You'll notice I didn't mention copyright, and that's because I don't think it's relevant. I think your analysis is correct. Keep in mind too that Stockfish – the engine I work on, although I wasn't at the time – was the subject of a high-profile defense of the GPL in Germany: https://stockfishchess.org/blog/2021/our-lawsuit-against-che.... That case was different as it involved wholesale copying of the source code, rather than porting ideas. Ideas shouldn't be copyrightable.

But copyright and plagiarism are orthogonal, and questions of morality are much more tied to the latter. It would be illegal (in the United States) for me to publish a copy of Nineteen Eighty-Four, but not immoral. It was not immoral, in my view, for Aaron Swartz to try to liberate JSTor articles.

That doesn't make plagiarism acceptable, either. And therefore, I'm willing to call it out when I see it.

No idea why you're getting downvoted. You're right.

> it's still eerie to me to see hackers coming down hard on a side of an issue we spend decades passionately defending the other side of

Yeah, it's surreal. It's not just intellectual property either, Hackers have already started complaining about automation now too.

Computers must have replaced tens of millions of jobs. It's the ultimate job destroyer, since it allows capital to do more with less. Automating me, a doctor, out of his job is essentially the holy grail of Hacker News.

But when hackers are automated away, then suddenly it's a crime against humanity.

I think that's true, but I'm not sure if this would characterize itself as "lifting ideas". I mean, it's one thing to personally take the idea and re-implement it yourself elsewhere. But I have to believe there could be an argument, that having an inference algorithm take the original source and re-implement it in the context of your engine might count as derivative work.

A particular work is deliberately supplied at inference time so the system can recreate or adapt it for another environment. It resembles commissioning a translation, port, or derivative adaptation through an automated tool, even where the output is not verbatim, so it seems it could hold a candle in court.

You have a few scenarios now up for legal debate:

    1. Is training a model using copyrighted material fair use?
    2. A model outputting chunks of identical coryrighted code from it's training material
    3. A model being given in it's prompt copyrighted source code, and asked to implement the same in a new source code.

I think #1 and #2 have open legal battles on it, but #3, which seems to be this scenario, is new to me, and I feel it could have legal ground for being considered infringing on the copyright.

Sidenote: how is your account "samatman" and created "April 24, 2010", that's a really funny coincidence