Can AI win an argument when you know you are wrong?