I locally host gemma 4 e2b on a very modest spec laptop to help summarize large texts, as a rudimentary translation tool, and just as a toy. It makes sense. I see small coding models on hugging face all the time, I think we’ve hit a point where it’s just too efficient not to atleast entertain a little bit of slop in your workflow.
TL;DR: Merged code has to meet existing quality standards. What tools were used to create it is irrelevant.
Fuck you, Linus Torvalds.
Truthfully I feel like this too sometimes. I get he mad eit and I understand that but what happened to the “community” and “open” part of it why does this dude get to call all the shots on everything related to Linux.
Reminds me of Satya Nadella with Windows. (Even though he didn’t invent it but calls the shots just the same way.)
Dude, this is lemmy. The headline should be: “AI Slop Allowed in the Linux!”
with that, they can do the open-source thing and fork it.
Ok regardless of your opinion on AI this is just stupid. No one is forking and maintaining the entire linux kernal.
The straw manning in this thread is amazing.
Linus allowing AI? This means the linux kernal will be flooded with unverified slop!
For all his great work on Linux (and git and others) Torvalds was never into activism so much. He declined GPLv3 (yes it would be impossible to get all contributors to sign for it). He’s always been an “open source pragmatist”, which is fine I guess. And I really respect his work. But at these important political points, we need someone like Stallman who has consistently proven to be on the right side of the debate
This. Google would not be fucking us with Play Integrity and other DRM craps if Linus made the kernel GPLv3 licensed. It would have been a dreamland to have hackable phones, TVs and a whole lot of IoT devices where the users would be controlling them instead of other way around.
Nah, they just would have either used something else, or fucked us with some other way. The kernel license just allowed them to do it like this.
Hot Take: IMO, using generated code is fine, if it goes through the exact same due diligance as normal code. (unit tests, is the algo optimised, etc.)
Only if you don’t care about your own cognitive decline, property theft, and climate damage
the destruction of the consumer electronics market …
This is like saying you should dig with bare hands even though drilling machines have been invented.
I feel like people parroting these tropes uncritically are the ones who should be worried about their own cognitive decline.
Are we wrong tho, and why?
The energy consumption trope in particular hasn’t been true for a while now. You can literally run models on your laptop for most tasks, and these are models that have capability that needed a data centre literally less than a year go. Meanwhile, the whole notion of people offloading their cognitive capacity to models is based on a handful of studies with tiny samples. So yes, you are wrong, and you just run around uncritically repeating nonsense thinking you’re being really profound.
Such a strong science backed rebuttal.
I’m gonna go rethink my entire life now.
Being aware of the numerous massive lawsuits currently being levied against AI companies for their theft, reading research papers on cognitive decline in AI users, and being aware of where the majority of energy in the US comes from, especially for new AI datacenters, really does have a cognitive impact, but it isn’t decline.
As strong a rebuttal as a parrot requires. I also love how you lump together a whole bunch of issues inherent in capitalism in your complaint further illustrating that you’re not able to put together a coherent argument.
Programmers already spend about as-much of their time, if not more, reading other’s code as they do writing their own. It doesn’t mater if that “other’s code” is AI generated or not.
There’s an argument to be made about excessive vs not-enough commenting, but that’s not where you went, and its clear you have negligible programming experience, or creative experience for that matter, to be coming after the concept of sharing code like-so.
One wonders how many books you’ve read, to be pretending that reading a book without paying for it, even borrowing from a library, is theft. Stick with the environmental costs arguments - its what you are personally suited to argue, and far more urgent than the rest.
Programmers already spend about as-much of their time, if not more, reading other’s code as they do writing their own.
And how is this an argument in favor of AI code? If 5-10 percent of you job used to be writing code and now you dropped it by half, that’s not very effective of optimization. Especially if the code you now have to review takes more time to review due to it being generated by AI.
and its clear you have negligible programming experience, or creative experience for that matter, to be coming after the concept of sharing code like-so.
As someone with several years of experience as a programmer: fuck off with this elitist nonsense
Lol. Way to show how little you know. I swear AI bros get the most butt hurt over provable stuff.
I’ve stated my experience elsewhere but I’m a senior software engineer, 20 something years of experience. I have actually created a neural network from the ground up for a previous company. But sure, go off about how little experience I have.
Only if you don’t care about your own cognitive decline
That sounds like Socrates’ argument that writing would weaken people’s memories.
And absolutely, people will probably forget the syntax by heart if they don’t type it as frequently. Personally, knowing syntax is not very valuable to me, as it’s just a means to an end. And whether that leads to cognitive decline or not, is really up to who’s using the tool.
Saying it leads to cognitive decline is saying you can’t use an LLM and have critical thinking, which I can’t agree with.
Socrates didn’t have access to modern science. This shit rots your brain.
That sounds like Socrates’ argument that writing would weaken people’s memories.
Saying it leads to cognitive decline is saying you can’t use an LLM and have critical thinking, which I can’t agree with.
That’s not me saying it. MIT, Harvard, and others have released numerous studies that show using LLMs does exactly that, reduce critical thinking. You can disagree all you want but until you do the science you’re disagreeing with a growing body of actual experts.
Are those studies in the context of software development? What are the tasks at hand? Do they evaluate critical thinking on matters that people actually care about or on chores? Were they instructed to use LLMs in a particular way that is equivalent to their personal preference?
You can’t pull a wildcard saying something like that because it’s too broad of a conclusion.
This is where you go look at any of the studies and start figuring out for yourself.
They’ve been looking at it in several contexts including software development, general problem solving, reading comprehension, writing ability, and more.
In some they were instructed to use LLMs certain ways, in others they weren’t. That’s the neat thing about so many studies being done is they’ve used a wide set of methodologies.
well, you didn’t link any and you’re the one generalizing it, so the proof is on you to provide. I doubt their conclusions are like you’re making them sound.
I didn’t link any because the internet exists. Spend literally five seconds on google.
https://www.mdpi.com/2075-4698/15/1/6
https://arxiv.org/abs/2506.08872
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6097646
Their conclusions are right in the abstract in black and white terms. And this is just a teeny tiny sample of the papers that exist all saying the same thing.
Yeah, they never are. Scientific conclusions are generally a lot more nuanced and cautious than what this guy’s claiming. He only read headlines.
Firmly agree. I spent 6 years coding in python as a daily language, then I swapped to using nodejs for 10 years, only using python on and off. I went to make a basic script with python the other day and I had to look up how to convert a set over to a list.
If you don’t use it, you lose it is fully valid.
Not true. You came into the problem knowing:
- I need to process data according to rules
- Python can solve this
- I need a set data structure
- I need to convert that to a list data structure
- …
That’s the thoughts of a good engineer. You don’t think in code and you shouldn’t have to. Loosing your memory on the particular syntax is hardly an issue. You’ll still churn out a small program in under a day. You need to remember that people who really don’t know how to code take months for a product half as good.
Why should anyone seriously concern themselves with memorizing all the syntaxes? That’s like memorizing all the dates in History class. It takes so much bandwidth away from other concerns with higher payout. Go learn some architecture, some risk management, stuff like that.
You’re the proof that it’s true.
Sorry. :D
Don’t be sorry, that poster walked right into it. Have you met anybody who actually checks code before just putting on production servers? Maybe, but the shareholders need moar monies guy. Proof-reading code is so 2010’s.
Exactly, the argument that whether the code was written entirely by hand or produced by an LLM is the wrong thing to focus on. To see why, we have to consider how software development actually works at scale.
There’s a view that code written by hand has to be more intentional, almost has to be by definition since it requires the maintainer to actually put it in there themselves. That’s, of course, true but once a project grows past a certain size or it has multiple maintainers, nobody really has the totality of the code in their heads. So, any new code that’s added is always done with limited understanding. Code being written by hand should not be equated with it expressing the intent faithfully; if that were the case, then we’d never have software bugs. Humans make mistakes all the time as is clearly evidenced by there being no lack of buggy code predating LLM use.
I’m also not intimately familiar with most of the code in the projects I’ve been maintaining over the years. Any code I’ve written even a few months ago might as well have been written by someone else. When I need to make changes, I read through the code and figure out what it’s doing, and I rely on the test harness to make sure I don’t introduce regressions.
It’s simply not feasible for humans to keep the entirety of large projects in their heads all at once. When you’re working on a project, you’re constantly forgetting and relearning code as you go. And the situation is even worse for projects where multiple people work together where nobody knows what everyone else was thinking. We look at the code and try to build up sufficient context in our heads to make the necessary changes. When we misjudge that context or misunderstand existing code, then we end up making mistakes.
The way we judge whether projects are actually solid is by the level of specification and testing they have, the experience of the developers, and amount of usage they see in the wild. All of these same tools work just as well with LLM generated code as they do with code written by hand.
Farming out design decisions to the LLM without reviewing the output or doing proper testing will almost certainly produce low quality code, but that is no different from somebody just slapping some code together to make a kludge rather than really thinking through a problem. Working with LLMs does not mean farming out your thinking to the machine. What these tools actually do is automate the mechanical aspect of producing the code. Once it is written, you can read it, understand it, and change it as you would with any other code.
I seen alot of people say “AI is just a tool” ngl There is nothing wrong with that it’s just a little overused
There is nothing wrong with that it’s just a little overused
That’s one of the problems of companies devaluing and abusing language so much that words are in danger to lose any meaning. It is a difference however whether Torvalds uses such a word, or a corporation advertising bullshit. It is really clear that corporations want to replace software developers.
Maybe your right
It feels like marketing speech to me at this point. Like a new form of “it’s the newest model.”
Ehh
Lmao. This went over exactly how i expected here. Oh well ludites gonna ludite.
Because it’s not like natural intelligence is always all that great either.
Wouldn’t be a proper Linus post without calling everyone else an idiot.
I’m sure he included himself in that group
it’s a safe assumption that he didn’t given his track record of drama.
Didn’t call everyone an idiot, but he’s not wrong, humans makes mistakes all the time
Yes, but allowing floodgates of AI work to come in it lowers the bar and allow more mistakes to flow in.
Just because humans make mistakes doesn’t mean we are ok with shipping mistakes
He’s not saying that, he is saying LLMs are a tool and can therefore be used. He’s not saying AI generated code gets a free pass into the kernel just because an LLM says it looks good.
We pretty much have to make some mistakes in order to learn.
Being overly afraid of making a mistake is a good way to paralize oneself.
Yes but AI doesn’t exactly learn like we do. It’s not about being afraid of mistakes. It’s about mistakes being made with no real accountability and a system that cannot critically think the same way a human can. AI can process massive amounts of information faster than us but it cannot apply knowledge the same way we can.
AI can’t be accountable for itself.
TL;DR
Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it.
In response to
I expect maintainers who want to act on sashiko reviews to triage and verify them first before bothering authors, yes.
I don’t generally take issue LLMs used as a tool, but I do have a huge problem with lazy slop slingers. I also don’t like that the frontier models are closed source and rent-seeking, especially when they were trained on copyleft code and by all rights should themselves be open source if they were respecting the licenses. I’d think Linus would have something to say about that.
IMO, a decent philosophy is that LLMs can be useful tools, but if anyone can tell you used a LLM, you failed. People should have a healthy fear of being ridiculed for outsourcing their critical thinking. Anyone using a ton of tokens shouldn’t be commended; instead, the quality of their work should be called into question because they’re likely using LLMs as a crutch instead of a tool. Commits by Claude probably mean the person didn’t review the diff to clean up the slop, and also probably didn’t understand the changes well enough to write a clear commit message themselves. I wouldn’t want to see any commits crediting Claude in the Linux kernel.
Edit:
I’d also think that if Linus likes LLMs and wants to see quality tech democratized, he’d be advocating for FOSS LLMs under the Linux foundation where all code and training data are open source. What we have now with Anthropic and “OpenAI” is the equivalent of cloud-hosted Windows, and open weight models like Qwen are more like Windows XP in that they can be run locally, but they’re still proprietary and can’t be inspected, modified, or built upon. We need the Linux of LLMs.
deleted by creator
e> Linux users: switch from Windows to Linux to avoid ai slop
Linus, an out of touch idiot: Actually vibe coded slop in the OS is good!
I don’t like LLMs for a number of reasons - the theft, the environmental impact, the deceptive marketing - and also because I am really pissed about the abysmal quality which my AI-pilled coworker produces.
But in difference to Windows and most commercial software, the Linux kernel project has high quality standards, and I have some faith they’ll keep them.
Also, the cited mail discussion is not about AI generating unreviewed code - which I still see as very questionable - but about checking code for bugs and deficiencies. Given the amount of CVEs found by LLMs, I see no way around that for important infrastrcture projects like the kernel.
Many people are not amused by the hype - me neither. But AI could unexpectedly help open source in a few ways:
- it will become clearer that quality matters. We alrrady see that with Windows - Linux is getting many new users.
- We might see a trend to source code rewriting and re-use instead of using libraries. Libraries are pervasively used now, and that has drawbacks like bloat and dependency hell.
- Copyright is rapidly losing importance. This is really bad for artists and book authors. But for most (not all) programmers not, since copyright in software was mostly a tool to protect profits of corporations. Software development at its core is a service.
- Issues of autonomy and control are becoming more important than ever. The kernel is a pragmatic project, but it being free to use and modify is absolutely central - that’s why it uses the GPL - and we are also seeing every day that Stallman was and is right more than anybody would desire.
deleted by creator
Tell me you didn’t read what he said without telling me you didn’t read what he said
deleted by creator
Doubling down on not reading what he said without saying you’re doubling down on not reading what he said is a bold move
deleted by creator
Not a single acknowledgement of the environmental or cognitive costs of LLM use. Disgusting.
You can run a local LLM capable of assisting in software development for less energy than running a AAA video game. I’m not denying the environmental impact of the current AI landscape, but I kind of disagree that it’s intrinsic to LLMs as a whole, I think it’s more a symptom of capitalism and its disregard for sustainability causing everything it touches to have a high environmental cost.
Also, nearly all modern computing has high environmental costs, certainly all cloud computing. I think instead of focusing on AI only, it would be more helpful to engage in a broader discussion on how computing can be made more energy efficient as a whole, and do proper cost benefit analysis of all things we use computers for, including but not limited to LLMs. We may well still conclude from that process that we need to stop using LLMs, in which case we should.
If you’re against LLM use on environmental grounds (which I’m not disagreeing with), I submit to you that we should take the idea even further and things like gaming, video streaming, high frequency trading, social media, and any other nonessential computing should be on the same chopping block for the same reason. Applications of computing that we also use at scale with high environmental impacts, but that have been normalised and practically seen as a right by many of the same people against any amount of AI use (not saying that’s you, speaking generally). Why should AI be the only thing we raise concerns about if we’re to raise concerns? Doing environmental protection piecemeal by independently targeting single things and not the entire system has been shown time and time again to not work at best and make it worse at worst.
The only thing local llms which regular PCs can run are capable of is generating basic boilerplate code. Which you only need to do if you create a new project every day and need to automate it. They are useless for working on established codebase.
Currently, LLMs impact on electricity usage and fresh water usage across the world is HUGE.
The painful part to me is the choice on where to put the stress. Which areas to highlight and talk about.
Yes some weak LLMs can use comparatively little electricity. Yes some other industries use electricity, generate CO2 and consume fresh water, too. But the existence of other problems, to me, does not mean that eco impact of LLMs should be swept under the rug.
While it may seem counterintuitive, cloud LLMs use less electricity than local LLMs. When serving a single user, an inference engine spends very little time and power doing calculations, and most of it reading in the model (weights) from memory
Cloud LLMs serve multiple users and therefore batch requests, so a model that has been copied once from memory can be used to generate hundreds of tokens
Fresh water usage has been debunked. It’s fake propaganda. Datacenter water usage is a rounding error if you compare it to the whole economy. This is why you’ll always hear absolute numbers from these people, and never percentages
So, if one was to check your post history, they would find complaints about the energy usage of video games? The water usage of golf courses?
Or would we find neither, and learn instead that anti-AI is the only bandwagon you’ve hopped on because it is the most popular?
I’ll skip discussions about my persona. But most certainly golf courses or video games have orders of magnitude lower impact on electricity and watee than worldwide LLMs (mostly in datacenters).
Doubtful. The task of running an LLM and playing a video game is the same, so it just comes down to which has more customers. A lot of people spend many hours playing video games. There might be more LLM users, but they use the LLM for a few seconds a day, as opposed to an hour or more.
That being said, they’re both non-issues compared to beef production, which involves torturing cows. This activism should be directed there, then you can move down the target list once factory farming is outlawed.
You would be mistaken.
Data Centers use about 0.06% of the USAs total Water per year.
Watering golf courses in total uses about 0.5 percent.
My personal hardware that hosts my AI uses zero water. It uses more electricity when running a game.
There are more video gamers than there are users of AI.
Your persona is indeed just jumping on the popular bandwagon.
You’ll hear no arguments from me. Let’s cut energy usage for a lot of things. Let’s read books more. Let’s spend more time outside. Let’s eat less red meat.
That isn’t some gotcha you think it is. A lot of people finding the alarm over AI have been working to reduce their carbon footprint
Everything in your first paragraph is contrary to your first sentence. Every single one of those is another argument, but using a turn of phrase and baiting was apparently more important that even making the first, best, and most relavent argument you brought-out.
The real clincher is that you then pretend the person you are replying to didn’t already make those arguments, well, and that you are disagreeing with them.
“This isn’t [the] gotcha you think it is”, really?
Bro whatever you’re smoking please share.
And his did you get that local LLM?
That’s well outside the scope of the Linux kernel policies.
The immediate external consequences of what tools are used for development?
That will change over time, as the expense of utilizing them wastefully or in general has already. This isn’t a news article or snap-shot of the state of things today - its a mandate for due diligence going-forward, and an elaboration on how its been addressed so-far.
Torvalds has slowly been slipping I feel. Just more cracks.
It seems like maintainers doing a thankless job in FOSS are some of the most affected by the AI mania
he has always been like this. god, he’s one of the most consistent public figures i can think of. nothing has changed except that now you can see.
There are other questions around AI (like what the economy of it will actually look like in the end), but “is it useful” is no longer one of those questions. Anybody who doubts that clearly hasn’t actually used it.
It’s implied to be included in ‘other questions’, and that whether AI is useful to development efforts is the only question being considered for whether to permit its use.
There are other questions
Not sure if that’s really _acknowledging_ the problem.
If he’s taking a clear position that the controversies over whether AI is a net negative for society may have merit but still are not going to be considered because that sort of thing is not the priority of the project, it seems reasonable to not lay them all out or get into why AI may or may not be overall bad. You can say that choice is wrong, but it isn’t an evasive statement.
I mentioned this in another thread, but if Linus starts acknowledging the ethical questions of where good code is coming from, things unravel fast. Microsoft is a major contributor. So are plenty of defense, oil, industrial prison complex, etc. Someone overtly evil is contributing code.
Further just the pure tech industry isn’t free from sin, Intel and AMD contribute lots to the Kernel and they certainly utilize slave or child labor at some point in chip production.
There are no good people within any part of the tech industry and yet AI is the line? Is it because that’s the worst evil in our industry or is it because this one is the most loud and inconvenient to us?
Linus is not a very progressive guy. Alas, it is what it is. Hence the political bans in the LKML, who has votes in The Linux Foundation, where such foundation is based, the embracing of (Rust based) permissive licensing of drivers, and now this AI bullshit. Linux is not GNU.
I’m pretty sure the money talked, and in the next few years we’re going to see walls being put around the kernel. We should rethink whether the ‘benevolent’ part of ‘benevolent dictator’ still applys.
IIRC Linus is self-declared as “woke”. The exact quote is:
I’m a card-carrying atheist, I think a woman’s right to choose is very important, I think that “well regulated militia” means that guns should be carefully licensed and not just randomly given to any moron with a pulse, and I couldn’t care less if you decided to dress up in the “wrong” clothes or decided you’d rather live your life without feeling tied to whatever plumbing you were born with. And dammit, if that all makes me “woke”, then I think anybody who uses that word as a pejorative is a f*cking disgrace to the human race.
Linus is Liberal as a pejorative.
He’s a political/economic pragmatist. If you have record of him promoting capitalism or bashing socialism, please share, but as it is, Norway is one of the dozen most-socialist-leaning coutries on earth.
Maybe don’t apply USian non-sense everywhere. Liberal vs Mamdani as a prejorative is relavent. Liberal vs Torvalds as a prejoritive makes about as much sense as saying the same about Xi Jiping, Kim Jong Un, Miguel Díaz-Canel, or Lenin.
It’s absurd, obtuse, and besides the point, every bit so as the DNC using identity politics only to silence, promote, or ignore real voices for change.
Ah, the liberal progressiveness. Not the seizing the means of production for the proletariat kind. My bad, I wasn’t being fully open in my past comment because I know the world ‘socialism’ triggers many in the technology bubble.
I don’t think its the LLM usage that is doing this I think using it is exposing what was already there since some people who use it are genuinely improved by using them, and not just the dummies, there are people who know their stuff, using them to do more stuff, and not showing obvious cognitive impairments.
I think its just that many people got to where they can use ai in these ways by single mindedly focusing on one thing which meant being bad at the meta skills involved in keeping brains healthy across different contexts.
