For my latest short story about surgery in 2050, I experimented with GPT-4’s storytelling abilities. First, I wrote the entire story myself and passed it through my editors. Second, I gave GPT-4 my finished story and asked it to rewrite it to be “more compelling”. Then, we passed that through myself and my editors. Finally, we compiled both versions into the published book.
What did I learn?
1. GPT-4 makes the same errors in spelling, grammar, punctuation, and storytelling that I make on early drafts. It needs human editors to clean-up its work.
2. GPT-4 made the story “more compelling” by using more adjectives and making every chapter “thrilling-all-the-time”. The story doesn’t have highs and lows, it tries to stay high all the time.
3. GPT-4 tends to remove dialogue, especially long exchanges. It deleted two pages of dialogue that were important to the story. I looped back with it to force it to keep and rewrite those two pages.
4. GPT-4 tells the story in fewer words. My original is 15,000 words. GPT-4’s retelling is 10,000 words.
Assessment: The AI is really pretty good. Its work has the same flaws as a human writer. It can’t rewrite a publishable story without the same editing that human authors require. But I learned some tips for improving my own writing from reading its version of my story.
You can download the stories for free in EPUB or PDF format from my BookFunnel site: https://dl.bookfunnel.com/mb03aayv3f
0 Comments