[dwoldrich]

> I'm referring to the model itself. The `.ckpt` file is clearly transformative wrt its training set. Or, at least, substantially more transformative than other things that have long received fair use protection.

Oh, I see. And the model weights are what one can make the copyright infringement claims on in the US?

Not to split hairs, but do you believe it's so transformative because you can't read plain text copies of original works in the weights or because the source material is so hopelessly discombobulated that the original work could not be reliably recreated?

I believe the 'hopelessly discombobulated' argument is probably pretty solid, but one could argue to a judge that the weights are something like JPEG compression. Sure the forged image of Mona Lisa is a bit foggy in the background and some of those details are incorrect, but the wry smile in the foreground is perfectly captured.

> On the contrary, I'm invested quite heavily in the exactly opposite hypothesis -- that the ChatGPT/Claude/Gemini UX you're referring to is not fit-for-purpose.

Oh! Excellent, carry on!

> rather than an anthropomorphic entity.

But it unfailingly passes the Turing test, at least with regards to an immature, non-discerning human mind like a child's. You may as well rub a lamp.