Literature is an ever-expanding canvas - welcoming new voices, styles, stories that mirror shifting challenges, values and fears of our society. Now a new frontier emerges - artificial intelligence (AI). In 2021, Vauhini Vara, a technology journalist and novelist, wrote nine short essays with an unusual co-author - AI (an invite-only version called ‘Playground’ developed by OpenAI). These essays, published in Believer magazine, went viral touching a chord (or nerve) with readers and of course writers. She turned to AI to tell the story of her sister's death, a story she couldn’t bring herself to tell.
Over the nine essays, Vara gives increasingly more detailed information to AI to work with. AI first makes stuff up, ending one story with a complete recovery. In another story, AI has the protagonist processing her grief through running and turning into a runner for cancer benefits (Vauhini Vara does not and is not in real life). Again, AI is fed more information. The author has to reveal more. With nudging and coaxing, the later essays capture her very real grief and tell a fictional story that is authentic. Vauhini Vara herself attributes one of the best lines in the pieces to the AI. However, she is cautious and warns of Big Tech swallowing language, in the same way it swallows things like friendship and community into online followers and marketplaces.
The evolution of AI as a creative tool poses profound questions about authorship. Unlike conventional tools like pens or word processors, AI transcends mere functionality. Its output is creative and involves intellectual efforts. A tenet of copyright law is protection of “expression” of an idea and not the idea. For example, we can all think - Love is best and the worst. However, both the Delhi Belly song lyricist who penned “I hate you like I love you” and Ghalib who wrote “Mohabbat mein nahi hai phark jeene ya marne ka” (Translation: In love, there is no difference between living and dying) expressed the same idea and would hold rights over the unique lines and not the idea. With writers collaborating with AI, this delineation blurs. Are we just human idea generators and AI - the expressors. Does the task of expression, once solely attributed to human authors, now shift to the AI itself?
In India, the Indian Copyright Act, 1957 governs “literary work” and the author holds all rights in her literary work. In computer-generated literary work - the “person who causes the work to be created” is the author. With AI who is the “person causing the work to be created”? The person who prompts the AI? The corporation that owns the AI (OpenAI, Microsoft or Google), the team of developers that built the AI? OpenAI’s terms of use say the user (prompt generator) owns output. Some platforms assign copyright to user. Yet the question arises: can they assign ownership or decide who is the owner when they may not inherently possess it?
Perceiving AI as an author provokes contemplation of its creative potential and its implications. Are we admitting machines are as creative as humans? More worrying than the human versus machine question is - are we then moving creation of art, literature, etc., to corporations? Given the time-saving nature of these technologies, will creation be the purview of a select few that can access and afford these technologies?
Currently, Copyright Offices globally dismiss AI authorship claims. In India, an application listing an AI (Raghav) as sole author of an artwork was rejected, later receiving registration with Raghav credited as a co-author, only to be withdrawn a year later. Our existing legal frameworks see authorship as human. However, the sophistication of these tools may change that.
Simultaneously, there are enough examples of the limitations of large language models. ChatGPT and BingAI have famously been trained on content it was not supposed to have access to. Assuming these are temporary hiccups, the fact remains that much of AI’s literary output remains pedestrian. Even Vauhini Vara who coaxed poetic essays out if it failed to recreate the experiment on ChatGPT3. She reached out to AI researcher Sil Hamilton, from McGill University to understand. He posits that ChatGPT will write badly because that is its job. OpenAI trains language models to sound safe, inoffensive, upbeat - i.e. boring. Great literature isn’t boring.
While AI is intelligent, it is an imitator, without its own unique perspective to share. It learns from and represents one small part of the world. Overwhelmingly, content on the internet used to train large language models tend to be in English, male, replete with corporate speak and cliches. Can AI ever transcend the limitations of the biases inherent in the language it is trained in? Again, the threat is not if AI will replace humans as literature creators but rather if we will see an unfair prioritization of one kind of literature - one that is just easier, faster and more cost-efficient to produce.
Merging of AI and creativity looms as an enigmatic realm - exciting yet unsettling. Our practices and laws need to play catch up. What we most need while we create new legal frameworks to govern AI generated work- is to ensure equitable rights for all. A challenge that we have not yet risen to.
This article is the first of a multi-part series on our "Future with AI".
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
