>> you cannot be both moral and successful in the US.

I assumed the use of massive scraped datasets, with copyrighted material and without consent, to train large AI models, had already established this.

Many people don’t think there is a moral case against training a model on copyrighted data without obtaining a license to do that specifically.