- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
GrapheneOS is currently defending its use of AI coding tools on Mastodon against complaints by various accounts claiming to be users.
We do not understand where you’re coming from or why you’re so incredibly angry with us. It’s not justified and does not make sense.
OSS projects have already started to recognize that LLM is just a tool and how you use that tool matters a lot more than whether you are using said tool at all. Lets bring StackOverflow for analogy: we didn’t ban browsers for code development because one can navigate to StackOverflow and copy-paste some really crappy code. LLMs are SO on steroids (actually trained on SO topics). So grabbing LLM to get a sense of direction or shape is an OK use in general, but copy-pasting it’s results blindly is silly and dangerous.
There’s a second argument that AI is ruining OSS communities which I believe is true, but refusing to accept AI code is a self-defeating action. Why? Because AI training needs unadulterated source of knowledge, one not tainted with AI and OSS communities rejecting AI would be that source. Also: if OSS community did not use AI it would hardly stop it from continuing it’s growth in Enterprise. So there’s not much of a winning strategy there either.
Which brings us to GrapheneOS decision and given above reasons I think it’s a well-balanced decision.
a lot of communities are aggressively clamping down on AI contributions. i think leaving the floodgates open really hasnt worked out. People touting “the kernel does it so get over it” gloss over the contributions being small and strict “your ai code reflects you and you own the consequences” policy.
additionally i dont buy the enterprise does it so get on board argument. the floss movement is literally built around pushing back on corporate norms. i dont know why anyone would take thay argument at face value.
for a lot of FLOSS communities, the ethics of the software is as important as the quality. which is why so many Debian community members felt betrayed by milquetoast ai policy adoption.
Graphene made a similar “look they’re all doing it too” argument on Mastadon which I feel just really undermines their image as the alternative.
“we dont police our developers usage as long as the quality meets out high standards. all code is at the end of the day ownedand reviewed by the humans in our teams” or some other equally silly statement would still have had critics but at least they wouldn’t sound like a grade schooler wining about how infair life is.
Is it just a tool really?
Using a hammer doesn’t atrophy the muscle in your arms as you use it.
However using AI has already been shown to atrophy the brain.
https://www.media.mit.edu/publications/your-brain-on-chatgpt/
You can achieve wealth and happiness way faster if you rob all other countries of their resources. But if it is feasible, must it still be done ?
Kind of the same thing with LLM. It’s not because you can that you should.
As you said, LLM is killing OS communities. It’s preventing people from working together, now they are all “I don’t need you, I can do it my way all alone”, and you get 50 clones or similar softwares, which aren’t finely tuned since it’s all LLM.
As you didn’t say, LLM has a huge impact on the planet, it costs a lot of electricity, it uses clean water which it evaporates in data center, but we don’t care because guess what, we have it in our hands and it’s so much fun to use ! And well, we ourselves are not responsible for the planet’s state, it’s the big bad company that create LLM, I’m just using it because otherwise, the other will do so !
What’s next ? After coding our own little games and softwares alone and ultra-quickly I mean.
Who pays the bills ? Who will pay the next one, and where is humanity going ? Not sure I want to know.
AI Searchers are quitting their jobs being scared of what they enabled Companies are being hacked by LLM without control People are losing their jobs because of LLM (and I think those are not the ones using LLM in the OS Community, because oh, the irony it would be)
Society clearly isn’t ready for any of this, and saying “don’t care” doesn’t make it fine.
Edit : this is cleary not on or against you, I’m simply expressing my frustration on this subject with arguments and examples I hope should make people realize things x)
Yeah, just one small problem, muchacho. Whenever a language recieves a breaking change you gotta retrain the model again, and a lot of your references are useless now. There is no “improving” it’s just a vicious cycle.
Your argument is that of peer pressure and not compelling.
“Join us or we’ll just steal your code anyway”
Here’s my rebuttal. Garbage in, garbage out. Most projects out there are trash. This tool doesn’t help you build good code, it helps you throw together a pile of trash. An actually helpful LLM would be one trained on the damn language itself and coding best practices. Almost everyone has lost the plot here. You DON’T want a massive LLM, the small specialized ones are good enough. So you can use it as auto complete, like the tech was intended.
You’ve made a few of erroneous conjectures which are orthogonal to my argument. But I’ll call out one: LLM is bad at code completion and it’s use not have to be limited to (final) code writing. Thus my theses on “what” vs “how”
Give me a break. Put down the thesaurus. You are the reason LLMs are so.useful. because, as one searches for a solution to the update that borked one’s servers, we find that 4 out of 5 forum threads on the topic devolve into jerks weaving unhelpful riddles, smelling their own farts and feeling superior.
LLM is just fine for next text prediction, in the same way my phone keyboard can sometimes guess what I’m about to say. But the notion of building this huge mass is where I draw the line, it’ll never be finished. Does overfitment mean nothing to anyone anymore?
your points support my statement though: how vs what is the real question here. LLM is fine performing repetitive mundane tasks that we find boring and uninteresting. Like refactoring small chunks of code out or “search and replace” when context matters more than matching string etc. Also, let’s assume someone had enough bitcoin mining gear laying about to repurpose it into LLM training infra and trains his/her small model on stuff they care for, suddenly we lose all the other arguments and use of LLM becomes more acceptable. So blanket ban at this point is not going to achieve any of the goals, really. Consider that majority of developers were mandated to use LLMs for work, consequently their skill shape changed, for project to draw from that pool of developers you need to offer comparable tools.
Context matters too: Graphene has to deal with AOSP, which lost it’s development transparency a while ago and Graphene developers likely have to contend with big (likely LLM-generated) code dumps within mainline AOSP with each new release, so to keep up with that onslaught they either need to expand developer base (and we just said that pool is shrinking) or leverage LLM to cut through that mess. Again “how” tool is used matters a great deal.
When people say they don’t want AI in their codebase, they’re talking about Claude, and GPT. I mean, we’ve had intellisence for ages. To a certain extent, we’re both fighting strawman arguments.
I get your point now, but I’ve personally had about 6 of my projects at work ruined due to AI slop, to the point that I don’t think my company is going to make it. I’m the odd one out, everyone else is loving it, despite our productivity grinding to a hault.