Linus Torvalds: AI Won’t Replace Understanding Code

[00:00:00]Let's get into Linus Torvald's thoughts on AI in the Linux kernel, which has become an increasingly contentious topic as AI assisted tools are uncovering vulnerabilities, generating lots of bug reports, and now even being used to generate patches for the Linux kernel.

[00:00:16]Lionus talks about how this increasing volume of work is creating burnout for maintainers, especially the reports that are lowquality in a driveby style are clearly generated without any human follow-through. Now, Lionus isn't completely against AI. He sees it as a useful tool. But let's get his thoughts on whether or not AI actually belongs in open source in the Linux kernel, or do these tools just overwhelm maintainers and flood projects? Anyways, here's a recent conversation between Lionus Torvald and Durk, who leads Verizon's open source program office. Let's get into what Lionus really thinks about AI in the Linux kernel, its benefits, and the problems it's creating, and if open source should adopt it. So, we we talked backstage about the impact all of this has on maintainers because the maintainers were already struggling with burnout before and now this flood of bug reports isn't really helping. But are there any good tools that help you as a maintainer with code review with with understanding the patches that come in to help you with your workflow? We have a I mean we have a ton of tools for that and uh I I'd like to point out that Linux kernel is actually doing really well. We have every single release we have over a thousand people involved and uh we have a solid cer of maintainers that are there and who get most of the time they get well paid for being there too. And so we're doing well. uh all these problems that I've been talking about. I've been talking about the Linux kernel because that's what I do obviously, but think of all the tens of thousands of random projects that people maintain that are not the Linux kernel and that maybe somebody's like pet project that they've been working on for a decade or more and they have one person or three people that are somewhat involved and they get really burnt out when they get a flood of these sometimes obviously AI reports that just it's a bug report and when you ask for more information the person has done a driveby and and doesn't even answer your questions anymore. So, so that's the the real burnout issue is is I think there uh and now I forgot what the actual question was actually that aside >> I asked you if there are AI tools that are helpful.

[00:02:57]>> Oh yeah, we we do use AI tools and and uh I'd like to point out again I actually work mostly with people. So for me as a top level maintainer I don't do a lot of coding. I do coding on my toy projects, on my guitar pedal, on my like I do coding for fun because I'm a programmer, but my job is working with people and I do not use AI to work with people.

[00:03:24]>> Thank you.

[00:03:24]>> And I I should suggest you don't do that either. But we do have a lot of automated tools including AI. We have like Google for example set up this patch tool thing that a lot of people are looking at. I forget the name Shashiko uh that looks at all the patches that go to the mailing list and does a review of them and sometimes the review is not great.

[00:04:00]uh but quite often it finds issues and it asks questions and says hey what about this issue uh and uh that's the public one and there's a lot of companies working on private tooling for inside inside companies but actually these days a lot of the main developers uh do local AI which I actually would suggest uh everybody to look into. You don't want to be entirely at the mercy of of the big companies that at some point decide, oh, we need to make money, too.

[00:04:38]>> Uh, so you're showing your age because your your comment about tens of thousands of projects is really funny.

[00:04:46]This week, GitHub announced that they currently have 480 million repositories.

[00:04:51]And um in a recent study, 600,000 projects were considered of uh important or critical use for enterprise users of open-source software. So it's literally hundreds of thousands and hundreds of millions overall. The the the space is so big it's it's hard to wrap your mind around. But let's you you say this is a social issue. You talk about working with people. If you are at the beginning of your career right now. Yep.

[00:05:21]>> And a lot of people do these doom and gloom. All code will be written by AI.

[00:05:27]And on the flip side, those of you who were here on Monday for Jim's keynote saw Jim's very optimistic note that said actually we are creating a lot of jobs in the tech industry and we are seeing hiring in the tech industry. So do you have something that you would tell someone at the beginning of their career here is where you should focus? So my opinion has always been that hey AI is a great tool but it's a tool and and when I see people saying hey 99% of our code is written by AI I literally get angry because those same people I pretty much guarantee that 100% of their code is written by compilers but they never say that right it's it's exactly the same issue. Uh I grew up writing machine code and when I say machine code I don't mean assembly language. I mean the numbers and I mean it leaves an imprint on you. I I I can I still remember LDA is uh hex A9 on the 6502. Right.

[00:06:43]>> Correct. Uh and I it took me a while to understand that writing down the numbers and calculating offsets for branches is kind of stupid.

[00:06:55]And people had come up with this tool called an assembler.

[00:06:59]Uh and then later on I figured out compilers are good too. And these days I'm figuring out AI tools are good too.

[00:07:08]I'm still writing the code. I'm just not doing it the same way I did when I was basically typing in numbers in in data statements. Uh, and the I I'm personally 100% convinced that AI is changing programming, but it's not changing the fundamentals.

[00:07:32]Exactly the same way that you all use compilers to actually generate your code. You will all use well not maybe all of you but a lot of people will use AI to generate the code that the compilers use to generate the code that the assemblers then use to generate the machine code.

[00:07:54]This is revolutionary in the same sense that we've seen revolutions before.

[00:08:01]And AI will increase your productivity by a factor of 10. And I claim that compilers increase your productivity by a factor of the thousand. So AI is great, but AI is not changing programming. It may be changing other areas, don't get me wrong, but I'm a programmer, so I don't care.

[00:08:23]uh and and one of the things sorry my voice is already fading. One of the things that I think has always been true when you look at the code written by developers good developers write good code other developers write code that then breaks. And the same will be true for the people prompting the eye tool.

[00:08:43]People who know what they're doing, who understand systems will be able to prompt tools to write good code. And people who don't understand the complexity of systems will also prompt systems and write code and it will fail.

[00:08:56]>> And I think you do want to understand how it all works in the end. I still like I don't program in machine code anymore. But I still look at the generated code. So when I use a compiler, even when I use AI for my pet toy projects, I will use AI to generate code. I will look at that code. I will actually still look at the assembly language end result because it's what I grew up with. It's kind of where my comfort zone and right now I'm playing with microcontrollers anyway. So you really want to actually look at at the generated code and make sure it's it's doing what you really want it to do. So I think even when you use AI for coding, if you do a project that you actually maintain long term, you need to understand not just your prompts, but you need to understand the end result too because that's the only way you can maintain it long term. And it's easy enough and we see that all the time where you uh basically generate a new project and it's a throwaway project and it's a one one time and it's over. And uh AI is great for that and we call it vibe coding. But if you want to make something serious, you're going to have to maintain it for 35 years or something. and and at that point uh it's it's a lot more than just writing the prompts to make somebody else generate the code.

[00:10:32]>> Well, unfortunately, we're out of time.

[00:10:34]I have a lot more questions. I guess I'll have to ask them next year in Vancouver when we'll do the same session. I hope you'll all be there.

[00:10:42]Thank you very much.

[00:10:43]>> Thank you.

[00:10:46]>> Thanks to the Linux Foundation for hosting this event and conversation between Lionus and Durk. If you want more of this video or you found this conversation interesting, check out the link in the description below or in the comments section. Linus and Durk talked about the challenges and opportunities that you get with AI. AI is clearly helping developers move faster by reviewing patches, finding bugs, and generating code, but it also is putting more pressure and burnout on maintainers through lowquality reports and contributions. Lionus clearly sees AI as a useful tool, but not a replacement for understanding. And the real challenge here is making sure that AI solves real problems and doesn't create new ones.

[00:11:24]What are your thoughts about AI and the Linux kernel and open source? Post them in the comment section below. I'll catch you in another. Thanks for watching.