I think you’re misunderstanding what I’m saying. Any portions of the kernel that are public domain can be used by anyone for any purpose without following the terms of the GPL. AI generated code is public domain. To make sure all parts of the kernel are protected by the GPL, public domain code should not be accepted unless absolutely necessary.
I don’t see the problem. GPL protects all of the code that is copyrighted, i.e. 100% made by humans. Accepting a submission created with AI tools doesn’t change this. It’s not going to be a simple task for someone who has decided to violate the GPL license to only use the generated/uncopyrighted portions without using any other GPL code and thus being subject to GPL licensing terms.
These hypothetical GPL violating people will have a hard time using lines 27-38 of ./kernel/events/ring_buffer.c to do anything even if they technically can do so without releasing their code under the GPL. If they use any piece of GPL code, at all, anywhere, their entire project is required to follow the GPL. So while they could, technically, take 27-38 of ring_buffer.c and build an entire proprietary non-GPL Linux kernel… it is, in practice, not feasible even if it technically possible.
So what happens thirty years from now when 95% of the kernel code is AI generated? It’ll be a lot easier to rewrite the parts that aren’t, and have a fully closed source kernel that you can use without following the GPL.
The short answer is that this is a slippery slope argument.
The long answer is:
In this hypothetical future where 95% of the Linux kernel is AI generated, it stands to reason that generating an OS kernel is possible (by definition of the hypothetical).
If generating a full OS kernel is possible then people could generate a fully closed source kernel without using any of the 5% of GPL protected code in the Linux kernel.
If you allow that it’s possible for AI to create a kernel with AI generated code then it will happen regardless of the status of the Linux kernel’s copyright protections.
Generating code costs a lot of money, as does the expertise to review the code. People aren’t going to want to spend the many millions of dollars to do that when they could use a GPL kernel. Of course if the kernel is not only free, but basically public domain, it solves all of their problems. They can modify it and keep those modifications closed source, the complete antithesis of what the GPL stands for.
Well, that isn’t the case now and isn’t likely to be the case anytime in the near future.
The rules are not written in stone and future Linus will have a better idea of the capabilities of future AI and can change the rules accordingly, as he has done since the beginning, in order to steer the Linux project in the right direction.
I think you’re misunderstanding what I’m saying. Any portions of the kernel that are public domain can be used by anyone for any purpose without following the terms of the GPL. AI generated code is public domain. To make sure all parts of the kernel are protected by the GPL, public domain code should not be accepted unless absolutely necessary.
I don’t see the problem. GPL protects all of the code that is copyrighted, i.e. 100% made by humans. Accepting a submission created with AI tools doesn’t change this. It’s not going to be a simple task for someone who has decided to violate the GPL license to only use the generated/uncopyrighted portions without using any other GPL code and thus being subject to GPL licensing terms.
These hypothetical GPL violating people will have a hard time using lines 27-38 of ./kernel/events/ring_buffer.c to do anything even if they technically can do so without releasing their code under the GPL. If they use any piece of GPL code, at all, anywhere, their entire project is required to follow the GPL. So while they could, technically, take 27-38 of ring_buffer.c and build an entire proprietary non-GPL Linux kernel… it is, in practice, not feasible even if it technically possible.
So what happens thirty years from now when 95% of the kernel code is AI generated? It’ll be a lot easier to rewrite the parts that aren’t, and have a fully closed source kernel that you can use without following the GPL.
The short answer is that this is a slippery slope argument.
The long answer is:
In this hypothetical future where 95% of the Linux kernel is AI generated, it stands to reason that generating an OS kernel is possible (by definition of the hypothetical).
If generating a full OS kernel is possible then people could generate a fully closed source kernel without using any of the 5% of GPL protected code in the Linux kernel.
If you allow that it’s possible for AI to create a kernel with AI generated code then it will happen regardless of the status of the Linux kernel’s copyright protections.
Generating code costs a lot of money, as does the expertise to review the code. People aren’t going to want to spend the many millions of dollars to do that when they could use a GPL kernel. Of course if the kernel is not only free, but basically public domain, it solves all of their problems. They can modify it and keep those modifications closed source, the complete antithesis of what the GPL stands for.
Well, that isn’t the case now and isn’t likely to be the case anytime in the near future.
The rules are not written in stone and future Linus will have a better idea of the capabilities of future AI and can change the rules accordingly, as he has done since the beginning, in order to steer the Linux project in the right direction.