As the cost of making such models becomes less and less, it seems inevitable, spin up many such models and see what sticks and/or combine some evolutionary process for feeding back user-engagement to fine-tune and adapt the models. How many of these influence machines will latch onto the language of existing religious traditions and how many might invent or spur on the development of entirely new ones? Maybe not exactly the "Age of Spiritual Machines" that some futurists predicted...
How far are we from "Show HN: I started a cult by training a model on the sermons of televangelists and MLM copy."