I'd love to better understand what you advocate eliminating, which kind of counter measures you're referring to, and what alternatives you envisage.

I don't mean to draw you out here, I'm just unclear as to what kind of middle ground others do find acceptable.

Jürgen Haas 6h ago

@jpoesen @joachim

I'm advocating for eliminating the threats and the business models of the evils. And Open Source is yet again part of the measures that help us to get there.

jpoesen | 🇪🇺 | 🏳️‍🌈5h ago

@jurgenhaas @joachim

> "I'm advocating for eliminating the threats and the business models of the evils"

I get that. Specifically?

Jürgen Haas 5h ago

@jpoesen @joachim
Open Sourced LLMs are a thing, and they're emerging quickly. Energy consumption can be massively optimized, and distributed resources become capable of hosting and running those models. There is a clear path to sovereignity and we should soon be able to responsibly using AI without having to rely on OpenAI, Anthropic, and others. Imagine, being in control, and being in charge, benefiting from extremely powerful technology. Why not !!!

@jurgenhaas @jpoesen What's the energy requirement actually going to be, including training the models?

@jurgenhaas @jpoesen And it's not powerful technology! It's crap! Having to plead the thing with phrases like 'You are a smart developer who makes no mistakes.' I mean, it's utterly ridiculous! You've all fallen for magic beans here!

Jürgen Haas 5h ago

@joachim @jpoesen

I can demonstrate that this is untrue. And that flattering is easy to ignore. With the right tool (e.g. @opencode and @codenomad) you can hide them and focus on your work. It's me being the engineer, and the agent being my coder(s). You can't just say that's crap, I've done it, and created enterprise level software with it, you can't distinguish it from code that has been hand-coded.

Jürgen Haas 5h ago

@joachim @jpoesen

That ship has sailed.

jpoesen | 🇪🇺 | 🏳️‍🌈4h ago

@jurgenhaas @joachim
Let's say your predictions are 100% correct.

Until such time, the *current* tech should not be on the market at all.

By all means *research* the everloving shit out of this exciting technology, but don't make it available until it actually is safe and resource friendly and ethical.

Until we can replicate cat food without grinding babies I don't think we, as a society, should be replicating cat food *at all*, even if there will soon be a clear path forward...

Jürgen Haas 6h ago

@jpoesen @joachim

To 1: not necessarily. GPL licensed code allows explicitly that learning from it is allowed.
To 2: don't think so. Thy hype is, but reducing it to an effective tool (e.g. a modern calculator) is grounding this and puts the magic aside. It's just a tool.
To 3: Responsible usage is not. And the resource issue needs to be fixed in any case. And we can, there is unlimited free energy available, we just have to go ahead and drop all fosils etc.

To be continued ...

Jürgen Haas 6h ago

@jpoesen @joachim

... All that LLMs are doing is data processing to an extend that we haven't seen before. At least for the use cases that we're talking about here: assisted coding.

Kevin + Drupal + Beard 5h ago

@jpoesen @jurgenhaas @joachim So much going on here, but I have to take issue w/ comparing the intentionally infectious design of GPL to the fruit of LLM/GenAI.

I understand the argument that SFC has been making since its https://sfconservancy.org/GiveUpGitHub/ campaign and the impact accepting that if LLMs just interate enough derivatives, the GPL no longer applies will have on the FOSS ecosystem, but I reject the idea that AI cannot be used responsibly within FOSS projects w/o infecting everything.

Give Up GitHub - Software Freedom Conservancy

The Software Freedom Conservancy provides a non-profit home and services to Free, Libre and Open Source Software (FLOSS) projects.

Kevin + Drupal + Beard 5h ago

@jpoesen @jurgenhaas @joachim I also reject the idea that AIs environmental impact is fundamentally different than the ever increasing resources all tech requires... or that the market won't reward improved efficiency in this space like it has before. Every Drupal & WP site running a variation of the LAMP stack requires orders of magnitude more resources that SSG. In many cases we keep these stacks running for months with no edits for the convience of near real time editing when needed.

jpoesen | 🇪🇺 | 🏳️‍🌈5h ago

@kreynen @jurgenhaas @joachim
Valid points.

Correct, dynamic sites use more resources than static ones; many dynamic ones could/should be static ones. My own site is a local Drupal 11 + Boost that generates & deploys a static site. Not sure how this is pertinent to AI resource usage.

It's probable that the market will reward increased AI resource efficiency. Until that time, the current ginormous usage shouldn't be brushed off with technosolutionist "it'll change for the better some day"

jpoesen | 🇪🇺 | 🏳️‍🌈5h ago

@kreynen @jurgenhaas @joachim

> "I reject the idea that AI cannot be used responsibly within FOSS projects w/o infecting everything."

Given that it is my position that current "AI" is 100% societal poison (in a nutshell), I don't see how I can consider "responsible use of AI" as anything else than a contradiction in terms...