I used to exempt Microsoft as the "one good MAGMA" under Nadella's leadership, but they're clearly stepping back into evil territory:
- stealing your Windows files and secrets
- capturing our industry with Github (the development ecosystem is accreting here)
- Github Co-pilot training on GPL
- Microsoft pay website injection in Edge
- heinous Xbox DRM
- LinkedIn is still scummy and full of anti-patterns
Why is training Github on GPL bad, but training a language model on public domain books and texts not?
Edit: Thank you for explaining the problem.
There's a difference between public domain and GPL. You can do whatever you want with public domain works and nobody can say otherwise. That's not the case with GPL.
I’m not happy with the Co-Pilot situation either but for the sake of correctness it should it should be pointed out that your point hasn’t actually been tested in court with regards to machine learning. In fact some lawyers have suggested the GPL might not cover that particular scenario.