Linus Torvalds tells AI critics to fork Linux or walk away

[midian182]

A hot potato: Linus Torvalds' stance toward AI appears to be softening. After pointing out that AI tech was 90% marketing and 10% factual reality in 2024, Torvalds just said that Linux is not an "anti-AI" project, and that anyone who has an issue with this can either walk away or fork the kernel.

Torvalds' comments came in response to a discussion about anti-LLM feelings on lore.kernel.org, the official public archive for Linux kernel development mailing lists. He said that while he realizes some people really dislike AI, it's an area where he's willing to "absolutely put my foot down as the top-level maintainer."

"Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it. Or just walk away."

The Linux creator added that AI is a tool, just like any other, and a useful one at that. He said that while there are other questions around AI, such as what its economic impact will ultimately look like, "is it useful" was no longer one of those questions.

AI can be a "somewhat painful tool, both for maintainer workloads and just from an 'it keeps finding embarrassing bugs' standpoint," Torvalds said. "But the solution is not to put your head in the sand and sing 'La La La, I can't hear you' at the top of your voice like some people seem to do."

"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."

Torvalds admitted that AI isn't perfect, but he believes anyone who points to the technology's problems 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," he said.

The Finnish software engineer appeared to take a more cautious approach to AI back in 2024. In addition to complaining that it was 90% marketing hype, he said it would take five years for it to become clear what AI is really useful for. It looks like that usefulness has become apparent in a shorter timeframe than expected.

– Tsarathustra (@tsarnick) October 21, 2024

In January, Torvalds said he had started using AI to write code, though only as part of a small personal project rather than on Linux. A few months later, he criticized Linux developers for filing poorly timed bug reports shortly before an RC5 release, noting that some relied on AI to flag trivial issues. He added that many of the proposed fixes were also AI-generated and often introduced unnecessary bloat into the Linux kernel instead of addressing the underlying problem.

Torvalds' comments suggest AI has already moved beyond the experimental stage for Linux, even if developers are still grappling with the extra workload, questionable fixes, and occasional kernel bloat it creates.

Permalink to story:

Linus Torvalds tells AI critics to fork Linux or walk away

 

[mbrowne5061]

Sounds like he isn't against the concept of AI/LLM assistance with coding, which makes sense broadly speaking.

But I fully expect him to be against relying on a for-profit corporate licensing structure (I.e. Claude, Copilot, etc) being involved in any critical Linux development. I suspect he will try to come up with some way to add a reasonably capable local-run LLM to most dev environments. The latest models from the big companies already seem to be achieving diminishing returns; the improvements to coding output relative to prior models are there, but marginal, while costs increases are exponential.

Pie in the sky would an FOSS dev environment that gave you AI-assisted bug detection, automatically wrote more 'standard' modules that followed best practices, and wrote documentation; all while running on a typical desktop GPU. Let the dev work on the 'new' bits, while the AI looked over their shoulder to write the 'boring bits' of code, flag bugs for review (inc. documenting how to reproduce), and generate a draft of a Wiki for the project.

 

[LordVile95]

Of course he’ll say that they’re the ones paying him

 

[Endymio]

After pointing out that AI tech was 90% marketing and 10% factual reality in 2024...he said it would take five years for it to become clear what AI is really useful for.

He didn't "point out" a fact, he made a claim ... a claim which he's already had to walk back, as less than two years later, he himself was using AI to write code.

 

[Hecate91]

Linus must be making quite the profit from some AI companies to be saying this,
it's sad to see that even Linux isn't safe from AI ensh#$ification.
Also forking Linux even further fragmenting it into versions that might have less or no AI crapware won't help adoption of Linux at all.

 

[Lowhangingfruit]

The Linux creator added that AI is a tool, just like any other, and a useful one at that. He said that while there are other questions around AI, such as what its economic impact will ultimately look like, "is it useful" was no longer one of those questions.

This is the most profound statement. That's where things are people. If you're still one of those people saying AI is just a fad, and a bubble, and it will go away you're in denial.

The problems to deal with now are code bloat, human skill and knowledge atrophy, leveraging business competitive advantages when everyone can copy you in minutes etc.

 

[LordVile95]

Linus must be making quite the profit from some AI companies to be saying this,
it's sad to see that even Linux isn't safe from AI ensh#$ification.
Also forking Linux even further fragmenting it into versions that might have less or no AI crapware won't help adoption of Linux at all.

He earns his salary from the Linux foundation which is funded largely by…

*Checks notes*

Google, Microsoft, AWS and Meta

Think that’s your answer

 

[justarando]

I am not sure why people keep treating this man like some sort of demi-God.
The man is loud and very rude even when he is proven wrong or changes his mind

 

[OortCloud]

All these pro-AI folks completely avoid the most important reason why people hate AI. They hate it because they can see where it will lead, and they can see the societal and environmental damage it has already caused and which is only going to get far worse. Almost nothing AI provides helps the common man and it never will. In fact, it is patently obvious that the common man will be shafted by AI at every opportunity; this has already happened repeatedly. But the bro's just ignore all of that because most of them are driven by nothing but greed or in the case of a few like Linus Torvald they can't see beyond technological benefits.

 

[bviktor]

Linus must be making quite the profit from some AI companies to be saying this,
it's sad to see that even Linux isn't safe from AI ensh#$ification.
Also forking Linux even further fragmenting it into versions that might have less or no AI crapware won't help adoption of Linux at all.

But it is safe.

Linus is consistently rejecting AI code that's actually cr@p.

All he's saying is that he doesn't categorically reject anything AI just because.

Because, unlike most people, he has brains, and doesn't think in black and white.

 

[Northy]

I am not sure why people keep treating this man like some sort of demi-God.

He's the guy who wrote the Linux kernel that the worlds computing infrastructure runs on, and didn't monetize it but made it FOSS. He also created Git and just gave that away too. He could be a billionaire many times over if he wanted but has chosen a much more noble path.

The man is loud and very rude even when he is proven wrong or changes his mind

Someone's feelings got hurt by mean words from Linus, many times.

Linus must be making quite the profit from some AI companies to be saying this,

He doesn't need money, has never tried to make computing about money.

it's sad to see that even Linux isn't safe from AI ensh#$ification.
Also forking Linux even further fragmenting it into versions that might have less or no AI crapware won't help adoption of Linux at all.

AI isn't a chat bot and image generator, that's just a "fun" thing for people who don't use data to play with. Making bullet proof usable systems has always been priority to Linux project. I run LLM's locally on my Linux tower for data processing, it'll be great to have more compatibility and not have to be using off distro projects.

All these pro-AI folks completely avoid the most important reason why people hate AI. They hate it because they can see where it will lead, and they can see the societal and environmental damage it has already caused and which is only going to get far worse. Almost nothing AI provides helps the common man and it never will. In fact, it is patently obvious that the common man will be shafted by AI at every opportunity; this has already happened repeatedly. But the bro's just ignore all of that because most of them are driven by nothing but greed or in the case of a few like Linus Torvald they can't see beyond technological benefits.

Doesn't matter whether a person is pro or anti, it is happening and there's no stopping it. Adapt or die.

Linus is consistently rejecting AI code that's actually cr@p.

All he's saying is that he doesn't categorically reject anything AI just because.

Because, unlike most people, he has brains, and doesn't think in black and white.

Yeah, he's a smart man, often able to see implications long before anyone else.

 

[Axeia]

People who think he's motivated by money - you might want to read up on the man. He's got enough of it to do what he wants to do in life and doesn't seem particularly interested in making more.
One of the few people that has my utmost respect.

For what he has been doing for the last many many years it's basically just approving contributions to the kernel. Probably not something where you run into the typical person without any knowledge sending in a ton of code that was fully AI generated, mostly very knowledgeable people that will double check AI generated code and correct it where needed. Shunning it can't really be enforced and it would just be shooting yourself in the foot trying to do so.

Other projects where every other submission (or even more) is some vibe coded POS from some random guy who thinks he's a coder after typing something in a prompt until it starts working... There I understand they do try to shun it, no one likes wasting their time.

There's a difference between being a tool and using AI as the tool .

 

[hahahanoobs]

Linux community in disarray after hearing this from their saviour, LT.

 

[LordVile95]

People who think he's motivated by money - you might want to read up on the man. He's got enough of it to do what he wants to do in life and doesn't seem particularly interested in making more.
One of the few people that has my utmost respect.

For what he has been doing for the last many many years it's basically just approving contributions to the kernel. Probably not something where you run into the typical person without any knowledge sending in a ton of code that was fully AI generated, mostly very knowledgeable people that will double check AI generated code and correct it where needed. Shunning it can't really be enforced and it would just be shooting yourself in the foot trying to do so.

Other projects where every other submission (or even more) is some vibe coded POS from some random guy who thinks he's a coder after typing something in a prompt until it starts working... There I understand they do try to shun it, no one likes wasting their time.

There's a difference between being a tool and using AI as the tool .

If he’s not interested in money why does he take a huge salary from a non profit?

 

[Axeia]

If he’s not interested in money why does he take a huge salary from a non profit?

Considering his skills it's rather modest. The man created both the kernel for the world's most popular OS and the most used source code management system.

He could make several times his current wage working for any number of companies. He could just make some random new project and start raising god knows how much money (imagine him announcing his own AI thing with how willing investors are at throwing money at things atm).
Tons of ways where he could monetize on his his clout far more effectively if that was the goal.

It seems pretty clear to me from interviews and videos that he just cares and feels responsible for how the code of the Linux kernel is managed. Git was a means to that end and he has taken care that the right people and systems are in place that things will continue as they have been once he retires or passes away.

 

[KnightSayNI]

With AI you can spend seconds and maybe accomplish something. With linux you can spend hours, days, generations and most likely accomplish nothing.

 

[LordVile95]

Considering his skills it's rather modest. The man created both the kernel for the world's most popular OS and the most used source code management system.

Over 30 years ago, software has moved on. That’s like saying Woz could still design cutting edge PCs

He could make several times his current wage working for any number of companies. He could just make some random new project and start raising god knows how much money (imagine him announcing his own AI thing with how willing investors are at throwing money at things atm).
Tons of ways where he could monetize on his his clout far more effectively if that was the goal.

He really couldn’t, being a coder from 30 years ago wouldn’t really get you that much.

It seems pretty clear to me from interviews and videos that he just cares and feels responsible for how the code of the Linux kernel is managed. Git was a means to that end and he has taken care that the right people and systems are in place that things will continue as they have been once he retires or passes away.

No he cares about the money he’s getting for it. The only way he’s going to maintain his income if for silicon valley to keep donating to the non profit that pays his wages.

 

[Endymio]

Over 30 years ago, software has moved on. That’s like saying Woz could still design cutting edge PCs

This is the most absurd statement I've heard yet from you. The Linux kernel and core components are still today written in the same languages and using the same tools and design principles as from 30 years ago -- and today's kernel is not only still being code-reviewed by Torvalds, but he still writes code for it.

The most ironic part of your comment is that the only major change between then and now is the adoption of AI in coding -- and you're knocking him for embracing it.

 

[Honest Abe]

He's accepted that the landscape is evolving. AI is getting way better than it was in 2024, especially in the cyber security field. Criminals will use AI to find vulnerabilities, and companies must use AI to find and patch them before criminals take advantage of it first.

We're in a cat and mouse game. But it's a good thing. The future of computing will be much more secure when you have red team/blue team hammering on software via AI before packages are officially released to the public.

The scary thing right now is our adversaries, such as China, evolving this type of AI faster than the west. They simply don't have the same bureaucratic limitations we do, so they really may win the race.

 

[wiyosaya]

After cases like this - https://www.rawstory.com/todd-blanche-2677234639/
Embarrassing is not quite descriptive enough, IMO.

 

[Endymio]

The scary thing right now is our adversaries, such as China, evolving this type of AI faster than the west. They simply don't have the same bureaucratic limitations we do, so they really may win the race.

Even more frightening is the amount of money China is spending to fund anti-AI groups and propaganda here in the US ... for the obvious reason they intend to win that race.

 

[Gezzer]

I've said it before LLM based AI isn't really intelligent in the human sense. As he said it's a tool... with limited uses. And like any tool it's the user that actually dictates how well it will perform in regards to those uses. The big problem as I see it is the unbelievable overvalued hype it currently creates. The massive amounts of resources it takes. And more importantly the idea that it can replace all human value in a vast array of production scenarios.

It simply can't, period. Even as LLM based AI improves, it'll never overcome the limitations inherent in the fact it's a language based comparative database. The AI has no concept of what it's spewing out, it simply links data with other data and gives the end user the result. Which can seem magical, at times. Other times due to those inherent limitations it craps the bed big time.

It's like a nail gun compared to a simple hammer. Neither can build a house. It takes a skilled worker to use either. The gun simply makes it faster to build one. And as someone that's worked in the construction industry in my youth, I can tell you I've seen more examples of poorly nailed joists that used a gun than used a hammer.

But because there's lots of money to be made (at least right now) the overvaluation will continue. And yes there'll be a bubble bursting eventually. But that doesn't mean that LLM based AI will fade away. Again it's a useful tool. It just needs to evolve beyond its current "miracle do anything" concept to a much more realistic use. Who knows eventually we might see a true AI that will fulfill the promise currently being given to LLM based AI. But IMHO we're a long way from that point and may never reach it.

 

[scoffer]

Ha! What the actual **** does Torvalds know about Linux?



/s ( because I'm sure someone will fall for it).

 

[RudyBob]

Good thing Carmack didn't say this

 

[scavengerspc]

Damn that dudes head is getting as big as, well, his head.