Linus Torvalds To Critics of AI Coding On Linux: 'Fork It. Or Just Walk Away.' (arstechnica.com) 58
Linus Torvalds says the Linux kernel will not ban AI-assisted coding tools, and if anti-AI absolutists have a problem with that, they can "fork it" or "walk away." An anonymous reader quotes a report from Ars Technica: Writing in a lengthy post on the Linux kernel mailing list this week, Torvalds said that "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. Or just walk away." The statement came amid a lengthy thread arguing about the use of Sashiko, an "agentic Linux kernel code review system" that its creators claim can, in tests, independently find 53.6 percent of the bugs that would end up being fixed by human coders in later commits. But the tool can also waste maintainers' time by sending "false positive" reports of bugs that don't exist, at a rate Sashiko's maintainers estimate is "well within [the] 20% range."
In discussing whether maintainers should be subjected to a flood of these kinds of automated, AI-powered bug report emails (true or false), one poster cited the Software Freedom Conservancy's recent statement that the open source community "should support, not just tolerate, those who outright reject LLM-gen-AI systems" and that "every FOSS contributor deserves self-determination regarding LLM-gen-AI." In the face of that statement, Torvalds said that he rejects those who demand that their open source projects not accept any LLM-generated code or revisions. "We're not forcing anybody to use [LLM tools], but I will very loudly ignore people who try to argue against other people from using it," Torvalds said.
Torvalds said his position on this is a pragmatic one that's "based on technical merit. Not fear of new tools." And when it comes to utility, Torvalds said that "AI is a tool, just like other tools we use. And it's clearly a useful one. It may not have been that 'clearly' even just a year ago, but it's no longer in question today. Anybody who doubts that clearly hasn't actually used it." [...] While Torvalds acknowledged that "AI isn't perfect," he urged detractors to compare the output of these tools to the performance of human code maintainers. "Anybody who points to the problems at AI had better be looking in the mirror and pointing at themselves at the same time," Torvalds wrote. "Because it's not like natural intelligence is always all that great either."
In discussing whether maintainers should be subjected to a flood of these kinds of automated, AI-powered bug report emails (true or false), one poster cited the Software Freedom Conservancy's recent statement that the open source community "should support, not just tolerate, those who outright reject LLM-gen-AI systems" and that "every FOSS contributor deserves self-determination regarding LLM-gen-AI." In the face of that statement, Torvalds said that he rejects those who demand that their open source projects not accept any LLM-generated code or revisions. "We're not forcing anybody to use [LLM tools], but I will very loudly ignore people who try to argue against other people from using it," Torvalds said.
Torvalds said his position on this is a pragmatic one that's "based on technical merit. Not fear of new tools." And when it comes to utility, Torvalds said that "AI is a tool, just like other tools we use. And it's clearly a useful one. It may not have been that 'clearly' even just a year ago, but it's no longer in question today. Anybody who doubts that clearly hasn't actually used it." [...] While Torvalds acknowledged that "AI isn't perfect," he urged detractors to compare the output of these tools to the performance of human code maintainers. "Anybody who points to the problems at AI had better be looking in the mirror and pointing at themselves at the same time," Torvalds wrote. "Because it's not like natural intelligence is always all that great either."
Pragmatic attitude works well on this. (Score:4, Insightful)
He's typically pragmatic about it, and the attitude serves well.
Is it a real bug? Are POSIX standards violated, or user code broken by a operating system's promise? If yes, it needs to be fixed. If no, it has already wasted too much time.
Re: (Score:2)
He's typically pragmatic about it, and the attitude serves well.
Here's a few posts from Dries Buytaert [wikipedia.org], lead of the Drupal project, (UserID 1)
Never submit code you don't understand [dri.es]
AI rewards strict APIs [dri.es]
Re:Pragmatic attitude works well on this. (Score:4, Interesting)
I wouldn't say pragmatic, no. He's tried it, found it works, and now won't listen to the people pointing out the numerous problems with it.
The slopped Linux kernel may well be a violation of the copyrights of several unknowns in Europe, for example. If you're not planning to build a Linux-based product in Europe, not a big issue I guess. Not an issue in the US. But it's amazing how many people think US copyright law is the only type of copyright law.
And that's before we get to changing the entire nature of Linux so it's no longer a project understood by human beings.
This is Bitkeeper all over again. That was the last time Torvalds made a "pragmatic" decision, effectively locking out a large number of kernel devs from kernel development, until he was forced to build Git to replace it.
Re:Pragmatic attitude works well on this. (Score:5, Insightful)
He has pointed out AI's flaws and limitations. He has also said the beneft is it gets him to the starting point more quickly to either figure out a bug or how to do something, or even how to rewrite something he's already done. He is not blindly accepting what it says.
Trust, but verify would be closer to his thinking.
Re: Pragmatic attitude works well on this. (Score:2)
I'm of the stance that AI could be good to check my code and test it, then let me decide what to do.
I need to understand what I have in my code to trust it to work and be secure so an AI generated snippet might be working well but be hard to understand. Such code can be very hard to maintain.
Re:Pragmatic attitude works well on this. (Score:4, Informative)
Ted Tso, a senior linux contributor and maintainer, appears to agree; he noted [kernel.org] on the thread that there are multiple use cases for LLMs and conflating all of them is not helpful.
Re: Pragmatic attitude works well on this. (Score:1, Troll)
It's a pragmatic FINANCIAL decision. Five of the top funders of the Linux foundation are hawking LLMs. Linus is typically compromised by the usual capitalistic conflict of interest.
Re: (Score:3)
Re: (Score:2)
In drinkypoo's defence, I can make what he said more compelling by changing just one one word: "It's a pragmatic EXISTENTIAL decision".
If those top funders were to withdraw support - as well they might given the influence that Linus and Linux bring to any such discussion - it might spell the end of Linux as we know it. I can't say that such a consideration actually crossed Torvalds' mind, but if it did I wouldn't be at all surprised.
Re: (Score:2)
If anyone is blindly accepting AI code, they deserve what they get.
I've found that AI is a great copywriter. It can write copy. I turn into an editor, accepting or rejecting things.
This is faster, and usually ends up with equal or greater quality, and absolutely meets functional requirements, testing requirements, and the stated "definition of done."
Re: (Score:2)
Let's say a coding LLM is trained on copyrighted works that weren't licensed in such a way to permit being used as training data. Let's then say Linux and/or any other numerous projects start accepting submissions that were in whole or part generated by said LLM.
How would you determine which submissions violated a copyright?
What if I, Joe Q. Human look at the source code for something, learned from it, and then started making submissions for the Linux kernel based on things I learned. Am I in violation of copyright?
Most interesting - I wish I had mod points to give.
Results over tools (Score:3)
And contributors being responsible for what they submit no matter if it is AI or not.
Re: (Score:2)
And contributors being responsible for what they submit no matter if it is AI or not.
Time will tell if contributors using LLM based coding tools will actually feel responsible for what they submit to the extent that they actually check and understand _all_ the code they contribute. So far, experiments have shown that one cannot expect people to remain as diligent after extended periods of LLM use. And then, there may be real trouble ahead, like LLM based tools injecting hard to spot back-doors into the code, not being noticed by overly confident LLM-tool users.
I agree with Linus' insofar
Re: (Score:2)
This is absolutely a problem if you work in a shitty sweatshop that forces you to be a sin eater for robot code.
The way we do it is, the tools are available, but you're not required to use them in any specific way or really, at all if you don't want to. You do, of course, have to get your work done and be responsible for your code. We have a spectrum of use, from all-in cascades of agents
Re: (Score:2)
You can expect Linus to become loud, if someone submits crap.
Re: (Score:2)
Peer review is still a thing.
If your peer review process is just a rubber stamp, then you're already fucking it up whether AI writes the code or not.
Re: (Score:2)
Peer review is still a thing.
If your peer review process is just a rubber stamp, then you're already fucking it up whether AI writes the code or not.
Yes, but when all peer-reviewers are accustomed to use AI tools, also for their reviews, you can easily end up in a situation where even multiple approvals signed off by humans (with all good intentions) mean nothing more than that one popular LLM, which happened to be used by all of them via their coding/review tools, liked the code it also authored before.
I would be less pessimistic about this possibility if I had not witnessed exactly this in a company I worked for: AI tools reviewing PRs written by AI
Not surprised (Score:2, Insightful)
Re: Not surprised (Score:2)
Of course he does. Corporations have made him wealthy. Wealthy people can afford not to have human-centric principles.
Re: (Score:3)
Re: (Score:2)
You're welcome to try a different method.
Re: (Score:2)
Re: (Score:1)
As I mentioned, Linux has never been a social warrior project.
If you don't have technical reasons, you don't have reasons.
That is unfortunately incorrect. Commits have been accepted into the Linux kernel that, for non-technical reasons, change or remove ungood words, such as blacklist.
Re: (Score:2)
Re: (Score:2)
The article mentions the first mail:
https://lore.kernel.org/linux-... [kernel.org]
Then Lauren Pinchart insisted that AI is bad and Linus carefully explained why he thinks they are useful:
https://lore.kernel.org/linux-... [kernel.org]
When Pinchart doubled down, Linus Torvalds wrote a second clear statement:
> So this is where the discussion ends.
>
> As I mentioned, Linux has never been a social warrior project.
>
> If you don't have technical reasons, you don't have reasons.
>
> You can choose not to use AI, but that's your PERSONAL choice.
>
> It has absolutely no impact on anybody else, and you should not expect
> it to have any.
and got quite clear he's not on the side of the FSF:
> It's why Linux is *not* GNU/Linux, and why we call things "open
> source" instead of "Free Software".
>
> So keep your ethics where they belong - in your personal life. Don't
> try to enforce your ethics on others.
https://lore.kernel.org/linux-... [kernel.org]
Ouch. Blunt and pointed in one exchange. LoL
Re:Second nail (Score:1)
That is how open source works! (Score:3)
When a groups of contributors doesn't like how the project is going, they can always fork it. That's how we got LibreOffice.
Linus is right, as usual (Score:5, Interesting)
He understands that ideologues only make technical endeavors worse, and he's is correctly inviting people to prove him wrong.
Re: (Score:1)
if you want to slam your dick in a door, he wont stop you
Re: (Score:3)
Here's what he had to say when asked when this holy war was perhaps at its peak:
I don't actually have any particularly strong opinions on systemd itself. I've had issues with some of the core developers that I think are much too cavalier about bugs and compatibility, and I think some of the design details are insane (I dislike the binary logs, for example), but those are details, not big issues.
(source [zdnet.com])
Re: (Score:2)
Until they release systemd-kernel. You can say I'm being ridiculous but they already have their own bootloader (systemd-boot). At this point, the kernel is just about the only thing left for systemd to consume.
Re: (Score:2)
Yes.
Re: (Score:1)
He understands that ideologues only make technical endeavors worse, and he's is correctly inviting people to prove him wrong.
Well, he was proven wrong regarding his "pragmatic" use of BitKeeper, but at least he fixed that mistake by writing something better when he, too, realized how BitKeeper was not suitable for an open source project. Also, his excellent track-record did not keep him from becoming involved in the bound-to-fail "Transmeta" activities. So yes, he has excellent credentials, but he cannot predict the future, like anyone else.
Re: (Score:2)
As for Bitkeeper, what alternative was there at the time that didn't have even worse issues? I don't see RMS suggesting any alternatives in his 2005 "thank you" letter to Larry McEvoy. In the event, Linus wrote a replacement that ended up taking over the world a second time, which makes him even harder to argue with.
Re: (Score:2)
Transmeta wasn't "bound to fail", that's ridiculous.
Transmeta was bound to fail, and several experts in the field of CPU design pointed this out early on. For example: https://books.google.com/books... [google.com] and https://news.ycombinator.com/i... [ycombinator.com]
Re: (Score:3)
But wait! Just because I'm trashing the hype doesn't mean I'm trashing the chips. Transmeta has employed some innovative technology in Crusoe, particularly for conserving power. Crusoe chips are perfect for notebook PCs because low power consumption means longer battery life and cooler operation.
As for the second link, yeah not putting much weight on a couple randos who claimed to be working at a competitor at the time.
Re: Linus is right, as usual (Score:3)
Both off the sources you quote claim that Transmeta had good chances too succeed in the low power market, not that they were donned to fail.
Re: (Score:3)
Why was he proven wrong? They used it as long as they could use it in a pragmatic way and when this changed, he pragmatically hacked together and alternative. Things worked out and the transition was rather smooth. Building git just because bitkeeper is non-free would have been a good choice, but not a pragmatic one. Building git when bitkeeper was no longer an option was the pragmatic choice. Maybe it could have been less smooth, but I think history proves that his judgement how fast they could adapt when
Re: (Score:2)
Re: (Score:2)
> He understands that ideologues only make technical endeavors worse, and he's is correctly inviting people to prove him wrong.
Actually, from what I've seen, we *SHOULD* make technical endeavors worse. If they were significantly worse we might not have invented AI in the first place, and that would be a better world.
The problem with AI is over reliance .... (Score:3)
Re: (Score:2)
Luckily, rules and standards are another thing AI is good at if you give it to them.
Re: pet project (Score:1)
Re: pet project (Score:2)
Re: (Score:1)
Start pulling more things into SystemD so that we won't need Linux and its benevolent dictator anymore.
Just walk away (Score:2)
Just rewrite it in Rust already (Score:1)
all those commercial sponsors should be willing to pay for it, right?
Re: (Score:2)
*monkey's paw curling*
https://blog.rust-lang.org/202... [rust-lang.org]
One day Linus will be the las coder (Score:2)
While the rest of us with our AI tools long forgot how that actually worked,
God bless the Swedes (Score:3)
And the Swedish Finns. And the Nordics in general. Their principled pragmatism is refreshing.
[Beep, beep, beep] (Score:3)
"Hey buddy. We've got a load of pull requests. Where do you want them dumped?"
Hey, I like unplugged music.... (Score:1)
Is the tool the problem or its it the way our civilization forces artistic expression underground in favor of the easy to clone and mass produce? LLM's are not causing anyone create the next linux or windows while the next easiest way to exploit the likes is always a year away.