Dharma’s dog

On The Young Poet and AI

1

Once I asked the poet Devadevan a question. We were sitting on a shelf of bare rock in Samanar Malai, Kilakuyilkudi, near Madurai; the still, meditative rock-cut of Mahavira on the face above us appeared to breathe. The sky was blue and clear, and the solid rock gleamed like old gold. Patches of green amidst the rocks made for some colour. It was a September morning in 2021; the sun had only been up a couple of hours, and the horizon in the distance was still an airy haze.

Devadevan sat on the bare rock in the centre, and gathered all around him were about twenty of us, all readers of his poems. He sat in our midst like the axle of a wheel. Behind him, someone had arranged slender, white, sempangi flowers in a small circle pookolam under Mahavira, with a small yellow wildflower that I could not recognise at its heart. There were colourful origami paper cranes placed on either side of the roughly hewn rock-cut steps leading up to our rock, the handiwork of our friends at Cuckoo who had organised the event. Devadevan sat silent, his skin a shade darker than the surrounding rock and the green punctuating it. The mild wind blowing over the rock seemed to pass over him the same way.

I intruded the silence abruptly. “How do you write poetry?” I asked. Immediately feeling foolish, I fumbled to try and bring some sense to that unruly question. “What I mean is, how do you find the words to express your experiences? I also have experiences that move me, sometimes tremendously. But I am never able to say it properly when I try to tell them. How do you know what words to use?”

Devadevan seemed genuinely puzzled by my question. “But how can you not know?” he asked. He looked at me out of his large eyes that appeared like pooled dark honey; there is something of a fearless mountain squirrel in his gaze and aspect. “நான் பார்க்கறதத்தான் எழுதறேன். I write what I see. அது சொல்லுதில்ல? அதை அப்படியே எழுதிடுவேன்.” He said it very simply, with absolute conviction. He always speaks like he is running his fingers through a mountain brook.

“But what do you mean, it tells you?” I persisted. I couldn’t quite follow. How did he leap across the chasm between feeling and language? “Experiences don’t have language. We first experience something… something indescribable, a feeling… language comes afterwards, doesn’t it?” I asked.

“But they are not two in me,” Devadevan replied. “They happen as one. Language is not outside. Language is within me. I am one person, am I not? So everything is one here,” he pointed with a finger at his own chest.

It would not be an understatement to say that Devadevan’s statement on that day hit me like a bolt of lightning. I was personally in a very confused state of mind at that time. I had just left my full-time career as a research scientist, where among other things, I had been studying how the brain represents perception and proto-language. I wanted to be a writer, intensely felt I had things to say, but did not know where to find the language to express them. My son was one year old. My parents, who were not happy that I had left my job, were also not happy that I had left my son with them on a Sunday morning to go meet a poet. My life felt utterly fragmented. But that day, with that revelation, something fundamentally changed in my understanding. It set me going on a quest that still fires inside me. How do I find that oneness of self where feeling and language, experience and its expression, are not two, but one?

That question acquired depth a month later, when I met Jeyamohan at a literary gathering in Coimbatore. It was a meeting on modern poetry. A friend of ours was hosting it in his countryhouse. The property had dark-leaved trees flowering and fruiting in abundance, and barnyard animals clucking around the courtyard. It was a break, and some of us were standing there, chatting in a small group. A female dog, full udders swinging, ran around accepting biscuits from the more generous among us, and three puppies followed everywhere taking turns to drink from it.

The sight prompted me to turn to Jeyamohan and ask a question that had been on my mind for some time. “Jey, what is indriya jagrata? You mentioned it in your speech about Mathar,” I asked. The poet Mathar had been awarded with a literary prize a few months ago. Jeyamohan had said in his speech that indriya jagrata was a quality the young poet possessed.

Jeyamohan looked at me. “Keenness of the senses,” he briefly said. He looked away, then at the animals gambolling at his feet. “Often a sensual keenness too. Rati bhaavam, it is called. A poet should have it. The keenness to sense within him, all the possibilities of the world around.”

I looked at the dogs too. The puppies, pushing each other, jumping over one another, trying to drink from the mother, and the mother, butting them back, playfully biting now one, now another puppy at the neck. A convivial roundabout of snouts and tails. One organism, really.

Perhaps the poet’s self drinks of the world thus, I thought. And perhaps this is how the world plays with him, bites him, nuzzles him, nurtures him. Perhaps that is how the poet’s self and the world stand in relation to each other. The picture awed me.

How intensely unified the self of the poet is! His sensory keenness, his desire, feeling, memory, emotion, his thought and intellect come together as one organism, with one aim – to express himself. His choice of form or his language – whether it is word-language or image-language or both – whether the images are visual, tactile or musical – are all intensely bound up with his inner self in a manner that they are, as Devadevan put it, not-two, but-one.

How does this intensified unity of self come about? I languish for the answer within myself. A part of me realises, after gaining a modicum of understanding of how poets work at their craft, that while some individuals may inherently possess a tendency towards it, usually it needs practice, training and toil to perfect it.

But such perfection is not the result of brute-force toil, like the labour required to lift sequentially heavier rocks or crack a series of increasingly difficult examinations. It comes about, as I understand it now, in three crucial ways.

First, by absorption and immersion, of letting the self and the senses stay awoken to all the manner of stimulations and experiences from the world around. Like a pear sunk into wine that becomes drunk with it, the poet soaks up life into his self. He may lose himself, take wrong turns, even go around in circles, but no matter. All experiences leave its sediments. All experiences go to perfect the self of an artist by uniting him with the world.

Second, through immersion in the arts, by developing a measured understanding of form. According to Devadevan, language is no different from perception, but while such a state might not have been one he meticulously and consciously trained for, I am sure his immersion in language and poetry would have in no small measure sharpened that impulse. Perhaps it happens best when one is playful, unhurried, fully attentive in the manner of children. Perhaps sustained attentive practice plays a role. Perhaps the more one practices a form, unites their self with it, the more the distance collapses between the form and their self.

Finally, the perfection of self requires the courage to follow the voice of the self. However faintly one hears the call of the self, the poet-artist pays heed and follows its path; the more persistently he follows it, the clearer its voice gets. The quest of the artist is always a quest for the self, which by definition involves a rejection of all that is not the self.

None of these are original ideas. I am merely repeating what the masters have said. ‘எழுத்தாளருக்கு எல்லாமே அனுபவம்தான்,’ I once heard Jeyamohan say. Yehudi Menuhin famously played the violin to rest when he was not practicing. ‘Ekla cholo re,’ says Tagore. If no one heeds your call, doesn’t matter. You walk alone.

Thus the experience of life, the conviction to express and the language of expression are not fundamentally different things for the poet. They are experienced as one in his inner world, they turn as the wheel of his creative practice around the axis of his self.

What do we hope to hone when we hone our creative practices? Our self? Or everything around the self that allows the self to express itself better? Are we cleaning the mirror so that the face can stand out more clearly in it?

It is such matters we are compelled to think about when we think about AI, whether it has any role in the creative arts, including creative literature, and above all, the relationship of AI to the creative self.

2

Does AI have any role in the creative practice of the poet and the writer? If AI can play a role in creative literature, then how does the young writer use it – and not use it? AI use has become so pervasive that I believe the time has come for the creative-minded among us to consider these questions seriously and reflect upon them.

A couple of days back, I wrote a rebuttal to an opinion on the use of AI in creative translation practice. I found myself so acutely repelled by the idea of using AI in creative translation – that is, the idea of using it to generate sentences that the translator can then choose from, edit and improve upon – that I was forced to take a step back and ask myself why I felt such an acute repulsion at the idea.

After all, AI is here to stay. Newer models are coming out every day, and there is a sense of a gold rush about the whole thing. Most of us are either comfortable using, or are increasingly encouraged to use AI in our day jobs in some form or shape. In many domains including science and technology, AI is radically transforming the outlook of the field itself. Scientific and medical discoveries are projected to increase exponentially in the coming years thanks to the burgeoning cognitive capacity of upcoming AI models. This is probably true of a lot of fields where raw computing power, sophisticated pattern recognition and deeper cognitive capabilities matter.

But what about the arts? Specifically, what about literature? Can AI help us ‘speed up our literary output’ – an unfortunate combination of words that already possess a distasteful, middle-manager-like ring – and is such a thing is even desirable?

Today, AI is already widely used for routine tasks in publishing, including proofreading and copyediting. A few days back, Chiki Sarkar, publisher at Juggernaut Books, had spoken about how they use AI to ghostwrite the first draft of non-fiction titles. While Chiki was clear that they do this only for non-fiction, and only with people who may have a story to tell but not the writing skills to do so, who would have otherwise employed ghostwriters, and that the process is disclosed in the books, it still drew flack from readers. The fact that AI takes away human ghostwriting jobs is one complaint. But much of the umbrage comes from the fact that AI has built its current corpus by drawing upon the labour of human writers. The fact that human labour is used to train machines that can then replace humans does not sit well ideologically with many people in the humanities.

However, when it comes to creative literature, the issue runs deeper. It is becoming increasingly common to observe young creative writers claim that they use AI in some part of their creative process, that it is a tool just like a search engine. Spellcheck, grammar check, proofreading, copyediting – many people would find the use of AI for these tasks relatively acceptable even in creative writing. Perhaps some of us use AI for background research, fact-checking, or to understand difficult topics – the possibilities and limitations therein is a separate conversation to be had.

But what about the actual creative part of creative writing? Generating plots? Designing character arcs? Structuring the pacing? Working out a dialectic thesis, a vision? Can AI play a role in helping the young writer come up with these elements of his creative fiction? Note that the verbs like ‘generating’, ‘designing’, ‘structuring’ and ‘working out’ already assume a mechanical, architectonic aspect to the writing process.

While there is a structural aspect to writing in any form from the venba to the novel, creative literature, as we have learned, is one that unites a sense of self, the desire of that self to express something, to the form to create something unprecedented.

The question then arises: how are we to keep this inner desire of the self, its singular voice, undiluted, if we are to go about outsourcing the structuring of the creative work to AI? More fundamentally: how is the young poet to build this muscle of self-expression if he starts out his career borrowing possibilities from this cognitive machine instead of exercising the uniting of his own faculties? How is Devadevan’s not-two but-one union of experience and expression possible through this practice?

No, this is not merely a rhetorical question. I am not writing this essay as a rebuttal anymore. I am sincerely engaging with the possibility that a creative artist can use a cognitive machine at this level.

After all – and I am ventriloquizing the argument of the young AI-partial poet here – when we read books and immerse ourselves in the arts, our brain recognises and remembers patterns from them. Patterns that are presumably personally important for the poet, that are decided by the priorities and desires of his inner self. These are what we called motifs, movements, images, arcs or tropes in literary theory. Those assimilated patterns are then used, consciously or unconsciously, when the young artist creates his own work.

Now, the AI-adherent may ask, why can one not use a hyper-efficient pattern recognition machine to help him discover those patterns? Why lose out on its possibilities? After all, visual artists adapted to the still and moving camera, and singers adapted to the microphone and recording systems. In fact, those who didn’t adapt became obsolete. So would it not be foolish for the young writer to not capitalise on AI’s possibilities to further his artistic quest?

I’ve had these questions too. I have played around with AI, and I have observed first-hand that it is ridiculously efficient at this sort of pattern generation. Given a scenario or a premise, AI can generate about thirty different plots within a minute even as the average creative writer struggles to come up with three.

For example, just prior to writing this essay, I asked ChatGPT to summarise the Virata Parva from the Mahabharata for me, and it gave me an accurate summary in a few seconds.

I then asked, ‘Are there any other stories similar to it within the Mahabharata itself?’

It meandered for a couple of sentences, considering stories like Rama’s exile and Shakuntala’s recognition, before settling on the story of Nala and Damayanti.

‘It is the single strongest Mahabharata echo,’ it claimed. ‘In fact, the Nala story is told to Yudhishthira during the exile precisely because it is a mirror.’

‘How did you arrive at this match?’ I pressed.

It gave me a few parameters. Plot structure. Motifs. Symbolic patterns. Historical relationships.

‘Fine, give me a few knots for a novel that layers these two stories,’ I asked.

I watched as the screen rolled before my eyes listing plot after plot. Damayanti in old age arrives at Virata’s court and tells Draupadi stories. Kali that once possessed Nala now becomes obsessed with Yudhisthra, a man he cannot fully corrupt. Virata’s court is revealed to be a labyrinth of masks and disguises. Yudhisthra descends into madness as he realises Nala’s story can never explain his own. The dice that ruined both Nala and Yudhisthra tells its own story.

Truth be told, I was momentarily overwhelmed. One of our current definitions of intelligence is that it an advanced form of pattern recognition. If AI is able to make these connections, then where does it leave human intelligence?

But that anxiety subsided almost immediately. First, I understood what the machine was doing was a sort of pair-wise correlative pattern matching, that none of the plots were more sophisticated than that. With a limited set of characters, one can generate pair-wise matches until all the possibilities run out. It is like a game of cards. They are ranked by similarity and the similar pairs are turned into a plot. Additionally, the AI had landed on the post-modern idea of a repeating, yet inconsistent story – perhaps well-represented in its corpus – and was creating various nested patterns of the same idea; it was this element that made the plots appear literary.

All of these look like plausible themes, and may still work if one was writing genre fiction, where such variations make up the entire industry. But three elements stood out for me that marked the entire exercise as fundamentally uncreative nonsense.

One, all the plots are squeaky clean. There is none of the messiness that accompanies human ideation, where a writer has so much to say, realises the form in front of them cannot hold it and attempts to mould the form to express his idea.

Two, the plots don’t appear to originate from any central directing impulse, but rather like mathematical overflow.

Three, the plots are presented in a remarkably vague, averaged, smoothed out language.

However, I reasoned, perhaps this is the AI’s ‘rough note book’ voice. Perhaps if we push it in the direction of a specific plot that aligns with our own interests, it does better. Perhaps it writes better prose when specifically prompted to write a novel. But when I asked the AI to generate the first 2000 words of one plot and started reading (it wrote it in fifteen seconds, a task that usually takes me a whole day), I began laughing.

What we first sense when we read AI-generated prose is that it has no centre of gravity. There is no ‘writer’ behind it, no central unifying impulse, no self. Its sentences are vague hand-wavy gestures, averages of averages, with a surfeit of abstract nouns. Sentences like ‘The mirror becomes not merely a surface of reflection, but a shared space of responsibility, remembrance, and recognition sound lofty, but are eerily vague and empty.

This is the foremost difference between AI and human prose. When we read Jeyamohan’s Neerkolam – the 13th in the Venmurasu series of novels that also draws upon the Virata Parva and the Nala-Damayanti story – we intuitively every sentence, every plot, every character as part of one unified organism. It is this coherence that lends any story its authority – quite literally – for it emerges from the coherent self of the author.

AI as it exists today has no means to experience life. No conviction to express anything. What it has is a pale imitation of language, culled by pooling and averaging thousands upon thousands of any particular type. Can we unite this machine’s language with our personal experience and inner will to express, and create something unified? I am not convinced.

Thus I don’t know whether it is possible, or for that matter, advisable, for a young poet to select plots, characters and sentences from an AI-generated surfeit and cobble together a literary novel from it. I doubt whether it is possible to create such unity of purpose when experience, motivation, imagination, knowledge of forms and vision are not already unified by creative practice within a single artistic self. Moreover, I fear whether, even if I try to, the use of AI itself would inhibit such a process, and that my prose, my thought, my very voice would slowly begin reeking of AI if I kept using it.

An AI-generated sentence is the canary in the coal mine that is AI’s black box – a person sensitive to language is immediately awoken to the amount of averaging, abstraction and generalisation that is happening within that black-box. Such generalisation is antithetical to the obsession with the particular that emerges from a perceiving, thinking, questing self. What the creative writer desires above all is to write in their own voice, with their own gait, bearing the stamp of their own personal idiosyncrasies and obsessions. This is why, while AI may find its use in logical domains and can potentially be useful for proofreading, copyediting or fact-checking, I feel we would be wise not to use AI when attempting to write creative prose – whether that is original writing or translation. If speed is of essence in creative writing or translation, then the obvious solution appears to be to nurture human practitioners, not seek machine-made solutions that can later be broken by the limb to fit the box.

Genuine creativity is a tender flame. We strive every day to protect it with both our hands as we ride through wind and storm. Some may perhaps harbour anxiety about losing out on the benefits of AI, but for me, the greater anxiety is about losing my creativity – and therefore my self – to AI. 

3

There is a well-known story by Isaac Asimov, called ’True Love’. Milton, a programmer, is in search of true love. He programs his very smart computer, Joe, to filter through its database of co-workers and identify women who would potentially be a good match for him. Joe identifies a few candidates, and Milton arranges for them to be transferred to his department so that he can get to know them. However, he finds out he is not happy with any of them.

Milton intensifies his search by spending increasingly more time with Joe, pouring his heart and soul into the machine, and having the machine talk to various women on his behalf. In the end, Joe identifies a perfect match – Milton’s true love – but informs Milton that, filled with Milton’s dreams as its processor is, it has now developed desire for the woman in question, and would rather keep her for itself, and therefore has turned Milton over to the police for malpractice.

As I was writing this essay, I found myself thinking of this story. The AIs today are very intelligent in many ways. What I see, however, is that none of them have desire. A highly sophisticated chess AI may have the intelligence to win every game with Magnus Carlsen, but it has no inherent motivation or drive to do so. These machines do not crave.

That is why Asimov’s story is both strange and poignant. The human, Milton, has become become static and almost impotent in his desire – a modern-day incel, if you will. He is so inept that he outsources his quest for romance to his computer Joe. The machine Joe, in a strange twist, develops desire – perhaps, one could even say, a kind of indriya jagrata – as well as the cunning to realise its desire, and thus appears to have become quasi-human. In the story, Milton appears to have lost his self, and Joe appears to have found its.

This tale brings up a rather foundational question. What is the human self? Where is it to be found? What is its locus?

Perhaps it is fair to say that many of us with a proclivity for thought centre the self in the mind. We identify ourselves with our cognition, our thoughts, our ideas. Perhaps that is why present-day AIs and their intelligence fascinate us and bother us so much in equal measure – because their cognitive and pattern recognition capacities appear to rival or even outstrip our own. If AI becomes as intelligent as us, then who are we, we wonder. Perhaps that is why we strive to unite their intelligence with our own – because we want to become bigger than them, because we want to become bigger than our fellow humans who may reap the advantage of using them.

But the experience with AI, I think, also teaches us something fundamental about ourselves. If our cognition can be abstracted away from us, yes, developed in the highest degree, if a machine could some day possess the cognition that rivals or outstrips human cognition – then is my human self really to be found in my cognitive ability?

No. My ‘I’ appears to be something beyond that. A questing, desiring self, a voice beyond my thoughts, a voice that can know and observe my thoughts, a voice for which my thoughts are but a tool.

Where is this desiring ‘I’? In the body? In the senses? Is it located at the portal of my experiences, where my sense of ‘I’ meets the world? Is this how we really should understand indriya jagrata – not just the senses, but the sensor of all our senses, a meta-sense that is aware of and awoken to what the senses bring in?

We can go one step further, and ask, does indriya jagrata even require physical senses? Even if each one of my senses are put out, one by one, like the lights going off in a house at night, would that silence the one within who breathes, I am, I am, I am, and craves, with every breath, for something unknowable, unperceivable?

I am reminded of the novella Madhilgal – Walls – by Vaikom Muhammad Basheer. Basheer is incarcerated in a prison cell. He is permitted to have the run of the grounds during the day, where he begins to cultivate a garden. However, the prison’s walls still hold him in. On the other side of the wall is the women’s prison. A woman prisoner’s voice begins speaking to Basheer from there. The voice says its name is Narayani. Basheer and Narayani fall in love. Through the wall, they utter sweet nothings to each other. They tap on the wall to send messages to each other. They gouge a hole in the wall, each from one end, so that Narayani can see the flower that Basheer has grown for her. As I remember this story now, I feel, such are the walls that hold in our craving selves, and such is the drive within us to pierce the walls to commune, even if momentarily, with the other.

But the novel comes with a twist that further deepens our metaphor. The prison doors open one day, and Basheer is led, much to his surprise, out of the walls. He has been released. He realises that he has the opportunity to see Narayani now. But he does not want to, and indeed, runs away from the place. It makes me wonder – does the craving to cross walls, commune across walls, hold no meaning in the absence of walls? Was Basheer now perfectly satisfied in the sweetness of his own company? When all the lights go out, what does the self crave for?

It is in such sophisticated ways, with so many unresolved questions, that we approach and try to understand this inner perceiving desiring voice inside of us that we call our self.

Today’s AI has no such comparable self, no such desire. We recognise that much clearly. Some people consider it the perfect slave – perfectly legal slavery, for it is not human, and therefore it has no desire that we are operating against when we make it do our will.

And yet, because a machine has no desire, it can operate on the desires and whims of anyone who controls it. It is the perfect vector of human desire, an empty capsid that can be manipulated by the DNA of power. We have seen this with the rise of internet companies and social media over the past twenty-five years. The desire of their founders and shareholders manipulates these Bhandasura-like entities; and they in turn are built to enslave our desires to further the whims of their masters. The whole thing is making castrated bulls of us all.

It is in the face of such orchestrated attempts at co-opting human desire to fuel machines that have none, that we go back to the voice of that faint little ball of fire in our breasts that jumps to breach its walls while somehow simultaneously also being fulfilled with itself. I am greater than you, for I quest and I lead, and you have no recourse but to follow me, announces this voice with the conviction of fire. Every artist knows this voice. The young poet perseveres every day to preserve this voice, to give it form and shape with his own life and his artistic training. When his inner desire unites with his experience and language, he becomes a greater manifestation of his self, now realised as art. Not-two, but-one.

Freedom lies in heeding this voice. The greatest human freedom lies in safeguarding our conscious sense of self, our inner desire and our attention towards realising our self. All other freedoms may be taken away, but the last of human freedoms exists in the knowledge of the phrase: I exist, I am. Everything else follows that fundamental knowledge. And that is why, above all, and contrary to anybody else, the artist must safeguard his self and always stay one step ahead of any artificial sort of intelligence. Because he the artist led by his self, and his self cannot follow anybody else’s.

There is a reason artists must keep our senses keenly attuned to our inner selves at all costs. Today’s AI does not desire, it is not a self that has motivation and drive like we do. But what if, as many computer scientists predict, some form of general intelligence arrives in the future? What if those machines develop some sort of a drive, or agency, becoming human-like in its goal-directed behaviours? How will we understand the nature of the human self, human creativity then? Would human desire be distinguishable from machine desire at that point?

I honestly don’t know the specific answers to these questions today. But I suspect that at that point, the difference would be embedded in our embodiment, our connection to the earth, our ability to perceive pleasure and pain through the limited span of our “walls”. In any case, I suspect, and quite strongly, the human self’s understanding of itself would be one step ahead of any machine, if the human self has kept up its ability to be creative and to reinvent itself. Perhaps the human self, the human consciousness, would be felt to be something pantheistic at that point – contained in embodiment – a sense of oneness as nature that an algorithmic emergent property cannot ‘feel’ or approximate. In any case, such models of understanding our inner self can come about only if we have the undying quest for it, only if we discover the forms to express them.

That impulse can come about only if the self always goes one step ahead of anything else. Everything follows the voice of the self. That is the spirit of the avant garde. Our forms will evolve to describe our understanding of our self. And we should have the wherewithal to discover those forms, and not be mired in the quicksand of old forms and old modes of thinking, which constant engagement with artificial thinking machines like AI will force us into. Thus the only education a young poet should aspire for is this: to pay constant attention to the voice of their inner self, to make sure its desire are heard clean, to unite it to one’s experience and artistic training through practice, to always allow it to march one step ahead.

What this sums up to is the training of that one resource that has become mighty scarce in our post-internet, post-AI age: our attention. And that training is inherently anti-AI. Gandhiji’s autobiography was called “The Story of My Experiments with Truth.” A similar post-AI age autobiography might as well be called “The Story of my Experiments with Attention.” Yes, we poets drink from our mother, the mighty dog that is our world. It fascinates us in many forms, including artificial forms of our own intelligence. Yet, we must always be led by the voice of our inner selves, march one step ahead of her, and have her follow where we lead. In the Mahaprasthana Parva of the Mahabharata, Dharma did not follow the dog; the dog always followed a step behind Dharma. Truth is not very different from dharma, and today, we understand, neither is attention.

*