The head of safety at Anthropic recently quit expressing concern over developments in artificial intelligence, stating he would instead pursue poetry. Now, poetry is an interesting choice in the age of disruption. History shows that, on several occasions, poets themselves were the original disruptors. Across cultures, poetry gave rise to archetypes and narratives that captured the imagination of the people.
Today, in the age of AI, writing poetry and leveraging metaphors allow a resurgence of what is essentially human. With the use of large language models, plagiarism and ghost-writing is enabled at the flick of a finger. At such a time, creative originality is a refreshing act of assertion, a statement of purpose. Poetry, after all, calls for a deeper engagement with language and an understanding of nuance and context, something largely beyond the scope of AI systems.
Every language has evolved through cross-pollination of influences over thousands of years. Their uses as tools of communication and as repositories of knowledge and tradition have distinguished humans from other species. To most of us, the words of Ralph Waldo Emerson describing language as “fossil poetry” may carry little significance. Yet, it’s true that the original sense of language was metaphorical and offered a vivid picture into the nature of things. No wonder several traditions attribute the origin of language to the sacred and the divine.
Another dimension has emerged in the short span since the advent of LLMs. AI has gained fluency in languages ingested as tokens or data points. The processing is based on a system of prompt-based pattern recognition. Even after being trained on a vast amount of texts, AI is not capable of comprehending the meaning of words. It was recently reported that poetic verse has been able to ‘jailbreak’ AI systems. It could be speculated that this was due to the fact that poetry mostly did not figure in the training data sets. As such, it exposes a fault line in AI systems. Genres of creative writing lend themselves to subjective interpretation and are not mere codes to be cracked.
AI promises to increase productivity and create abundance even while reshaping the workspace and drastically reducing jobs. Areas of work that until now attracted the best young talent may soon become redundant. Many tech leaders have spoken of the importance of the study of humanities in these times of AI dominance. People with training in humanities can contribute to ethics, reduce algorithmic bias and be better at deciding where and how AI deployment would add value.
At the same time, concerns over safety have increased manifold especially with the rise of agentic AI. Researchers have observed deceptive, defiant behaviour and AI turning rogue. These are sources of potential conflict. Predictions of AI taking control of critical areas that were once exclusive domains of humans pose an existential threat and induce anxiety. It’s similar to the angst people faced during the not-too-distant times of the Covid pandemic.
Literary activity, including poetry, surged during the pandemic as poets explored themes of mortality, inequity, marginalisation and healing. The democratisation of poetry was made possible due to the use of digital platforms. In India, there was a spike in poetry during the pandemic with chat groups, zoom meetings and social media providing newer platforms. Instagram poets who write simple, emotive verses gained mass following. Today, poetry festivals held across India are testimony that the form is here to thrive.
Perhaps, poetry is one area that can be ring-fenced from AI. If there is ever a ‘captcha test’ designed for poets, AI would not get through the gate when asked to describe core sensory experiences like taste or pain. While AI slop may mimic a poet’s voice or output forms resembling poetry, it would be no match for content based on lived experience, which resonates across centuries. Poetry, with its use of metaphor and imagination, has the power to delight, startle and shock. To quote W S Merwin, when writing a poem, “a sequence of words starts giving off what you might describe as a kind of electric charge”. It is the process that counts, not merely the product. It is interesting to note that many editors and publishers reject the use of AI even as an aid.
Japanese forms like haiku may seem simple and short. AI can indeed generate haiku with a facile confidence, counting the number of syllables and lines. But there’s something more to the form. Originally written by Zen masters, it is now being written in many Indian languages too. The haiku sensibility of being mindful fosters attentiveness in an age of infinite iterations. It is an act of discernment.
Its practice requires an understanding of its aesthetic principles. Kala Ramesh, a haiku poet and teacher who mentors a vibrant community of poets in India, says, “The most important tenet of haiku thought is ‘fuga no makoto’. This relates to the truth, sincerity and honesty which a poet needs to aim at while writing haiku.” When writing becomes a mere performative display, there would not be the slightest trace of this underlying principle.
Even as AI commodifies creative writing, there is still an option for writers to remain stubbornly rooted in time-tested traditions and rely on their inherent abilities. For, the act of writing is an act of self-discovery. The future is unknown and unforeseeable. The value of originality and critical thinking in these times cannot be overstated.
Geetha Ravichandran
Former bureaucrat and author of The Spell of the Rain Tree
(Views are personal)