AI is a tool, just like other tools we use.
This argument has, to me, always reeked of naivety. Heroin is a tool, but it has clear ethical implications, and we make attempts as a society to manage those. So is nuclear, and a particularly relevant one, since many leading AI figures have officially and explicitly stated AI is as dangerous as nuclear weapons. There are countless other examples of “tools” that require moral judgement (all of them, really).
I think Linus is being naive, and given his position of power, irresponsible.
And no, AI isn’t perfect.
I could write an essay about how much of an understatement this is, to the point of being wrong. Many, far more capable than me, already have.
This is NOT some kind of “social warrior” project, never has been, and never will be.
No, of course it’s not, but he seems to be treating it like ethics and context are for children.
It makes me worried for the Linux project. Not that he’s willing to use AI, but that this is how he sees it. It’s a learning experience.
I’m reminded of this quote that, to be honest I don’t really understand, but feels relevant
Our conventional response to all media, namely that it is how they are used that counts, is the numb stance of the technological idiot. For the ‘content’ of a medium is like the juicy piece of meat carried by the burglar to distract the watchdog of the mind…The effects of technology do not occur at the level of opinions or concepts, but alter sense ratios or patterns of perception steadily and without any resistance.
– Marshall McLuhan, Understanding Media
Once again, Linus steps in and delivers a reasonable take that cuts through the whole argument. It’s hard to imagine what Linux culture will be like without him.
Yeah without a single strong voice with implicit authority there’s really no telling what will happen i think
All it takes is one person using an LLM tainted with proprietary code which then just gives them that code line for line to undo decades of courtroom defense.
Not only that, but AI output can’t be licensed/copyrighted. The GPL license no longer covers the kernel in legal terms.
The GPL license no longer covers the kernel in legal terms.
The uncopyrightability of AI-written code only applies to the actual strings of code generated by an AI, not to the entire project.
A person could ignore the GPL if they only copied the AI-written portions. But, how could they know for sure which lines were AI generated and which were not? A wrong choice would leave them civilly liable for copyright violation and all they stand to gain would be tiny portions of the Linux kernel code which are worthless by themselves.
There’s no reason to steal the AI generated portions and risk a lawsuit, when you can just generate your own code.
There seems to be legal discussions about that. It’s not quite as simple as you say:
However, there may be cases in which a different assessment is justified, namely when users use and operate the LLM as a tool that merely implements their personal creative intent. This could be compared somewhat more vividly to using a paintbrush. If the brush merely rolls over the paper, for example because it is dropped, no copyright-protected work is created, even if paint remains on the paper. However, if a painter deliberately swings the brush in a certain way, a protected painting can be created. If AI is used in a comparable way a copyright-protected work can indeed be created.
https://kpmg-law.de/en/ai-and-copyright-what-is-permitted-when-using-llms/
Best to not believe anything KPMG says about AI.
Actually if KPMG say the sky is blue you should probably go outside and check and also make sure you still have your wallet.
yeah and the paintbrush somehow has abstracted access to millions of proprietary and copyleft licensed source code in forms of weight.
this is a clear misuse and abuse of any fair use rights, and clear push to centralisation of copyright to only a few companies with big budgets that can defend themselves.
i mean, can you really challenge and win against openai, a company backed by the govt, that your copyleft source code are misused as training data?
Yeah any decision would be on a case by case basis, which is normally something you’d want to avoid.
I’ve seen a couple of Linux devs talk about how they just give a prompt to claude and walk away leaving it alone to spit out the code, none of which can be licensed as GPL. But good luck working out what specific lines of what specific patches of theirs used an LLM vs. were re-written or such.
I’ve seen a couple of Linux devs talk about how they just give a prompt to claude and walk away leaving it alone to spit out the code
While I share Linus opinion on LLMs, I think doing this shit is extremely stupid and lazy.
And extremely abusive, since they don’t review the code fully, but a human must review the whole commit before accepting it. They save their time but consume that of others.
this is FUD and not true. projects are licensed not lines of code.
But if there are proprietary lines of code that end up in a project with an open license, you get a violation. 🤷♂️
what is a proprietary line of code? how do you look at a for loop or uf statement and say “i own that”. i can tell you that in all the huge applications i have worked on in the last 27 years, there isnt special proprietary code. there is proprietary data but not code.
the only time you really have proprietary codeis specialized code talking to a specialized device. so maybe a closed source driver.
I have worked at two companies back-to-back for a total of about 8 years, so significantly less time, that both definitely had proprietary code.
It is code that does things that nobody else is doing, or able to do, and patented, I believe(?).
But nice anecdote. 👍
Ist that a common thing that LLMs using proprietary code for coding tasks?
Because I don’t think so.
What do you think the LLM is using for reference when generating code? It can’t create from nothing.
They use everything for everything, that’s the big issue. Also gpl code. Anything they can trawl through they use. And replicate, in part or in full.
They take code snippets and copy and paste them? Or do they create own code based on what they’ve learned by trawling?
LLMs don’t “create”. Under the hood, they’re tokenizing the queries, looking for “clouds” of tokens that are similar to the query, then returning a sequence of tokens (with some random noise thrown in) that match what their training data says the answer should be.
In short: all LLM code is an amalgamation of their training data by definition. If there’s nothing similar in there, it’s literally not possible for it to be part of any response.
You’re exactly right. I should have used „generate“ instead of „create“.The point is I don’t think LLMs normally use copyrighted code in a way that would hurt open source projects.
Under the hood, they’re tokenizing the queries, looking for “clouds” of tokens that are similar to the query, then returning a sequence of tokens (with some random noise thrown in) that match what their training data says the answer should be.
Lol, so how do humans code in comparison?
Lol, so how do humans code in comparison?
By copy pasting from Stack Overflow
You’re exactly right. I should have used „generate“ instead of „create“
Did you purposely respond like an AI?
Damn, I forgot the em dash.
The point is I don’t think LLMs normally use copyrighted code in a way that would hurt open source projects.
I don’t know. I’m not a lawyer, and copyright for code was a hot mess even before LLMs got involved. With how many opportunistic copyright/patent trolls there are and how easily convinced judges have been in the past, it could go either way.
Lol, so how do humans code in comparison?
The good programmers normally code by breaking down the problem into constituent parts and logically working through the problem, step by step. What differentiates this from tokenization is that instead of just looking for code that is similar for a similar problem, programmers can usually understand the effects of each line of code, visualize what the state of each variable will be in that step (or dump out the variables to look directly if unsure), and then move on to the next step. This logical problem-solving approach is fundamentally different from a tokenization+noise looking for a similar-looking problem approach. For one thing, you can solve problems that haven’t been solved before.
Human programmers at least can tell you where they got a snippet they copied, whether it was in the docs, stack overflow or elsewhere, and you can try to keep attribution if you care about compliance. Not only that, but most of our skills are related to designing stuff and recognizing which pattern to use, the specific implementation isn’t necessary the same unless we go look for whatever we saw in the past, as our memories don’t just record everything and repeat it word by word. And after picking up a new language or framework I only need to look around when using a third party library or some API I’m less familiar with, or when something breaks.
I personally think both pro LLM and anti LLM are wrong. One group think they are gods. The other think they are demons. LLMs can be useful for programming to some extent. But they will create a disaster if you don’t know what are you doing. I have recently published a post about the matter on me blog. I think the best part is:
I strongly believe that LLMs are useful for programming to some extent. Imagine you have a shop and you get a robot to do the moves for you. So you instead focus on the main business concerns.
So if you want to make some changes to the code which don’t require intelligence, that is they are just mechanical tasks, LLMs are good. If you want the LLMs to understand semantics of your code, you have chosen the wrong tool. Maybe in future we’ll have new AI software and tools which also understand semantics to some extent. But I highly doubt a transformer will be able to do it. They just predict the next likely token.
There is something I haven’t yet added to the post. So I am writing it here. Our computers are Universal Turing Machines. There are some fundamental limits to what a turing machine can’t do. Those are called undecidable problems. For instance a turing machine can never check if two pieces of code are semantically equivalent[1]. But that’s what human programmers can do. That’s why I emphasize on tasks which require no intelligence.
[1] That’s about the general case. Sure there are exceptions. But as we say “exception is not the rule”.
This right here. It’s a tool.
Most of the “undecidable” are only undecidable for a subset of the problem instances, while a vast number of instances can be even trivially decidable. For example in the undecidable halting problem, both you and a computer can trivially deduce that
while(true)will not halt. In the same way a computer can deduce that many instances of two pieces of code are semantically equivalent.I’d like to see an instance of the problem where a human could decide it and the computer could not.
Hey. The number of problems which can are decidable are infinite as are those which are not. But as soon as there is a backward jump in your code, a Turing machine most likely won’t be able to decide if it’ll halt or not. The
while(true)is an exception. In the real world we have a great number of programs whose loops cannot be decided by a Turing machine. But the programmer who has written the code knows when the loop will terminate.If we see the machine code, if there is a conditional backward jump(unlike
while(true)which is unconditional), in the general case it’s undecidable.
LLMs are just tools, the problem is managers who all think they’re god machines and it causes hell on earth
But that’s what human programmers can do
I don’t think humans can solve the undecidable problems. If I understand them correctly, they’re like a law of nature; extends into mathematics, logic, and any formal system (e.g. Godel’s Incompleteness Theorem).
I didn’t mean humans can solve ALL undecidable problems. As I have written in me reply to wicked, a programmer usually writes loops and the turing machine cannot decide their halting, for a vast majority of them.
Yes! And the problem with decidability is you don’t which problem will be undecidable. You can only just try and they it’s like “well couldn’t find a solution, so it might be undecidable”. So we’ll need to put limits on how many tokens an AI should use before giving up, put limits on what it tries to do or we’ll be burning thousands of dollars on tokens and coming up with nothing. There will need to be a lot of judgement used on where we apply the algorithm.
Judging when to use an algorithm and when not to use an algorithm based around how expensive it will be in terms of resources? That’s just another day for a software engineer.
You have a valid point. I think currently there are assists which we know an LLM definitely can do for us. Like mechanical tasks. And tasks which we know LLM can never do. In between there is a gray area. I think we will learn about the gray area over time.
Also I would like automatic generation of programs. But I don’t think an LLM is the last stop for it. I personally research in the field of Genetic Programming. I strongly believe in future we could have tools which generate or optimize programs when guided by a human. Currently we have Evolutionary Art and Music which do the same. Tho they are far from being actually usable, this is the same with every technology in the beginning.
Also see this talk: https://cr.yp.to/talks/2015.04.16/slides-djb-20150416-a4.pdf
Human written code these days feels equivalent to a unique and soulful artisan made item whereas AI code is like a soulless and defected factory made imitation. I’d much, much, much, rather support artisans over factory made slop and even before AI, artisan work has been well known to be significantly higher quality than factory made stuff. For something as foundational and important as a kernel, I really think AI has no place in it.
All I care about is whether it works and is secure. Bonus points for cheaper and faster development. If artisan code gets us there, sure. If AI code gets us there, great. I trust Linus to know what works and what doesn’t.
I care about the environment. 🤷♂️
I suppose the big issue for me is that I’m also an artist and I view code similarly, especially when I combine code with my graphics. I understand though that with a large project you are going to have some amount of people using AI even if you try to filter them out, so I can partially understand his stance. However I really disagree with him when he says it’s a useful tool, given how AI causes brain rot, productivity losses, and environmental destruction.
As for productivity losses and environmental destruction, that’s sort of what I was referring to when I said “bonus points for cheaper and faster development”. The environmental destruction refers to energy usage. Productivity can be thought of as time usage. If AI can be cheaper than humans in both time and energy, then that’s a win.
There’s a related discussion about how the context here is mainly about finding vulnerabilies with AI. And that is one where AI does seem to be faster and cheaper than using humans. Since AI is now finding kernel vulnerabilities that have been lurking there for 15+ years.
Maybe code generation will get there too.
As an artist you might also be worried about how AI is trained on copyrighted data. But I personally don’t really care about copyright
While I am very anti-genAI, I’m much more okay with analytical AI as long as it’s not blindly trusted. Technically they don’t need an LLM as they could use something smaller and optimized just to find vulnerabilities. I still take issue with ethical and environmental concerns but I think there are solutions to them. We just need to slow down, take the time to do things right, and ditch the whole “move fast and break things” mindset commonly seen in tech, but I doubt that will ever happen.
What’s your argument against generative AI but not analytical AI? From what I’ve seen the arguments for/against one also work for the other
My viewpoint is that genAI will always cause cognitive decline no matter what however I haven’t come across any studies yet about analytical AI also brain rotting people. If it’s trained on data with consent from it’s owners’ and they find a solution to the environmental problems, I’d be fine with it as long as it’s used for good purposes.
This post about AI detecting cancer makes me believe it might be genuinely useful, although of course it can be used for despicable, evil purposes as well.
Also, humans really like making/designing things. I don’t see the point in automating activities that many people love doing. Every time I come across an ad advertising image gen or whatever it just feels so out of touch with humanity.
Fair. I think if somebody gets brainrot from using AI that’s their own problem, just like a programmer losing their coding skills after become a product manager. I don’t see how that makes AI bad.
I don’t see the point in automating activities that many people love doing.
Some people like the process, some people like the product. For those who only care about the product, why not let them use AI? Doesn’t stop you from enjoying the process.
I admit that this is getting pretty far from the original discussion though so no worries if you’d rather end it here. I just have fun thinking about these things
Even artisans use tools.
OP didn’t say they didn’t
Artisans own their tools, that’s what makes them artisans (rather than wage workers). I believe after the bubble pops there will be legitimate uses for the tech, and we’ll be able to run a pretty good iteration in our own hardware, but as it is now I’m uncomfortable with my employer having the power to decide whether I have the tools they want me to use for the job, whereas with code that’s less of a concern.
There’s already pretty decent open weight models that you can run on your own hardware. They won’t be as fast as the models running in a datacenter but they will get the job done while not adding more money to the garbage fire that is ‘The AI Industry’.
I don’t think that the future of AI is as a massive subscription service industry, that is just rampant capitalism on full display (with all of the damage that it causes).
I was thinking more that lots of people don’t have the hardware required to begin with, and it’s really hard to purchase now (especially in poorer countries like mine), hopefully memory prices will come down at some point after the bubble bursts so we can afford shit
It’s really a race between the AI bubble bursting and hardware manufacturers building more production to chase the higher prices.
One of China’s largest chip producers is building several new fabs and their domestic fabs (i.e. ones not made by ASML) are capable of producing DDR5. Once this price shock ends I wouldn’t be surprised to see hardware prices drop to below levels they were pre-2020.
What we’re seeing is the effect of companies having hundreds of billions of dollars in cash on hand and all deciding to dump it into computer hardware at the same time. They can’t burn capital forever and AI isn’t remotely what they’re selling it as so they’re finding a hard time generating any meaningful revenue. All they’re doing is shoveling money to hardware manufacturers, who will use it to build out extra capacity and once the spending frenzy ends we’ll have way more supply than demand.
Of course, given how much money is floating around, that could be 5 years from now.
This “argument” really shines the light on just how dumb pro AI people are.
This ‘ad hominem’ really shines the light on just how toxic anti-AI people are.
Did you have to ask ChatGPT how you felt about my comment then to create a response?
Are you naturally an asshole or do you just enjoy fighting strawmen on the Internet?
That’s why you like AI. You can say anything and have it stroke your ego. Real people challenge you and I bet that feels icky.
No, it is you that likes AI. I can’t believe you support datacenters being built in endangered habitats, you’re such an bad person to be carrying water for AI. How does Sam Altman’s butthole taste?
Oh, did I just make up shit about you and insult you using this fabricated information? What an ignorant thing to do on the Internet. Don’t you agree, AI lover?
Is it really his foot, or is it just vibe-coded?
I find it difficult to trust someone named linus
The discussion should stay technical.
Restricting discourse to a single realm makes it by definition a monoculture, which then intrinsically debases it to a fragile and brittle construct.
LLMs may well be useful and/or effective from narrowly myopic perspectives, but I’m yet to identify any which could legitimately be framed as justifiable or ethical.
It’s particularly galling given a fundamental aspect of how LLMs function is to incorporate myriad dimensions of input, yet to arrive at anything resembling a defence requires cherry picking arguments.
