Perhaps I should have said idealistic. I mean I've seen several people make demands like they're somehow already in a courtroom with Microsoft. No mention made of the fact that this is likely covered under fair use. If it isn't, then basically every deep neural net trained on a web dataset is in violation. This means open works like from University at Heidelberg (Latent Diffusion, VQGAN) or EleutherAI (GPT-J) would be impossible.

Like, maybe you could all get together and learn how machine learning works and train your own clean model? That feels far more positive and in the spirit of progress to me. I guess from my point of view it's clear - no laws will save you from the next wave of large language models. The weights are trivially distributed meaning once it's trained it's more just a fact of life we all have to deal with. So making demands when you have basically zero leverage rather than admitting defeat and working within the new constraints of progress.

I just feel like it's an inherently philosophical position and people are acting like code theft hasn't been common practice for the history of all software.