ai in the linux kernel, from the perspective of a non-kernel dev

Earlier today, I was reading lobste.rs, as part of my daily routine to keep myself informed, when I saw "Linus Torvalds on LLM usage in kernel development"— A link to a response Linus sent in the Linux Kernel Mailing List (LKML) in regards to "Sashiko".

A quick search led me to the sashiko-dev/sashiko GitHub repository, which seems to be for an AI "Agent" that reviews Linux Kernel patches by going through threads in the mailing list (which is available on the web via lore.kernel.org), GitHub, and Gitlab PRs.

I mean, in theory it sounds like a good idea. I do not follow the mailing list, so I don't have a great view of what responses the agent might reply with. Either way, this isn't about the system itself, nor the idea of "letting a large-language model review your patches and tell you that they might have some bugs"—I'll probably talk about that later on, if I remember—Instead I wanted to kind of dump my thoughts on Linus' stance on LLM usage during kernel development, just for fun (but also because I want to see if other people have similar thoughts as me, or if they have any criticism to give me, because I am generally neutral on these things).

You can read the thread in your own time if you so like, i've left the link just below this paragraph for your leisure.

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.

I generally agree with this statement. LLMs particularly have gotten extremely good, extremely fast lately, at an almost scary pace. It's definitely useful.

However, it's not really that it isn't, rather, what has got people worried is the environment that it fomented.

A lot of the marketing surrounding AI in general has to do with "Hey, you don't have to do this anymore, you can do something else instead! (like giving us money)", and in my opinion, it's a very dishonest way to look at this technology.

People also really, really care about their intellectual property—so, books, art, music, videos, articles, websites, etc. And when they heard that AI is trained on top of the same things that they worked so hard to make, of course they got mad.

"They didn't ask me for my consent!"

Of course, I am very biased on the topic of copyright, because I would rather put everything I make into the public domain (or at least, under a extremely permissive license) than tell other people "hey, don't you dare use the code of this tool I made!!!", because I believe that other people might find it useful and would want to use it as a base for their own projects.

Another sentiment that I've seen gone around is the environmental impact of the data centres that host these models, which, let's be honest, isn't negligible. I am not blind.

However, I would argue that there are other things that we use in our day-to-day lives that have a greater, wider impact than a data centre.

We were already speed-running the demise of our planet, long before the AI "boom", so I don't really understand why people care more about data centres (which mind you, have already existed for other purposes) existing to host AI.

Anyway, I think that's more than enough for a single paragraph, let's move on.

We're not forcing anybody to use it, but I will very loudly ignore people who try to argue against other people from using it.

And no, AI isn't perfect. But Christ, anybody who points to the problems at AI had better be looking in the mirror and pointing at themselves at the same time.

Because it's not like natural intelligence is always all that great either.

Most of what I said on the first section also applies here.

But yeah, both are imperfect. I don't actually know if anyone has ever disputed that though.

Sure, the social angle of working on open source is important and often a very motivating part of the project, but in the end that's a side benefit, not the point of the project.

Yes, it isn't the point of the project. But it's still critical to the project, so I wouldn't really call it a "side benefit".

Calling it that is similar to calling it a happy accident, even though it was eventually bound to happen as more and more people began contributing to the project.

In the kernel community we do open source because it results in better technology, not because of religious reasons.

Though, it's not for religious purposes. For the most part.

I hope.

In conclusion, this is just Linus' opinion. It's not like we need to burn him alive just for expressing his stance on LLMs, or for even hinting that they might be useful.

The beauty of open source is that you can simply fork a project if you don't agree with the direction that it is taking, however, maintaining it is a whole other task of its own, so people resort to publicly shame others for not having the same "correct" opinions as them.

I should probably revisit my "On AI, Humanity and Art" article, because it was written when I was angry at posts on BlueSky, which hilariously is the same thing that I am criticising others for here.

In hindsight, I didn't expect this to be as long as it is, but it's good practice for my vocabulary and grammar.

Whatever, someone will probably make a post telling how wrong I am about this. Ah, how fun the internet is.

Well, I think that's about it for now. I'll probably talk about something else soon, though I am not sure what, and it will probably be in my main blog anyway.

Goodbye!