Cana Fuest, PhD Candidate

I conducted an experiment in my writing last week. I gave Gemini a chapter of my manuscript and asked it to write the next chapter in my voice. I gave it no clues as to the story plot but I did tell it to pace the story as if it were a full-length novel. The first attempt produced a decent chapter, although it was very different than the second chapter I had produced on my own. I also had to instruct Gemini to steer clear of cliches such as “sent a shiver down my spine.”

Next, I gave Gemini the second chapter that I wrote and asked it to write a third chapter based on my story. It changed from first to third person, so I had to tell it to correct that. And something else sort of creepy happened: Gemini inserted an element into chapter three that I had in my original manuscript. This freaked me out a little at first, until I remembered that Gemini has learned from humans and perhaps my story was also a little cliche! In other words, what I had produced had been done before in other stories of the same genre and Gemini was just generating that information to align the task with the prompt.

I next switched to ChatGPT 3 to see which machine produced a better story. I felt that ChatGPT didn’t quite capture my voice the way Gemini did but ChatGPT also inserted a different story element into its chapter that I also have in my original manucript. Creepy….or just cliche? I don’t think the machines are creepy, really. They have learned what they know from us, so anything produced by them speaks to what they know about storytelling.

My conclusion is that neither of the machines wrote quite the way I would have but they are useful for eliminating story cliches!