AI can’t capture meaning without narrative structure
In a recent post by Ethan Mollick, he brings up an intriguing challenge that’s bound to resonate with anyone invested in the future of storytelling: an award-winning Argentine novelist competing against GPT-4 in a creative writing contest. Spoiler alert—the human author wins, and it’s not even close. But there's a bigger takeaway here, and it’s one that aligns perfectly with what we do at Subtxt.
We need more grand challenges in AI, like when Deep Blue beat Kasparov, to better viscerally get progress. This is a good one: an award-winning Argentine novelist in a writing contest versus GPT-4. The human wins easily, but lots to learn (like GPT-4 is less creative in Spanish)
Let’s be honest: it's tempting to think that more data and better algorithms are all it takes to produce a story that hits the mark. Yet, as Mollick highlights (and the linked research confirms), even the most advanced AI language models struggle with creative writing on a truly human level—especially when the narrative needs to capture more than just words strung together in a pleasing way.
Here’s the thing—until an AI can model the underlying narrative argument of a story, it’ll always fall short. It’s not about how well it mimics a voice or generates compelling prose; it’s about whether it understands the core ideas driving the narrative. In other words, the difference between a great story and a merely clever one is meaning. And meaning doesn’t emerge from word counts or stylistic tricks—it comes from a deeply structured narrative framework that reflects the human experience.
That’s where Subtxt comes in. Our approach isn’t about chasing the next flashy feature or dumping more data into a model. It’s about helping you uncover the hidden layers of your story’s meaning—whether you’re outlining your first draft or refining your final scenes. The AI in Subtxt isn’t just playing around with language; it’s tuned into the underlying narrative arguments that make stories truly resonate.
The research Ethan mentioned illustrates that while AI can dazzle with technical feats, true creativity requires a nuanced understanding of conflict, theme, and narrative purpose. In the contest, GPT-4 fell short precisely because it couldn't grasp the complexity of a fully realized story the way an experienced author like Patricio Pron could. This gap isn’t just a matter of “more training”—it’s a fundamental difference in what drives creativity in storytelling.
If you're a writer, you've likely felt that storytelling is more than just a string of events—it's about weaving together meaningful connections that stay with your audience long after they’ve closed the book or finished the film. That's the magic Subtxt aims to amplify.
Subtxt goes beyond surface-level feedback to help you craft stories that connect on a deep, narrative level. Whether you're exploring the conflict driving your Main Character or testing how well your story balances multiple themes, our platform gives you the tools to understand what makes your story tick—and how to make it even better.
AI has a long way to go before it can challenge the world’s top storytellers, but that doesn’t mean it can’t be an invaluable creative partner in your process. With Subtxt, you’re not just getting suggestions—you’re getting insights grounded in a narrative science that’s all about elevating the meaning of your work.
So if you're ready to dig deeper into your story and bring out the best in your narrative, give Subtxt a try. The future of storytelling may involve AI, but the stories that stand the test of time will always have something more—meaning.
Happy writing!
Don't miss out on the latest in narrative theory and storytelling with artificial intelligence. Subscribe to the Narrative First newsletter below and receive a link to download the 20-page e-book, Never Trust a Hero.