Once thought of as a uniquely human trait, creativity has now become a debated topic. And that’s why we’re asking questions such as:
Can machines learn creativity?
Can they be considered novelists, poets, or artists in their own right?
If there was a one-word answer to these two questions, I’d say ‘Yes.’ AI has created a new landscape of creative opportunities. And as a novelist, I feel threatened and intrigued by the possibility of machines putting authors out of business.
AI already has a profound impact on our everyday lives — Google search predictions, email spam filtering, voice recognition, and weather forecasts are all the examples. We are heading towards the future at an immeasurable speed, and AI is a spear that’s tearing through barriers.
This technology has paved its way to literature. Machines have evolved from ‘tools that help human creators’ to ‘creative entities’ themselves. Writers might scoff at the idea of AI-generated stories and texts, but fiction-savvy AI is already in the market.
Take a look:
The Rise of Robot (AI) Writers
The very premise of Artificial Intelligence and Machine Learning is its ability to learn from the data collected. The more data there is to analyze, the better the machine performs.
Artificial Intelligence & Machine Learning models, trained on millions of words, give computers the ability to write human-like texts. These ‘robot-writers’ are prepared to learn the fundamental structures of a language by following a statistical pattern.
Take, for example, GPT-2, one of the most recent ‘AI-writers’ that has generated synthetic text samples and adapts to the different writing styles. It performs tasks such as machine translation, rudimentary reading comprehension, summarization, and question answering — all without task-specific training.
To put it differently, despite the influx of highly-creative novelists, ‘robot-writers’ have managed to make their presence felt, in the same way as any amateur (but promising) human author. So, let’s take a step back and look at recent AI breakthroughs in digital content, academic, and novel writing:
AI & ML written content has started to expand beyond the formulaic writing. And that explains why notable media organizations like Associated Press Reuters, Yahoo!, Washington Post, and The New York Times are implementing AI solutions to produce hundreds of more stories in a sliver of time. It would take humans to write each one individually.
AI-Generated Digital Content & Stories
AI-Writer for Academic Books
Lithium-Ion Batteries: A Machine-Generated Summary of Current Research is the first AI book published by Springer Nature. Written by an algorithm ‘Beta Writer,’ this book provides an overview of current research and study in the field of Lithium-ion batteries.
Henning Schoenenberger, Metadata Management at Springer Nature, says: “This prototype is the first important milestone we reached. It will hopefully also initiate a public debate on the opportunities, implications, challenges, and potential risks of machine-generated content in scholarly publishing.”
‘1 the Road’ is an experimental novel generated by Artificial Intelligence.
In March 2017, Ross Goodwin, a gonzo data scientist, and creative technologist, traveled from New York to New Orleans. He went in a Cadillac car with a GPS unit, a surveillance camera, a clock, and a microphone — all connected to a portable AI writing machine that received real-time input from the surroundings to capture seed words or phrases.
As Goodwin drove, the AI machine continued weaving a fictional plot and printing the phrases on long scrolls of paper. By the time the Cadillac completed it’s four day’s journey, these scrolls filled the back seat of the car, and AI’s first novel was accomplished.
The AI that traveled with Goodwin was an LSTM (long short-term memory) neural net — known for performing tasks such as speech recognition and handwriting. Goodwin (previously) fed LSTM sixty million words of raw literature to act like a human writer — one-third science fiction, one-third poetry, and one-third ‘bleak’ novel.
And these emerging technologies surely makes one question:
Can AI Measure up to Jane Austen?
AI has the potential to enrich our literature but, be that as it may, it is still somehow off writing a coherent novel. Developers are training machines to think and program like a human mind, but now, this is more of a probabilistic approach.
AI-powered tools like Beta-Writer, GPT2, or LSTM are just using different methods of statistical analysis to predict which phrase to add next. The most compelling evidence of AI’s inefficiency in novel writing is a predictive keyboard experiment.
Take, for example, what GPT-2 wrote when fed the opening line of Jane Austen’s ‘Pride and Prejudice.’
The narration is puzzling for many reasons. The narrative sounds rather like a political bloviator. The second ‘that’ is superfluous, and the text, overall, has poor readability.
As of now, novelists see these machines as a smart plagiarism tool that mashes together bits of phrases, stories, and news written initially by humans and creates a praiseworthy spin-off text. Goodwin, too, accepts that “1 the Road is very much an imperfect document, a rapid prototyping project. The output isn’t perfect. I don’t think it’s a human novel, or anywhere near it.”
When Janelle Shane, a programmer, read GPT-2 generated texts, here’s what she tweeted, “Some GPT-2 sentences were so well-crafted I wondered if they were plagiarized, plucked straight from the training dataset. But for most parts, the computer journeyed into a realm of dull repetition or “uncomprehending surrealism.”
Writing is Not Data
Provided that, the question of the hour is, ‘how worried authors ought to be about ‘AI’s fiction-writing potential’?
For the most part, it’s too early to add ‘fiction writing’ to the long list of jobs that AI & ML will eliminate. The excess hype surrounding beta writers or GPT2 text generators is because of AI’s capability to work as a huge-data crunching system. Text can be mathematically manipulated and encoded by a machine.
But fiction writing is not data. It’s a composition of imagination and expression, which means that you’ve to have something to imagine & express. A non-sentient machine or program has no imagination or experience to share with the world.
To put it differently, I’d like you to think of other creatures on this earth. Let’s say, a crocodile. If a crocodile could express their emotions, would we relate to him? I think not. Being a reptile is presumably different from being homo sapiens, and hence, there’s no scope of mutual comprehension.
Of course, it’s easy for AI & ML technology to self-improve, but it lacks intuition. There’s emotion, instinct, real-world experience that can’t be replicated via coding or algorithms. Training a machine on a wide variety of formulas can enable it to reshuffle components and write ‘seemingly-original texts’ but we’re decades away from using robot writers as a replacement for authorial work.
The Future of Fiction Writing
Perhaps in the future, AI will begin to accomplish human-like imagination and creativity with its lightning-fast processing powers and comprehensive knowledge. Such leaps can open up new creative avenues for us to explore and introduce new forms of literature and art.
But right now, AI can only act as a creative prosthesis, a cognitive enhancement tool for humans to explore the highest level of creativity. So, the best way forward for AI & humans is to lean on each other’s strengths and live harmoniously.
Instead of thinking of AI & ML as tools to replace novelists, we should examine ways that AI can be used to augment human creativity. It’s time to stop worrying about AI’s fiction writing abilities and evaluate how machines and humans can work together for creative collaborations that have never been thought of before.
As an illustration, if AI can fact-check narratives, go over details, untangle plot elements, and point out contradictions in description and minutiae, authors will be able to focus more on what matters in fiction writing: the emotions, the plot, the themes, and the people.
Here’s an excerpt from Melvin Kranzberg’s book that best summarizes this idea —
Replace the violinist & his violin as ‘Authors & AI-enabled writing tools’, and you’ll get the context.