A quick idea

As I understand it Markov chains are based on random. So it will produce a series that could have been. So to me. That resembles handwriting. With Markov chains you got many different outcomes. So I think they could be useful as training examples.