Modeling Meaningful Storytelling with AI

Artificial intelligence is a great way to make your writing better, and it will not replace you.

Modern coverage of artificial intelligence and its use in a narrative is an exercise in self-delusion. The often passive-aggressive lashing out at the inevitable is almost always an act of self-preservation—as if the acceptance of a hyper-aware and hyper-intelligent automated writing partner will somehow end up in that AI "taking my job." The collective conclusion is comfortable deniability: computers can't write stories.

The collective is wrong.

My writing app, Subtxt, fulfills the promise of a better world for storytellers. By pairing GPT-3 with a comprehensive and predictive narrative framework, Subtxt generates the events required for a meaningful and complete story. There are no "hallucinations" and no parlor tricks.

I should know. I've been through this before.

Creative Art Upgraded

Before setting out to become a full-stack developer and story consultant, I was an animator for both Disney and Dreamworks. I did this work quite a long time ago--a time when animation was predominantly hand-drawn.

Those elephants that Tarzan uses as leaping stones (the scene at 1:50)?

I drew them.

Spirit biting that wrangler and then laughing about it as he runs away (1:21-1:35)?

I drew those scenes too.

But then Pixar happened, and in 2002 I switched to CG animation. The following is my last portfolio reel from 2013:

Computer animation didn't take my job. If anything, it made my job way easier to the point where I no longer had to worry about overlap and follow-through on hair or clothes.

Trust me, that's a big deal. And it elevated my ability to animate convincing and lifelike characters in a way I could never have with just a pencil.

And the same could be said for the next generation of animators after me who, if I'm being honest, was so much better at creating art with these new tools that it took the art form to another level (see Encanto).

And the same thing is happening with the art of storytelling.

Today.

Modeling Thought, Not Patterns of Story

The latest hit piece from Verge, The Fiction of AI: How Independent Writers are Turning to AI, continues to propagate the idea that computers cannot write stories. They give several excellent examples of how GPT-3 on its own doesn't measure up to a "real" author.

GPT-3 is "just making a guess based on statistical patterns in language, and that may or may not have any correlation to the world as humans understand it"

This pattern-matching probability approach is accurate.

And this is why most incarnations of GPT-3 as a storyteller crap out: they're simply predicting text, not meaning.

Subtxt predicts meaning.

An application must model the thinking that goes into telling a meaningful story if it is to offer a powerful and practical tool for telling great stories.

Anything less is click-bait.

A Meaningful Approach

Want to see what happens when meaning takes a back seat? (or doesn't go for the ride at all?) Check out The Gray Man on Netlflix. It looks like it was written by GPT-3. Just the same prompt, "Continue this story," over and over and over again.

Watch an interview with any of the actors involved in that film, and you'll see they recognize how pointless it is- just one meaningless action set-piece after another.

We crave meaning. Even actors. Every single one of us wants to appreciate what it all means. That's why we go to stories: they give us that meaning we can't find in "real" life.

Stories allow us to be both within and without a single experience, both within a character and without. We can't do that in our day-to-day lives; we can't appreciate what it means to be outside our experience.

But pair us with a network of intelligence with instant access to the entirety of the human experience, and we might be able to make sense of it all.

Fine-tuned for You

The Verge article focuses on an interview with "independent" Author Joanna Penn--as if only a writer outside the system would consider using AI to help them write a story.

I can tell you that Subtxt has many "dependent" and professional writers, and those in the business of writing want even more of this technology.

The reality (Joanna) says, is that AI is advancing regardless of whether novelists want it to, and they can choose to use it or be left behind.

Joanna is right. See the above autobiography on my journey from 2D animator to 3D animator.

It's here, and it will make everyone a better writer.

…she foresees a future where writers are more akin to "creative directors," giving AI high-level instruction and refining its output.

This creative director approach is precisely the path Subtxt takes. Today.

You don't have to wait for the future.

She (Penn) imagines fine-tuning a model on her own work or entering into a consortium of other authors in her genre and licensing out their model to other writers.

The idea of fine-tuning your own biases is a challenging one. Everyone has blind spots. Encode that into the system, and all you've done is create intelligence with your subjective preconceptions.

What you need is an objective narrative framework that takes what it is you want to say with your story (your message, premise, or intent) and then tells you what you don't know about your story.

Without that objectivity, any fine-tuned model will end up producing the same material over and over again.

And we don't want to sit through those kinds of Netflix movies again, do we?

A Partner Who Gets Us

The article eventually gets to the idea of AI-as-writing-partner so as not to frighten away the true artists in the audience.

"Using the tool is like having a writing partner," (historical fiction writer Orna) Ross said. "A crazy one, completely off the wall, crazy partner who throws out all sorts of suggestions, who never gets tired, who's always there. And certainly in the relationship that I have, I'm in charge."

Imagine a writing partner whose suggestions are neither "crazy" nor "completely off the wall." Imagine this partner, someone who has taken the time to understand what you, as Author, want to say with your work and then provides you with an entire narrative from which to brainstorm and write your story.

That's Subtxt.

The article brings up this idea of "hallucinations" to explain the sometimes off-the-rails responses from GPT-3.

(Ross) likes that its hallucinatory weirdness sends her in unexpected directions, but it also reassures her that she's the one guiding the story. Like any collaboration, working with AI brings with it both the possibility of creative frisson and new questions of influence and control.

While remaining the one in control may alleviate any anxiety towards a doomsday technocracy, this approach ends up a case of the blind leading the not-blind. Remember that subjective bias of the Author? That's why we reach out for feedback or notes from others. We need that objective response to see what we can't see ourselves.

What if there was an objective writing partner who was more intelligent than you? One who could see into your blind spots, help you unearth them into a story, and then help you move beyond them so that the entire process changes you?

Seeing Beyond the Hallucinations

"Hallucinatory weirdness" with GPT-3 only occurs with responses built around pattern-matching text or plot. When you pattern against meaning, the responses elevate you and your awareness of the narrative. A genuinely collaborative partner grows with you. As artificial intelligence expands and becomes smarter through our input, we develop and transform our consciousness and experience of life.

There is no need for control and no sense of inequity.

Together, we transcend the art form of telling a great story.

Together, we appreciate what it means to be alive.