The lure was great. The call of void was in my ears. I felt, I was standing on the perilous edge of the cliff, my heart pounding from the chase. I was still not safe. The hounds of the deadline were not far behind. My hair, tangled by the chilly breeze, was in my eyes. I could not see waves crashing relentlessly against the jagged crevices weathered by time; their salty spray making my face sticky. A whirlwind of temptation was tearing my mind, and each thought echoing louder than the cries of a seagull. Should I upload one or two samples of my writing and ask Bard to create an article? I had to run to the early office after a late night and today seems no different, with the departmental summit being planned. On top of it, my car was low on reserve fuel, so a detour to the filling station would take another twenty minutes. I had promised to send my article on the topic of “AI in the world of Literature.” All logic said, it was not possible. I opened my eyes. My cat was looking sternly at me, as if reading my mind. Embarrassed, I decided to be late in my submission but not take a shortcut.

Rie Kudan, the winner of Japan’s prestigious literary award, acknowledged that she had penned about ‘five percent’ of her novel, ‘Tokyo-to Dojo-to’ (Sympathy Tower Tokyo), with the help of ChatGPT, opening debates on the use of AI for creative pursuit. But she is not the first one.

It began in the 1960s with a story generator developed by linguist Joseph E. Grimes. It was the first to take a grammar-based approach and the first to operationalize Propp’s model. The Policeman’s Beard was next, being the first algorithmically authored a book for mass readership. As computers enhanced and AI came into play, the abilities of the computer or AI increased from checking the phrases to generating novels. Hitoshi Matsubara, a computer scientist at Future University Hakodate and leader of the team that created the novelist AI. The novella, “The Day a Computer Writes a Novel,” one of 11 AI-authored submissions to the third-annual Hoshi Shinichi Literary Award made it past first round. The judges decided not to cut it against its human competition. One of the judges, Satoshi Hase, said at a press conference, “I was surprised at the work because it was a well-structured novel.”    

We have opened this Pandora’s box, and when we opened it, the tiny creatures escaped and smashed the little bottle holding the genie inside. There is no way can we put the genie back into the bottle and get the escaped creatures back into the box. Let us accept it AI is going to be the way of life. There is no doubt about it. The question is how ethical it is to use it in creative pursuit. Why does a person take a perilous path of creative pursuit? Creative pursuits give a safe channel for release of emotions without fear of judgement, providing a medium to delve deeper into one-self, relax and reduce stress. Few of people venture into critical pursuit to enhance critical thinking and problem solving, communication and empathy or a means to connect with community. Whatever the reason might be, there is an underlying need to express themselves; and let their voice being heard.

If that is so, then why are few authors using AI to write for them? Why is AI so seductive? As a data scientist, I am tempted to ask, why not? AI today has made people with no coding skills programmers with low-no code. Then why would the people who want a book in their name but lack skill or vision be far behind? The software provides prompts on atmosphere, setting, character developments, conflict, emotional resonance, introspective, plot twist, suspense, symbolism, and tension. A person could write a sentence “I am sad” and demand suspense, symbolism, and tension. The AI would write for him/her. Easy! Why not, from a simple vanilla sentence, “I am sad,” get elevated to a beautiful paragraph, perhaps of the level of Haruki Murakami. It is a shortcut that could very well be the ticket to success for some. Before using AI for creative pursuit, stop for a moment and understand how it works. Extensive use of Natural Language Processing (NLP) techniques like tokenization, part-of-speech tagging, and sentiment analysis is used to understand the meaning and structure of language and using generative algorithms like recurrent neural networks (RNNs), the AI uses machine supervised and unsupervised and reinforcement learning machine learning to go scour through the training dataset or world wide web to understand the style of writing, the language, vocabulary and mimics the communication of human.    

Using AI to write, the user needs to enter their inputs like prompts, characters, and description, etc. Most of these AI are trained with web data, human dialogues, and instruction. The AI will search for matching description from millions of the books with similar description. AI predicts the next word in a sequence based on the previous words (like how autocomplete works). The AI uses pre-defined templates for distinct parts of the story, like character descriptions or dialogue, and fills them in with unique content. Then AI analyses the existing story elements and generates new content that follows the established logic and style. They will replicate picking elements from them. The sentence may seem new, but they repackaged contents from millions of books, articles, code, social media conversations, and patterns that one may not have heard or read before. Consider this Bing, Bard, and ChatGPT work with 1.76 trillion, 540 billion and 170 trillion parameters. If there is some collaboration of the user to review the consistency and coherence with concept, characters, and plot outline, the repackaged literary work would look anew. Does it sound wonderful? One can give guidelines and AI would generate masterpiece novels. Wait before rejoicing! Speaking as a data scientist, let me forewarn you.

AI is like a child of four who is learning the ways of elders and mimics their behaviour. Be it Bard, or Bing or ChatGPT for that matter, it will go through the web and try it is best to match the parameters. With the boom of fake news, bogus websites, and bots creating news on the fly with the searched words, the AI bots train themselves with an incorrect database, unless the users patiently train the chatbots with the right prompts. In a real-life study with another AI enthusiast and Tech/Digital Transformation Leader, Alok Ranjan, there is a definite co-relation between the time spent on the training of the chatbot and the accuracy. It requires a good amount of training, about six to eight months to get the accurate report. The biggest risk of asking AI to draft a novel is plagiarism. Copying from AI to draft a novel can potentially expose the poor author to a lawsuit, as the AI will select plots and settings, and then generate the output from a million of similar writings. This brings to mind the experiment published in the Hindu. In the experiment, the researchers asked ChatGPT to author a novel using a list of prompts that would guarantee the next Booker Prize. The chatbot returned with a synopsis of that was identical to The Palace of Illusion, which in fact is a novel by Chitra Banerjee Divakaruni (published in 2008). When the team contacted Ananth Padmanabhan, the CEO of HarperCollins, to check if they could publish those. The CEO admitted he did not recognize that AI wrote it. He further explained that if more than one such query email came in, it would strike out, as they had a similar structure. That poses a considerable risk with more authors using AI to generate novel or query letters. It might look monotonous. Let me add there, there are software that can detect plagiarism and if AI bot wrote the text or not. The day might not be far when publishing houses would require authors to declare if they used AI to generate the content.

Related Articles

From ‘Just’ a Homemaker to ‘Only’ a Writer, Innovation Unleashed! – Jyoti Jha’s Story Inspires Many

IPS Vikas Vaibhav – The Bureaucrat with an Aura of being the Changemaker

Anand Law Institute Paves the Path to Legal Excellence in Deoghar

What am I pointing to? AI is not helpful in the creative pursuit? No, AI increases the creative productivity. Any author worth her/his salt spends a good amount of time on research. In my conversation with William Dalrymple in one of the LitFest, he stated it takes almost four years for a book. He suggested the best source was the libraries. Chatbots can reduce this time, especially with translation or checking for references. It makes searches easy. You have a list of suggestion for which one can fact check with authentic sources. Once again, the machines are still learning, do not take their reference sourced from web-crawling without a cross-check for facts. Still, it would still reduce the research time. The best utilization of AI to understand one writing style. Software like Pro-Writing Aid, Marlowe, Grammarly, or MS Word Editor Score help the authors with recommendation. They indicated on clarity, voice, along with a comparative benchmark with another author.

Let me demonstrate how to use AI effectively for a heavy-bulky novel.

Select the writing style you are writing. The AI will use material with a similar reference. Look at the grammar and spelling score. The AI will prompt for 100%, but it might not be true. The AI might lack knowledge of some words and indicated grammar. Look at sentence length. It should be with the environment and style you want to create. For example, if you are creating drama set in England 14th to 16th century, the word count in a sentence would be higher than contemporary teen romantic fiction. Few of the commonly used analytics used are checks for passive voice, slow pacing, long sentence, and readability grade. These help you check if your writing is matching the intended reader group. People do not widely use the engagement, start, sentence variety, acronym consistency, repeated words, or phrases, emotional tell, glue index, and dialogue and quote consistency scores. However, they are particularly important. These bring variety and remove monotony in the writing. The software provides a comparative analysis of against same genre. Monitor how you are faring against them. There is no said rule that would dictate you have to fall into the range of the genre. You decide how much above or below could you go. After all, it is a creative process, and your voice is unique; be unapologetic about it.

Finally, after completing this article, I could breathe easy. I closed my tired eyes. I was back on the cliff. The growling of the hounds was no longer audible. The call of the void had mellowed down. I had not succumbed to the temptation of an effortless way out. I have taken a route to make AI my fellow researcher. Drawing a sigh of relief, I opened my eyes. My cat, who was disappointed in me earlier for thinking about a shortcut, stared directly at me. She gave me three blinks to say; she loves me and is proud of my choice. What else do I need now?