ChatGPT and other AI writing apps like Jasper, Sudowrite, and Claude have been used to write longer-form pieces for over a year now, but developer Liam Swayne took the concept one step further by asking AI to complete the Game of Thrones book series (A Song of Ice and Fire) that George R.R. Martin has notoriously been slow on finishing.
Posted on GitHub and completely free for anyone to read through, Swayne’s ChatGPT version of both The Winds of Winter and A Dream of Spring are expectedly long, but surprisingly good.
So how did Swayne get ChatGPT to stay on track for that long and write something coherent? In an interview with IGN, Swayne explained that he worked through a series of prompts that started with an outline for each chapter until all 45 chapters of The Winds of Winter were done. He then fed each outline back through ChatGPT and asked it to flesh out the basics with more and more details. Eventually, the bullet-point list of ideas became full-fledged stories that grew from scenes to an actual flowing plot.
“Meera, her breath visible in the frigid air, glanced at Bran with concern etched upon her face. “Bran, we have traveled far and faced countless perils, but the path ahead grows more treacherous with each step. Are you certain that we are on the right track?”
Bran paused, his gaze fixed on the distant horizon. The fading light of day cast a warm orange glow upon his face, contrasting with the icy world around them. “I can feel it, Meera,” he replied, his voice tinged with a mix of certainty and wonder. “The magic that lies beyond these mountains is calling to me. We must press on.” — Excerpt from A Dream of Spring by ChatGPT.
On Swayne’s GitHub page, he details other plug-ins and Chrome extensions he used, including the ChatGPT to Markdown extension that exported chats and fed them back in for the outline expansion and the free trial of Keyboard Maestro to automate the feeding of thousands of prompts back through ChatGPT.
“Large language models can be very scary, but this project makes me more optimistic about the future of writers and AI,” Swayne told IGN. “This project demonstrates that large language models like ChatGPT can take hundreds of pages of text into account when making a narrative decision, which could help writers quickly fix plot holes. It also shows that AI can only do what has been done many times before. This project has given me confidence that AI won’t be replacing unique literary works any time soon. I believe the AI had trouble writing character deaths specifically because most writers (and consequently, most of the training data) are hesitant to kill off major characters. This is part of what separates George R. R. Martin from other writers: his stories make unconventional, surprising decisions. At this point, AI can only do what is most commonly done, which means it struggles to create stories that aren’t by the books. To my surprise, I’m more confident than before starting this project that writers making creative and unexpected decisions are not replaceable.”
In total, Swayne created over 1.2M words in these two books, which is closer to reaching Brandon Sanderson territory, but it’s clear that it was no easy task. While Swayne didn’t simply ask ChatGPT to finish the Song of Ice and Fire series with one prompt, the process was still obviously much easier than actually sitting down and writing the novels, as Mr. Martin can attest.
But as AI gets more and more advanced (OpenAI just announced character sheets to better keep track of writing characters in novels), we can’t help but wonder where we will go from here. It’s obvious that AI will be able to write novels that are indistinguishable from human writing soon, but will the stigma remain?
Want to chat about all things post-apocalyptic? Join our Discord server here. You can also follow us by email here, on Facebook, or Twitter.
