Comments by "JBird" (@jbird4478) on "Game over… GitHub Copilot X announced" video.

@Sammysapphira The fact that something is on the internet does not mean it is without copyright. Do you think movies are copyright free? Because they're all online as well.
76
@internallyinteral And both Kazaa and Napster were sued over copyright violations and subsequently went bankrupt.
40
@samuelbullard8978 They are being sued in a class action lawsuit. So hopefully, absent any new regulation, courts will rule that it already is against current copyright law.
32
@ghost mall AI is not doing similar to what humans do. Even from a technical perspective this isn't the case. Machine learning is a very simplified model that somewhat resembles only one aspect of human learning. No human programmer learned to program by reading millions upon millions of lines of production code. Ever. Likewise, you can teach a child to draw endless variations of cats using just one picture. That's not relevant to the law though. To the law, all that matters is that it is just a tool with no agency of its own. It is the people who train the AI who are violating copyright, by feeding that copyrighted material into an algorithm. A reminder that they themselves admit this violates copyright. They just claim it falls under the "fair use" clause of copyright law.
10
@ghost mall Open source code is owned and copyrighted just like closed source. People certainly do lay claim to it, and this claim is legally protected as well. And yes, Copilot has been found to exactly replicate code sometimes, even including comments. This tends to happen on the exact opposite condition; when a problem is so obscure it has had only one or two examples in its training data.
9
They are in fact being hit with a class action lawsuit; filed in a federal court in Cali in november last year against Microsoft, Github and OpenAI on behalf of several FOSS projects.
4
@everyhandletaken You don't have to prove that, because they don't deny that. The lawsuit nonetheless provides a few examples where code is reproduced verbatim (made obvious by the fact that code doesn't work because it has specific placeholders that are meant to be replaced by the user).
3
@ghost mall I don't see it as much of a dilemma. On one hand, you have people who've invested a lot of work in their software that they rightfully own. On the other hand, you have a company whose software wouldn't work as well without taking those people's work without permission. And what they sell is the ability to recreate said code. Mind you, they could simply train it using only code for which they have permission. They don't do that simply because their product wouldn't be as good. Somewhat worse even, because it is trained on everyone's code except Microsoft's own code.
3
Unless they loose in court, which they should.
1