zlacker

>>ssgodd+(OP)
I'd bet they win, but how do you possibly measure the dollar amount? If you strip out 100% of NYT content from GPT-4, I don't think you'd notice a difference. But if you go domain by domain and continue stripping training data, the model will eventually get worse.

>>mritch+W2
Take estimated losses of the NYT from this "innovation" and multiply by 10^x where is "x" high enough to make tech companies stop and think before they break laws next time. That would be my approach at least.

>>kfk+Z3
which laws are broken exactly? it's not remotely settled law that "training an NN = copyright infringement"

>>necrof+N4
The training isn't the issue per se, it's the regurgitation of verbatim text (or close enough to be immediately identifiable) within a for-profit product. Worse still that the regurgitation is done without attribution.