[opine-at-random]

If you'd ever read even a single one of the licenses to the software I'm sure you use everyday, you'd understand. This is such an obvious and pathetic strawman.

I notice often on hackernews that people don't seem to understand anything about free or open-source software outside of the pragmatics of whether they can abuse the work for free.

[RamblingCTO]

You read a lot into my not so serious comment. Maybe internet comment sections aren't the right place for you.

But I'll bite: I know licensing, thank you. But what's copyrightable is not so easy. Licenses are not so easy. Copilot does not copy entire works and it's very questionable if a few lines of code are "piracy". It's a repeating discussion again and again, there's nothing novel about it except for the fact that a machine learns (and overfits for small portions of code). So please get off your high horse. I don't care for your fundamentalism.

[an1sotropy]

If you know this area of IP law then you know that LOTs of licenses, copyleft or not, require attribution (which copilot never does, and can’t do by its construction), and you know that what’s problematic is when the model output is arguably not “fair use”. Examples of that abound.

You don’t need any fundamentalism to know that copilot’s output carries huge and untested legal risk. If this lawsuit clears some of this up, that’s a big win for everyone.

[RamblingCTO]

Yes, you're right. But my point also was that it's not so easy when it's just a few lines, isn't it? Especially since this is an international issue and the definitions of copyrightable work is not easily definable.

> You don’t need any fundamentalism to know that copilot’s output carries huge and untested legal risk. If this lawsuit clears some of this up, that’s a big win for everyone.

I agree with that! I also see this as the only proper takeaway that I think is ok. The rest is making money off this thing. But the US has a different law suit culture anyway, which I find weird.