Considering it used training data of human generated text I would assume so.