Comment Comment by tensor They don't really need to, most are trained on roughly the same data: everything they can get their hands on. parent tensor 10 days ago