Kernel Summit: "The time has come to make it harder for new people to contribute code to the kernel." -- Linus Torvalds reacting to AI assisted code contribution flood

3 days ago by HaraldvonBlauzahn to c/linux

cross-posted from: https://lemmy.world/post/53100937

"Torvalds said that the time has come to make it harder for new people to contribute code to the kernel. People like Kroah-Hartman have spent years trying to teach people how to contribute; it is time to stop doing that, he said."

https://lwn.net/Articles/1099594/

It is difficult to summarize exactly how Linus got to this point. Most of the discussion at this years Summit appears to have been related to AI assisted code and how the maintainers are dealing with the increase in code submissions to LKML. Many of the maintainers expressed varying levels of frustration with AI assisted drive-by code submissions. Linus' train of thought is that they make it intentionally more difficult for humans to contribute. A later suggestion of his was to create an AGENTS. md that makes generating kernel code expensive in compute and token time.

HaraldvonBlauzahn 30 points 3 days ago

Copied from another thread:

When he came out talking about allowing LLM code contributions everyone told him this would happen.

He has criticized the excessive hype around GenAI/LLMs long before. Torvalds has generally a very pragmatic approach, not a black-and-white one.

save

path: 0 26342005, hotness: undefined, score: 30, children: 4

HaraldvonBlauzahn 25 points 3 days ago

Three things more:

The Linux kernel is Torvalds project.

The fact that experienced kernel developers can sometimes use LLMs to their advantage does not mean that the average Joe Codemonkey can do it. These algorithms are often described as a kind of competence multiplicator - but experience from education shows that they are more risky than that and can hinder learning.

And third, we should in this question not put too much weight in the opinion of influential people. Behind the GenAI tech are huge corporations which pour billions of money into it. It is likely that they can pressure and bribe people and that they also pay influencers and bloggers to tout their stuff. Making decisions dependent on single influential persons, however competent and acclaimed they are, is therefore dangerous (see Eric Raymond, Eva Kaili or Gerhard Schröder). Torvalds is not Jesus, and I am pretty sure he didn’t apply for that role.

save

path: 0 26342005 26342015, hotness: undefined, score: 25, children: 0

atzanteol 8 points 3 days ago

This was going to happen regardless. Not allowing veteran devs to use AI tools wouldn't stop this.

save

path: 0 26342005 26344090, hotness: undefined, score: 8, children: 1

HaraldvonBlauzahn 2 points 3 days ago

true.

save

path: 0 26342005 26344090 26344251, hotness: undefined, score: 2, children: 0

strop 5 points 3 days ago

vErY pRAgmTiC

save

path: 0 26342005 26343924, hotness: undefined, score: 5, children: 0

XiJinpingStanAccount 16 points 3 days ago

I wonder how effective a master/apprentice situation could be for this? Like someone gets taken under the wing of a veteran contributor and all their work is vetted and signed off on my that person before being submitted. They get taught how to do things and in return maybe do some of the busywork for the veteran. Because your always going to need new contributors but you need a way to vet them.

save

path: 0 26343234, hotness: undefined, score: 16, children: 2

Dran_Arcana 7 points 3 days ago

The problem here is people will beg for the opportunity and plead that they'll take it seriously and it'll be life changing and then ghost you at the first sign of actual work.

Finding a decent apprentice that won't waste your time is a chore in and of itself

save

path: 0 26343234 26345376, hotness: undefined, score: 7, children: 1

XiJinpingStanAccount 8 points 3 days ago

Yeah I wouldn't leave it up to like individuals. Guilds back in the day had good systems for it. You'd get someone who had been vetted and verified to be competent assigned to you. Usually the apprentices had to "buy in" that doesn't mean money persay. But labor could be the buy in. If you want to be given some of a skilled mentors time then you'd have to put in labor to prove you were serious about it and earn it.

save

path: 0 26343234 26345376 26346846, hotness: undefined, score: 8, children: 0

Duchess 8 points 3 days ago

Can’t wait to see the bananas AI kernel forks coming. Those guys aren’t going to just stop contributing.

save

path: 0 26341882, hotness: undefined, score: 8, children: 6

Shayeta 11 points 3 days ago

As is their right, and the right of everyone. Linux is free and open source. The whole beauty of it is that anyone can take it and do what they want with it.

save

path: 0 26341882 26343498, hotness: undefined, score: 11, children: 0

floofloof 9 points 3 days ago

Doesn't mean anyone has to use them. Or anyone serious at least.

save

path: 0 26341882 26341998, hotness: undefined, score: 9, children: 4

NateNate60 11 points 3 days ago

One of the main benefits of Linux is that it has very proven security. The Linux kernel is possibly the most looked-at and audited code project in history. So any security issues are quickly spotted and patched. Any forks of Linux wouldn't inherit the swarm of security auditors attached to it. In addition to that, LLMs generate a voluminous amount of code, which is mostly impossible to audit for security anyway.

There is great promise in AI-driven security testing, but in terms of pure volume of eyes and brains working on it, you cannot really compete against Linux.

Organisations know that when they use the Linux kernel, they are using a project that has an excellent reputation for reliability and security.

save

path: 0 26341882 26341998 26342089, hotness: undefined, score: 11, children: 3

Duchess 5 points 3 days ago

I don’t think any of that is going to stop them from making bananas Linux kernel forks. They aren’t going to just stop generating slop. Legality, practicality and usefulness haven’t stemmed the tide of non kernel madness one bit. If they can’t submit it to the official kernel we are going to get forks where they can.

I honestly am interested in seeing what comes out of it. I’m not going to install it on MY hardware but I am curious what happens when caution is thrown to the wind and it’s pedal to the metal.

save

path: 0 26341882 26341998 26342089 26342232, hotness: undefined, score: 5, children: 1

Quazatron 4 points 3 days ago

I expect that some forks will actually have useful features or corrections that, when cherry picked, cleaned and polished, can improve the main kernel.

The rest (98% or more) (yes, I pulled that number from my arse) will be absolute rubbish.

I do not envy the kernel maintainer's job at this point. They are sifting sand with a spade and a sieve while the world is dumping truckloads of rubble at them.

save

path: 0 26341882 26341998 26342089 26342232 26343185, hotness: undefined, score: 4, children: 0

atzanteol 1 point 3 days ago

LLMs generate a voluminous amount of code, which is mostly impossible to audit for security anyway.

They can but they didn't have to. If you tell claude to "minimize changes" or "keep it simple" then it does.

This is why people need to familiarize themselves with AI tools. Even those of you who feel they are "of the devil". Because how one uses these tools matters.

save

path: 0 26341882 26341998 26342089 26345317, hotness: undefined, score: 1, children: 0

dusk 5 points 3 days ago

Recent photo of linus

save

path: 0 26342043, hotness: undefined, score: 5, children: 0

Rules

Related Communities

Rules

Related Communities