If the models were trained only on data that was out of copyright it wouldnt have been an issue, but nobody want to train on only 100+ years old data. Copyright laws are too long when their content arent culturally useful by the time they are free for public use. It hinders the creation of useful tools, among them these generative AIs. Maybe time for some reduction of those laws to something useful, and at the same time increased strictness for businesses to misuse copyrighted content?
If the models were trained only on data that was out of copyright it wouldnt have been an issue, but nobody want to train on only 100+ years old data. Copyright laws are too long when their content arent culturally useful by the time they are free for public use. It hinders the creation of useful tools, among them these generative AIs. Maybe time for some reduction of those laws to something useful, and at the same time increased strictness for businesses to misuse copyrighted content?