News for hackers. Stuff that matters.
Re:Should GitHub be sued for training Copilot on GPL code? by UncleMeat on Thursday June 23, 2022 @09:45AM (#31849045)
All of the large language models can emit text that has never been seen in the training set (unless you go so far to consider each character to be a snippet).
Re:Should GitHub be sued for training Copilot on GPL code? by BeefWellington on Thursday June 23, 2022 @11:24AM (#31850422)
They can also emit text that it verbatim copied.

Infringement isn't about how the infringing system works, it's about the product of that work.

Re:Should GitHub be sued for training Copilot on GPL code? by jhugo on Thursday June 23, 2022 @12:24PM (#31851365)
> Infringement isn't about how the infringing system works, it's about the product of that work.

Exactly this. It makes zero difference that you produced your infringing work with the help of a program that happens to be extremely complex and marketed as "AI".

Re:Should GitHub be sued for training Copilot on GPL code? by thesz on Friday June 24, 2022 @04:05AM (#31858934)
So do smaller models and I have to note smaller models are better at that.