Slop code can’t be licensed, the license Lutris had is invalid. That means you can’t even fork it unless you can identify the point slop started and fork prior to that.
There’s no guarantee the developer who uses slop code even knows what it’s doing. Github is full of nonsense like that now.
That part isn’t true whatsoever. Public domain doesn’t work like a poison pill. If I put the collected works of Arthur Conan Doyle on my website, that doesn’t mean the site can’t be CC BY-SA 4.0; it means someone is free to take that specific part without worrying about CC BY-SA 4.0.
Even for the slop code where it’s more mixed in, it’s on the person trying to take it to affirmatively show the part they’re taking was generated by an LLM, i.e. you can’t just say “this is slop, I’m taking this” and expect a project to have to affirmatively defend that it’s human-written.
And lastly, AI assistance doesn’t automatically even poison the code being created/changed with it. This one is murky, but so far, the general consensus is that so long as you, the human, worked on it – i.e. that you didn’t just let the slop machine run wild and take whatever it outputted without your own changes – it’s still your copyright.
The way the Linux kernel developers use llms is to help identify potential problems, the llm doesn’t write any code. It’s a human that looks at the potential problem to determine if it actually is one and then writes the code to fix it.
Microsoft generates large swathes of code and judging by how poorly it preforms when they immediately push it to production there is little to no human oversight.
To answer your question, no to my first example and yes to my second.
Even if it runs fine there are still problems
That part isn’t true whatsoever. Public domain doesn’t work like a poison pill. If I put the collected works of Arthur Conan Doyle on my website, that doesn’t mean the site can’t be CC BY-SA 4.0; it means someone is free to take that specific part without worrying about CC BY-SA 4.0.
Even for the slop code where it’s more mixed in, it’s on the person trying to take it to affirmatively show the part they’re taking was generated by an LLM, i.e. you can’t just say “this is slop, I’m taking this” and expect a project to have to affirmatively defend that it’s human-written.
And lastly, AI assistance doesn’t automatically even poison the code being created/changed with it. This one is murky, but so far, the general consensus is that so long as you, the human, worked on it – i.e. that you didn’t just let the slop machine run wild and take whatever it outputted without your own changes – it’s still your copyright.
—someone who doesn’t do slop
Don’t all the big software companies use AI, are you suggesting that everything actually has an invalid licence?
The way the Linux kernel developers use llms is to help identify potential problems, the llm doesn’t write any code. It’s a human that looks at the potential problem to determine if it actually is one and then writes the code to fix it.
Microsoft generates large swathes of code and judging by how poorly it preforms when they immediately push it to production there is little to no human oversight.
To answer your question, no to my first example and yes to my second.
In a good and just world, this would be the case.