Most projects don't have "lieutenants" that triage incoming patch sets. I totally get why projects smaller than the Linux kernel ban AI code out of self-defense against the tidal wave of crap.
To be honest, I think banning "AI-pioneered" (i.e. code fully written by AI without a single human thought involved) code is still a valid idea since 99% of the time it's solving a non-existent or very edge case-ish issue by adding extra bloat code to the codebase, but what I'm against is banning experienced engineers with good understanding over code from using AI just like any other tool you have to be more productive
I’m sympathetic to his argument that if it’s good code then it’s good code and if it isn’t actually useful or has flaws those are already reasons to reject it.
And since you can’t actually know for sure how much of a piece of code (or commit message) was written by AI you don’t really have a choice, do you? As time goes on and more people try to learn programming using AI a bunch of them may end up picking up some of its stylistic stuff. And if the submitter uses proper kernel formatting standards, will you be able to tell?
Hopefully, people don’t look back from a few years on and state that this was the last truly stable or secure versions of the kernel. Yeah, I assume Linus is assuming people understand their submissions but people are lazy. If they were given code in some fashion, it wasn’t thought through.
A bug is a bug is a bug. But not everything reported is a bug.
I watched a significant portion of that video, it represents what I expect well. You can't bury your head in the sand because an AI reported a bug, it's, potentially, still an exploitable bug. They can however raise the standard. They can require that reporting a bug comes with a fix, even if they reject the fix provided, it gives them an opportunity to evaluate that the bug is real, that someone put some effort in resolving it for the community (even if an LLM provided the fix); it gives something to evaluate the can benefit rather than only being a determent to the community.
I've sent a half dozen patches over the year to projects, I've always provided my perceived fix for the issue, occasionally they've been implemented as-is, and sometimes they were replaced outright with an alternate solution. I never really cared about the solution, just that the issues was resolved.
The first was a commercial library that dealt with multiple modems (interrupts and daughter cards)on dos based machines. They took the fix and integrated it and provided me with an updated library. Not much of a thank you, but the library was fixed.
More recent ones was another commercial library, open source but proprietary, and they sat on it for 3 months waiting for me to fix it their way, before the accepted it as-is and released appropriately adjusted patch. (I'm not their employer or employee, their complaints were tedious undefined rules. I fixed the bug and had working code. I waited them out.)
Good responses, bad responses, none of them mattered, what did matter to me, and others, is the developers involved cared enough to eventually implement the right fix for their codebase. The bugs need to go. Set the ego aside and fix it.
In Europe, there is this stereotype about how southern countries produce more extrovert reactions, and northern countries create silent, calm people. Untill you piss them off, then you get exactly the kind of reaction Linus produces. I think the world got used to the USA corporate blandness, and gets surprised when people just don't fit in that culture.
I expect the participants in this discussion to grumble for a few days, then continue working amically together.
He has made his position clear, repeatedly, for years now. He’s heard the arguments. He’s been in debates over them.
But people will not stop bringing it up over and over again. They just had the kernel summit where it was discussed again.
I imagine he’s REALLY REALLY tired of people continuing to try to start it up again.
Given Linus’ stance that it’s just a tool, which makes sense as a position, it’s really odd to ban only that one tool especially when you know your ban is completely useless to “make a statement”.
I guess he prefers to spend energy writing a long response explaining things, instead of just ignoring the person OR pointing to prior message and leave it at that.
Just interesting to see how different people handle these situations.
Pinchart is not just some random person. He’s been in the kernel community for a long time working on stuff like V4L, I think.
He’s someone who should know the policy.
Linus agreed that the “assisted by“ tag that was invented for AI wasn’t working and it got removed. There was already discussion that lead to this.
So here he is, again (presumably), taking this patch series removing that tag to try to push the topic again. That kernel summit I mentioned was 6 days ago as I type. LWN reported this was talked about.
I’m sure Linus would ignore me (a nobody) posting. Or someone who has made a few patches over the years.
I think Pinchart is high enough in the rankings Linus felt this was needed, or maybe more likely couldn’t hold his frustration.
I can’t imagine how frustrated by and tired of this topic Linus must be.
Most projects don't have "lieutenants" that triage incoming patch sets. I totally get why projects smaller than the Linux kernel ban AI code out of self-defense against the tidal wave of crap.
To be honest, I think banning "AI-pioneered" (i.e. code fully written by AI without a single human thought involved) code is still a valid idea since 99% of the time it's solving a non-existent or very edge case-ish issue by adding extra bloat code to the codebase, but what I'm against is banning experienced engineers with good understanding over code from using AI just like any other tool you have to be more productive
I’m sympathetic to his argument that if it’s good code then it’s good code and if it isn’t actually useful or has flaws those are already reasons to reject it.
And since you can’t actually know for sure how much of a piece of code (or commit message) was written by AI you don’t really have a choice, do you? As time goes on and more people try to learn programming using AI a bunch of them may end up picking up some of its stylistic stuff. And if the submitter uses proper kernel formatting standards, will you be able to tell?
That quote is nowhere in his message.
We've replaced it now. (Submitted title was "Linus Torvalds: "banning AI is stupid and so are you"")
Thank you.
Correct, I was paraphrasing.
Hopefully, people don’t look back from a few years on and state that this was the last truly stable or secure versions of the kernel. Yeah, I assume Linus is assuming people understand their submissions but people are lazy. If they were given code in some fashion, it wasn’t thought through.
I'm substantially less worried about the Linux kernel than almost everything else.
Honestly, seeing people go from concern trolling to what I like to call _concern farming_ has been utterly disappointing.
_concern farming_ is when people build entire crops of concerns with no intent to do anything other than fertilize them.
This video from the linux CVE team significantly increased my confidence they will not only be ok, but thrive
https://www.youtube.com/watch?v=NnV_cWeoo5Q
I really like your concern farming, will be using it going forward.
A bug is a bug is a bug. But not everything reported is a bug.
I watched a significant portion of that video, it represents what I expect well. You can't bury your head in the sand because an AI reported a bug, it's, potentially, still an exploitable bug. They can however raise the standard. They can require that reporting a bug comes with a fix, even if they reject the fix provided, it gives them an opportunity to evaluate that the bug is real, that someone put some effort in resolving it for the community (even if an LLM provided the fix); it gives something to evaluate the can benefit rather than only being a determent to the community.
I've sent a half dozen patches over the year to projects, I've always provided my perceived fix for the issue, occasionally they've been implemented as-is, and sometimes they were replaced outright with an alternate solution. I never really cared about the solution, just that the issues was resolved.
The first was a commercial library that dealt with multiple modems (interrupts and daughter cards)on dos based machines. They took the fix and integrated it and provided me with an updated library. Not much of a thank you, but the library was fixed.
More recent ones was another commercial library, open source but proprietary, and they sat on it for 3 months waiting for me to fix it their way, before the accepted it as-is and released appropriately adjusted patch. (I'm not their employer or employee, their complaints were tedious undefined rules. I fixed the bug and had working code. I waited them out.)
Good responses, bad responses, none of them mattered, what did matter to me, and others, is the developers involved cared enough to eventually implement the right fix for their codebase. The bugs need to go. Set the ego aside and fix it.
I recently worked through a security regression with the OpenCode team
https://github.com/anomalyco/opencode/pull/51337
The Pi dev who's handling my request there is just not getting it.
https://github.com/earendil-works/pi/issues/10895
He is pretty combative and abrasive.
In Europe, there is this stereotype about how southern countries produce more extrovert reactions, and northern countries create silent, calm people. Untill you piss them off, then you get exactly the kind of reaction Linus produces. I think the world got used to the USA corporate blandness, and gets surprised when people just don't fit in that culture.
I expect the participants in this discussion to grumble for a few days, then continue working amically together.
Yep, he always has been. But I found he was often right and I did have some respect for him.
I think he's very wrong on this, though.
He has made his position clear, repeatedly, for years now. He’s heard the arguments. He’s been in debates over them.
But people will not stop bringing it up over and over again. They just had the kernel summit where it was discussed again.
I imagine he’s REALLY REALLY tired of people continuing to try to start it up again.
Given Linus’ stance that it’s just a tool, which makes sense as a position, it’s really odd to ban only that one tool especially when you know your ban is completely useless to “make a statement”.
I guess he prefers to spend energy writing a long response explaining things, instead of just ignoring the person OR pointing to prior message and leave it at that.
Just interesting to see how different people handle these situations.
Pinchart is not just some random person. He’s been in the kernel community for a long time working on stuff like V4L, I think.
He’s someone who should know the policy.
Linus agreed that the “assisted by“ tag that was invented for AI wasn’t working and it got removed. There was already discussion that lead to this.
So here he is, again (presumably), taking this patch series removing that tag to try to push the topic again. That kernel summit I mentioned was 6 days ago as I type. LWN reported this was talked about.
I’m sure Linus would ignore me (a nobody) posting. Or someone who has made a few patches over the years.
I think Pinchart is high enough in the rankings Linus felt this was needed, or maybe more likely couldn’t hold his frustration.
I can’t imagine how frustrated by and tired of this topic Linus must be.
Consider over 30 years of protecting the Linux project from idiots and vultures.
The combativeness is a feature, not a bug.
Linux is funded by BigCorp and has fallen in line with BigCorps views no their AI investments.
For reals though: Comparing AI to editors or compilers is a massive false equivalency.