Current methods for training AI on copyrighted work means “less journalism will be produced, and the cost to society will be enormous,” says the NYT.
