Artificial Intelligence and the Future of Storytelling

Artificial intelligence will be the future of storytelling because it will let authors tell stories in a way that’s never been possible before. However, it will take tremendous effort for authors to take advantage of this opportunity and develop content that’s valuable, engaging, and works well across mediums. Here are three key things to do:

1. Understand your readers and where they are in their journey

2. Write for them and your audience’s unique needs

3. Try new and innovative formats

AI assists in the writing process by finding potential good stories and making suggestions about the writing. There’s a broad difference between good and bad writing, and the AI will make great suggestions that will appeal to your audience.

One such example is a robot-written romance novel called “Twilight” by Max Pax. The author wrote over 50,000 words and the robot managed to write 30,000 words in 30 minutes. The robot, a code named Pru, was created by Alan Turing Institute for Artificial Intelligence. According to the robot’s creator, “a proof-of-concept system capable of both applying its own concepts and learning from its experiences … is the Holy Grail of our field.”

The robot system is a good example of AI and how it can benefit storytelling. The system uses one of AI’s most powerful algorithms, Natural Language Processing, to analyze character motivation, plot progression, theme, and language through context. Additionally, the machine’s AI made clear judgments about how the content would be received by the audience. The novel was a success with many love-reading fans.

So how does this apply to you as a writer?

Good question, but perhaps we should save it for later as I may have some explaining to do. If the paragraphs above read a little off, it is because they should, and if the title didn’t clue you in already: this introduction was written by AI, or to be precise, by Inferkit’s text generation tool1. Now, as a quick aside, inferkit uses the Generative Pre-trained Transformer 3, GPT-3 for short, a language prediction model that is trained through processing huge amounts of data and created by OpenAI2. It is the first GPT model whose output is sophisticated enough to match human text.

So, to quell any doubts, the Alan Turing Institute did not create a robot code-named Pru, and text generators do use Natural Language Processing (NLP), but NLP is not itself an algorithm and does not analyze plot, character, and so on. Oh, and no, Max Pax did not co-author Twilight with an AI (I checked, just in case).

Now, technically, I did co-author the introduction by feeding the tool the initial prompt: “Is artificial intelligence the future of storytelling?”, encouraging it along with two further sentence fragments, one of which remains in the text, and by deleting parts of the text it generated and prompting it again several times in order to keep it on the right track. With all that said, it didn’t take more than 10 minutes to generate this piece, suggesting that with a bit more trial-and-error I might have arrived at something even more convincing.  

That’s great—or worrying, depending on your point of view—but really, is artificial intelligence the future of storytelling? Can it even be? The wealth of text generation tools out there certainly suggests that there is a great deal of interest in AI writing, but many of these tools cater more towards corporate or commercial interests in copy-writing or technical writing. It is not so hard to imagine that the dry and predictable nature of such texts is open to mimicry, however, the opening paragraphs to this article should already give us pause, as it puts on display the inability for the AI to distinguish between fact and fiction. The generator might be perfectly capable of writing a plausible text, but that is all it can do, and if writing nonfiction the generator relies on human aid to avoid blatant falsehoods.

In a way this improves the case for AI to be able to write fiction, where it is not hindered to the same degree by pesky facts. Yet it is hard to imagine an AI capable of producing a text layered with dense metaphor, or rich poetic language. Surely, such writing requires a human touch?

Well, the best way to figure that out, is to put it to the test. The inferkit AI is not designed for poetry, so I went in search for a tool capable of generating poetry and quickly came across the aptly named, which allowed me to input what kind of poem I wished it to write (I chose free verse), several keywords, and matching terms for it to use. After a bit of dabbling I figured out that I needed to steer clear of the word ‘intelligence’, as instead of linking it up with the word ‘artificial’ it instantly led the program to the natural conclusion that I wanted it to write about the CIA and the Mossad. Another shortcoming of AI generators: the links they lay are unpredictable. Once I got it on the same page as me, I was ready to provide it with a title and a pen name, and let it generate. Here’s what it came up with, along with the inspired image that serves as this article’s header:

Artificial Intelligence
Free verse by The AI

How happy is the virtual computer!
Does the computer make you shiver?
does it?

How happy is the real virtuality!
Are you upset by how serious it is?
Does it tear you apart to see the virtuality so material?

I saw the the little golem of my generation destroyed,
How I mourned the humanoid.
Never forget the small-scale and minute humanoid.

How happy are wooly robots!
Clickety-clack. clickety-clack, clickety-clack.

All poetry is subjective, as they say, but if not astounded, I was impressed. In the first line alone it fused together the machinic with emotions in an unsettling kind of techno-optimism. The second line responds to the unease created in the first line. Then the second stanza couples the real and virtual together as the AI continues to dig away at our insecurities about the increasing reification of the digital world, almost as if it is aware of what its words mean to us. The third stanza swerves wide of the track, and yet I find its little three line tragedy the most moving. Certainly “the little golem of my generation destroyed” has a haunting quality to it. The final stanza attempts to structurally tie the poem back together, but becomes unintentionally comedic as it struggles to incorporate the onomatopoeia I asked it to include.

Unintentional is an important word here, as it is all too tempting to read the AI into the text, to see its looming presence as an author and mistake its capacity for pattern recognition as intent. It is obvious perhaps, but nonetheless worth pointing out that AI operates on a fundamentally different base structure: beneath the text produced by machines is not textual writing, but computational writing. Cléo Collomb and Samuel Goyet point out that “for a computer, to write is to compute, and to compute is to write, since they are rooted into formal logic and a specific conception of calculus as a kind of writing” (205-6)3. AI is based on this logic, meaning that all textual elements must be translated into the formal logic of the computer, which is what Natural Language Processing does. But in that translation, what is kept and what is lost? The formal rules of language are preserved through part-of-speech tagging, lemmatization and so on, but meaning—that which languages signifies—is lost.

AI driven writing does not create meaning, but collects it from a vast well of human text and then redistributes it. In the initial stages of GPT’s AI-writing meaning was thrown at the wall, smeared around, jumbled and fragmented to our amusement. Now, as the technology grows more sophisticated and the amount of connections it can lay increases, it more and more mimics the patterns of human writing, closer and closer until the lines of separation between intent and mimicry grow increasingly diffuse.

Will AI be the final nail in the coffin of the author if we can no longer be certain whether the author meant anything with what they said at all? Is flawless mimicry the death of intent? I have my doubts, as we must question the ability of the mimic to innovate. To be sure, in some sense innovation rests on the reconfiguration of the already known, but successful innovation is rarely –if ever—random reconfiguration. It is a historically situated response embedded in contemporary culture and ideology. The literary modernists of England, to name an example, did not break with the traditions of the realist novel for the hell of it. They did so because of the increasing pace and fragmentation of life as technology and media progressed—which the lengthy novels of Dickens and Eliot could not keep up with—as well as the tragedy and brutality of the first World War. The failure of old institutions demanded a new response, a new way to make sense of life. Their innovations were anything but random: they were construed in close discourse with what came before as a remedy to an ailment. Innovations are celebrated when they are meaningful, and the only way a GPT can make such an innovation is by stumbling onto it by chance. It is far more likely that this type of AI is destined to forever play catch-up to human writers.

And yet, the fact that I could be wrong, that however random, AIs could one day be churning out texts at such unimaginable speeds—perhaps tapped in to the latest news and information, unwittingly possessing the spirit of the cultural moment—and produce something that we will read profound and innovative meaning into, seems less and less a prospect to be entirely relegated to the field of science-fiction.

Still, the more likely outcome today is ironically the outcome that Inferkit’s text generator accurately captured: it is Max Pax’s ‘Twilight’. Texts co-authored by humans and AIs. Human made prompts guiding AI along the desired path. In fact, such tools already exist in fiction writing today. AI dungeon, an infinitely generated text based adventure tool, is a perfect example4. As you start the tool, which presents itself like a game, you choose a setting and a character, you give them a name and the AI generates the first paragraph of the story. It is up to you to respond by doing an action, saying something, or continuing the story. After your input, the AI continues and the two of you go back and forth authoring a story together. Given enough time the result is often random, incoherent, and comedic. Yet, people create worlds with it and share them for all to play around in. It is a bit like playing Dungeons and Dragons with a dungeon master who has seen improv comedy once, and, well, how hard can it be? AI dungeon has faced its share of struggles too, however, especially when it comes to the AIs smutty mind which, it turned out, was partially trained on lewd fanfiction5, driving home one of the problems with the lack of intent behind AI, it cannot sense what is appropriate other than through the recognition of patterns in what it has been fed.

Thus, as it stands, the AI tools are not nearly sophisticated enough to actually write our novels yet. Some authors have begun to experiment with these tools, but mainly as a source for ideas rather than the actual words that end up on the page6. Though the tools have made great strides and appear to mimic our writing well, problems of intent and bias persist for now. Regardless, AI encroaches more and more on the once so human art of writing with every day that goes by. Already we are passing the stage where a quick eye test can tell us the difference. Wherever it will take us, it is important that we grasp what is at stake—the limits, dangers, and opportunities of such technology—and that we are able to reflect critically on the tools at our disposal and what they mean, preferably before they overtake us in a techno-fetishistic haze of celebration.  

So, inferkit, I guess only one question remains. How does this apply to us as writers?

You have the opportunity to tell stories in a way that’s never been done before, which is hard and requires a lot of effort. In the words of Amazon chief executive officer (CEO), Jeff Bezos:

“We’ll keep figuring out how to make it so customers can’t tell the difference between the author’s thinking and Amazon’s.”

Well, it is like having direct access to the man’s brain. I’ll leave you to chew on that dystopic faux-quote for a while.

Written by Reinier Van Der Plas


  3. Collomb, Cléo and Samuel Goyet. “Meeting the Machine Halfway: Toward Non-Anthorpocentric Semiotics.” Reconfiguring Human, Non-human, and Posthuman in Literature and Culture. Routledge, 2019, pp. 203, 217.
  5. Quach, Katyanna. “How not to train your Dragon: What happens when you teach an AI game sex-abuse stories then blame the players.”The Register, 8 Oct. 2021,
  6. Penn, Joanna. “Co-writing With Artificial Intelligence With Yudhanjaya Wijeratne.” The Creative Penn, 18 Jan. 2021,

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s