Do you have a reason to believe this ain't already being done? I would assume that the big guys like openai are already training on basically all text in existence.