I have been thinking about applying GP3 to manuscript scoring. Better yet, line up manuscripts with similar books already published. Then publishers or agents could get a constant stream of submissions that they are likely to accept. The writing quality and story could be scored. Can an AI determine quality? I think so.
I've noticed that the AI generated texts in novels tend to lack narrative cohesion over more than a few paragraphs and that the people who use them often have a human write the beginning and the ending of the book and then the middle is just AI hash that looks perfectly okay if you only read a few pages, but if you actually read the whole book, you realize that the central portion is completely fragmented. If an AI read such a text and concluded that it was great based on stylistic elements and if an agent didn't actually read the book all of the way through, that tool could completely sink the publishing industry because the platforms woudl be saturated by books that are 90% unreadable by humans. You would end up engineering an illiterate population! Kids would pick up those books and not be able to follow the plot in the middle (because there isn't one) and then they would conclude that they don't like to read.