I didn't indicate which way I care for the ruling to go. I think you're giving current AI models much more credit than they deserve. Additionally, it's hard to

I didn't indicate which way I care for the ruling to go. I think you're giving current AI models much more credit than they deserve. Additionally, it's hard to have enough computational power to run these models even if you have top of the line consumer hardware, and that's for generating a single frame. NERFs are interesting from a film-making and art point of view. AI isn't going anywhere. Any sort of ruling would have zero impact on this kind of AI. Tech will move forward regardless of the ruling. The issue is about datasets, not the software itself. I expect all of the suits against the underlying technology to be dismissed. If the lawsuit is successful in blocking AI, it would have a lot of negative implications for fair use and be a gross mischaracterization of the underlying math. I think we need a clear ruling on how copyright should be handled from things generated using AI. Once you modify it in some way, you clear the bar for making a transformative work, but the difficult questions is how and if the copyright of the input dataset applies to raw generated media. No one is having the same discussion regarding the input and output of ChatGPT. Even the discussion against CoPilot is weak. But a ruling in a case will affect every other one. It's not really possible for laws to differentiate between text and photos, because photos are stored digitally as text. During my research, my PI explained to me that copyright work is fine to test with and make datasets with, but I can't include benchmarks against it or for it in my final paper. But the weights generated from such work is fine. This is only really a problem because these models are being used in commercial ways. https://theaisummer.com/nerf/ https://theaisummer.com/nerf/