There is literally never any discussion on these forums about case law around software copyright. It’s not like the resultant jurisprudence is illogical or arbitrary. The courts have plenty of limitations on what can and cannot be restricted by copyright and the related opinions make perfectly reasonable arguments.
If you want these things to change you need to arm yourself with the knowledge necessary to not only convince the courts but also your peers outside of the FOSS movement. Merely complaining about things being “unfair” is not going to achieve much outside of some virtual pats on the back.
> The absurdity of Microsoft and OpenAI’s legal justification for GitHub Copilot.
> Note 1: See Comment Regarding Request for Comments on Intellectual Property Protection for Artificial Intelligence Innovation (application/pdf) submitted by OpenAI to the USPTO
With where I predict the author is heading, I'll point out that that this document clarifies:
> > We do not claim that AI systems are invariably beneficial or non-infringing. Rather, we address the narrow question of whether training AI systems on copyrighted data constitutes copyright infringement.
> > Generative AI systems might generate output media that infringes on existing copyrighted works. We think that this is an unlikely accidental outcome of well-constructed generative AI systems, though it remains possible due to overfitting or developers’ intentions.
----
> It contains a screen recording of Copilot suggesting this Quake code.
I'd call the fast inverse square root algorithm "folk code", passed down and altered multiple times before being misattributed to Carmack. It appears in many repos, often with permissive licenses like WTFPL. I don't think it's generated by Copilot anymore, after steps to prevent exact regurgitations.
As a challenge, since I see the same one example used so much, I'd be interested to see if there are any examples of:
1. With the public version of Github Copilot, not opting out of any protections...
2. Using a prompt that could be entered incidentally...
3. Generate code of problematic length and similarity to existing code...
4. Which is not (even ostensibly) under a license allowing such copying
This tests whether Copilot would be an issue for someone using it earnestly. Whether it could be used to intentionally generate protected code is another question.
Elaboration/reasoning:
1. To allow corroboration and demonstrate that it's an issue on the service being sold commercially. My understanding is that Copilot is deterministic
2. As opposed to, say, intentionally copying 90% of a source file and having it complete the last function. The prompt code appearing in other unrelated repos would be sufficient to show this
3. Doesn't have to be an exact regurgitation, but I think it should at least by unambiguous that the given code is what's being copied
4. Most licenses require at least attribution. "even ostensibly": someone might share code under public domain without permission to do so, but here even an honest human could make the same mistake of copying it
If you want these things to change you need to arm yourself with the knowledge necessary to not only convince the courts but also your peers outside of the FOSS movement. Merely complaining about things being “unfair” is not going to achieve much outside of some virtual pats on the back.