AI's have certain input words that break the conversation, presumably to keep you safe/give you an off switch. Google's AI doesn't seem to be aware that a users' NO will make it stop and drop everything and tells a joke in the format of a yes/no question that expects a 'no'-response if you don't know the joke already. Following, it ends the conversation with an ok because the user said no. Luckily that breaks our expectations of what should have happened in a way that makes some of us laugh anyway.