A community for everything relating to the GNU/Linux operating system (except the memes!)
Also, check out:
Original icon base courtesy of lewing@isc.tamu.edu and The GIMP
The actual body of his message seems to be more about him taking a stance that AI is a tool and people working on the project are free to choose to use it because it does have technical merit, which he considers the deciding factor in the case of the Linux kernel.
It's actually a pretty calm and reasonable message, I don't think the sensationalist title is warranted.
But, we don't really want to encourage the use of AI as a technology, despite its helpfulness. Its destroying so much already like clean water, habitats for animals, the sound generated from there data centers drives off wildlife. The pollution (Grok running on jet engine turbines to power the TX plants ), etc.
Unless there are those running their own tiny instances, it's not really worth it.
As people mentioned in other publication about this those problems, which are devastating, are not being caused by the technology itself but the corpos (and the diabolical and insatiable greed of those who run them) behind it. As you are concluding, local llms can be fine, though I'd think he could have a stronger instance about that point - trying to discourage people about using corporate llms and incentivating local llms.
I agree with you, and also consider the current AI setup destructive and unethical all the way from the model creation from mostly stolen data to the data centers to run it with the issue you listed.
I would also prefer if more people spoke up on that to fasten the demise of this current hype.
That being said I also realise that Linux (and therefore the kernel) is incredibly important in a different fight for privacy and owning our own devices. If it were to fall behind because they completely lock themselves off from the admittedly useful aspects of AI, it could backfire in a different way.
I think it's a lose-lose situation here and he understandably picked the side that's more impactful for the Linux kernel (and him).
He does literally say if people have issues with that stance, they can fork Linux..
Yes, and that sentiment captures exactly 2 sentences of a many paragraph email. It happens to be the part that can be interpreted most like drama from it, hence its a sensationalist title.
Yeah but in FOSS sense.
it can be really dumb, but it can also be extremely useful for some things. i think that's pretty much linus' stance--and if you do use it, you are the one wholly responsible for what comes out of it.
This headline is being unreasonably melodramatic
That's true of pretty much every headline on the topic of AI.
The implication is that the contributor is responsible for the AI-generated code, but I look at this as an inevitable slippery slope as maintainers feel pressured to rely more and more on AI for reviewing the massive amount of slop contributions that will be submitted if you don't put restrictions on AI use. The end result is an exponential growth of tech debt and making the codebase truly awful to work on, until AI contributions are reigned in.
I've been in a position of reviewing a ton of slop. It was this way before AI and even worse after. I think the appropriate response is that I don't care if it's human slop or AI slop - I'm not going to do a ton of work to review it if I see slop. I'm going to make a comment that you should go deslop it and let me know when you're done.
I was also wondering about the whole "AI content is not copyrightable" thing... Does that potentially screw up copyleft licences?
I was also wondering about the whole “AI content is not copyrightable” thing… Does that potentially screw up copyleft licences?
In practice, no. Simply containing AI content doesn't invalidate copyright or licenses.
Purely AI generated content is not copyrightable, but the Linux kernel does not fall under that category.
Works that have a human in the loop are, for the parts that are human created, copyrightable. So if a developer uses AI to assist them in writing code (fixing formatting, detecting bugs, refactoring) then the code is copyrightable.
Someone could technically take the AI generated portions of the Linux kernel and use them outside of the licensing restrictions... but since it isn't possible to determine which lines of code were AI generated and which were human generated then it is very risky to do. If they were sued by the Linux Foundation, and found to have used code that was verifiably human-written (a developer testifying that they personally wrote the code in question would satisfy that requirement) then they would be civilly liable.
Since the Linux kernel is mostly human-written, simply extracting the AI-written parts would likely not provide any actual useful functionality and it would come with the risk of violating copyright if you chose the wrong lines to copy. It's a large risk for very little reward.
After all, why copy AI generated code which would potentially lead to you being sued when you can just make your own AI generated code.
Better debate than endless culture wars.