[cozzyd]

Copyright law is far from perfect, but the concept is not morally bankrupt. It is certainly abused by large entities but it also, in principle, protects small content creators from exploitation as well. In addition to journalists, writers, musicians, and proprietary software vendors, this also includes things like copyleft software being used in unintended ways. When I write copyleft software, it is my intention that it is not used in proprietary software, even if laundered through some linear algebra.

I'm also far more amenable to dismissing copyright laws when there is no profit involved on the part of the violator. Copying a song from a friend's computer is whatever, but selling that song to others certainly feels a lot more wrong. It's not just that OpenAI is violating copyright, they are also making money off of it.

[CaptainFever]

With the exception of source code availability, copyleft is mostly about using copyright to destroy itself. Without copyright (which I feel is unethical), and with additional laws to enforce open sourcing all binaries, copyleft need not exist.

So it is not good when people use copyleft as a justification for copyright, given that its whole purpose was to destroy it.

[cozzyd]

Source code availability (and the ability to modify the code on a device) is the most important part, IMO , regardless of RMS's original intention. Do you feel that it's ethical that OpenAI is keeping their model closed?

[CaptainFever]

No, because I think such restrictions are unethical in the first place. However, in regards to training, I think it might be a necessary evil to allow companies to ignore copyleft, so smaller entities can ignore copyright to train open models.