Linus Torvalds expressed his views on the use of AI in Linux kernel development, stating that 'Linux is not anti-AI' and 'if you don't like it, you can fork it.'

https://lore.kernel.org/linux-media/CAHk-=wi4zC+Ze8e+p3tMv8TtG_80KzsZ1syL9anBtmEh5Z40vg@mail.gmail.com/

Torvalds' remarks came up in a discussion about 'Sashiko,' a bug detection system, on the Linux kernel development mailing list. Sashiko is a system developed by Google employee Roman Gushchin that uses AI to find bugs in Linux kernel patches.

AI bug detection system 'Sashiko' developed by Google employee, named after the Japanese embroidery technique 'sashiko,' detects undiscovered bugs in the Linux kernel one after another - GIGAZINE

The mailing list has been discussing topics such as 'the pros and cons of using Sashiko for reviews,' and Gushchin has also joined the discussion, sending a message that reads, 'I understand the pros and cons of using AI to contribute to open-source projects. However, it is up to the project leader to decide what stance the Linux development project should take towards AI.'

In response, Torvalds said, 'No, that is not the Linux kernel's position. I know there are people who really hate AI, but this is an area where I, as a top-level maintainer, am firmly opposed. Linux is not an anti-AI project,' making it clear that the Linux kernel development project as a whole is not opposed to AI.

Furthermore, Torvalds stated, 'If someone feels there's a problem with our approach, they can either fork it the open-source way or simply walk away.'

He added, 'AI is just a tool, like any other tool we use. And AI is clearly a useful tool. That wasn't so clear even a year ago, but now there's no doubt about it. There are other questions about AI (like what the AI economy will actually be like in the end), but there's no doubt about whether it's useful. Anyone who doubts that has clearly never actually used it. From the perspective of the maintainer workload and 'continuing to find embarrassing bugs,' AI is a somewhat troublesome tool. But the solution isn't to ignore it, as some people are doing. The solution is to make LLMs help maintainers, not just cause them pain. There's no doubt about that.' He thus acknowledges that AI is a useful tool.

Torvalds continued, 'We're not forcing anyone to use AI, but we clearly ignore the opinions of those who oppose others using it. Sure, AI isn't perfect. But those who point out the flaws in AI should look in the mirror and point to themselves, because natural intelligence isn't always great. The kernel project has always been, and will continue to be, focused on the technology itself. Certainly, the social aspects of working with open source are important and a major motivation for the project. But ultimately, that's a secondary benefit, not the purpose of the project. The Linux kernel is not, and never will be, a so-called 'social activist' project. The kernel community works with open source not for religious reasons, but because it leads to better technology. In other words, we make decisions based primarily on technical merit, not on fear of new tools.'