I sent the manuscript to various people to solicit feedback. I did not tell any of them that this was computer-generated — just that it was my first attempt at fiction and that I had decided to play safe by writing about subject matter that I was comfortable with in a short story format that did not require me to extend myself too much. For a dash of verisimilitude, I explained that my initial motivation was to generate a series of case studies that I could use to present concepts of AI risk at many of the workshops I am asked to conduct (which was not entirely untrue, even if the actual project quickly became far more ambitious than that initial concept).
Almost everyone I sent it to knows me well enough to suspect that AI had been used in some way, shape or form. Some were tipped of by the richness of geographic and cultural detail that was far more than I could have ever accumulated through personal travel. Others suspected something was up, given that the register in which the stories were written was nothing like that which comes across in my non-fiction writing. One of them, clearly a regular user of AI himself, figured out that I had used AI heavily and was tying himself up in knots trying to figure out a polite way to tell me (in an ironic twist that was not lost on me, he turned to AI to help him find the right words to say)
No one figured out that every single word of the book had been generated by AI. Either they simply did not think it was possible and as a result, rejected the possibility outright when it crept into their minds, or they gave me more credit than I deserved.
I am the first to admit that this is not a brilliant book. While some stories of the stories are interesting, tugging at your heartstrings, and presenting a turn of phrase or stylistic choice of sentence construction that truly gives pause, many of the stories are pedestrian, stretched out longer than they should have been, telling when they should have allowed the reader's imagination to do the heavy lifting.
But I had not set out to write a brilliant book. All I wanted to demonstrate was that it was possible to get AI to generate long-form stories without any direct human intervention in the text. I was far more interested in the architecture I had to build to enable this and what it would take to implement at scale. As a result, I spent more time on developing the harness and formulating the process that created the story than working to ensure that the actual text itself was what a human would write.