When you’re training things on what is pretty close to the entire existing corpus of human knowledge, those things are gonna turn out similar at their roots no matter what, is my feeling.