All the data used to train the AI models was arguably stolen. Is it so surprising that the models may just be doing very good search over the entire present of human thinking?