Artificial Intelligence for Artificial Poems
A friend asked me recently if I thought there was a place in poetry for AI. It’s a good question, given how AI, in the form of flattering chatbots, is already offering us the ease of creating word documents that adhere to the rules of grammar and seem logical, well structured, and somewhat intelligent.
With poetry, if we ask, chatbots can mash up poetic forms or free verse with the particular styles of particular poets and any popular poetic subject matter in lightning speed. What they spit back at us looks and feels like a poem — and, indeed, might even move us emotionally, the way good poetry does. We could read it and weep — or laugh.
So my answer is that, for better or worse, AI already has or will continue to have a place in all human interaction, including poetry. Recent research already indicates that readers generally can’t tell the difference between AI-generated poetry and human-generated poetry. In one study, readers actually preferred the AI poems to the human-poems and assumed that the AI-poems were written by humans. The report notes:
“We found that AI-generated poems were rated more favorably in qualities such as rhythm and beauty, and that this contributed to their mistaken identification as human-authored. Our findings suggest that participants employed shared yet flawed heuristics to differentiate AI from human poetry: the simplicity of AI-generated poems may be easier for non-experts to understand, leading them to prefer AI-generated poetry and misinterpret the complexity of human poems as incoherence generated by AI.”
I assume that, in time, AI will get even better at writing poetry — maybe learn how to use incoherence effectively as an element of irony or truth. It may be that the best-selling poetry books in the future will be machine generated. Hard to say. But I also told my friend that I have no interest in AI-generated poetry, no matter how good it might appear to be. And I have no interest in asking AI to help me write poetry — even if it could help me improve a stanza or a line I’m struggling with or check facts for me (if facts matter in the poem). Even if AI could help me fool readers into thinking I’m a poetic whiz, draw accolades, win prizes, I would have no interest.
Why? It’s not just because the poetry that matters to me can only come from human experiences, a life lived. To ask a machine to produce a poem or rework a draft of a poem seems to me a pointless exercise. I think I’d also be too embarrassed to try to pass off such a poem as mine. I get that AI can sift almost instantaneously through all published poetry (or all that’s available online) and produce a piece that might move many readers. It may even be that I’ll admire the poem. And you could argue that, while AI is not human, the poetry it produces is an amalgam of lived human experience. In its way, then, it does represent us and our experiences. But when it comes to poetry, I’m only interested in what the individual human mind can produce — for better or worse. If we rely on AI for poetry, we’re essentially erasing our individual experiences, the very cool workings of our individual imaginations. Or we’re folding them into some kind of composite of experiences of millions of people. We, the poet.
May someday I’ll come think of my views as naïve. But I am 71 years old, and for all my life the poetry that has interested me has come from the living. From a person who has a heart that pulses at varying speeds throughout the day. From a person whose imagination is impossible to predict. The poets I admire are those whose way of being is tattooed all over their poems. I can feel their genius and their shortcomings, their insights and confusions, their quirkiness and brilliance. Their work will have the kind of limited perspective that bleeds on the page — or, despite all struggles, that sings on the page. I want the love and resistance, the empathy and anger, the petulance and reflectiveness, the actual hope and dreams of a real person. The poetry doesn’t have to be perfect. It just has to be human.
We’re already reading news about writers who use AI to generate their work — and have profited by such use. For some, it’s a matter of pure deception. They don’t think they can write an award-winning poem or story on their own so they ask AI to help. I’ve read that some romance novelists use AI to churn out whole novels for them quickly. And then there are the writers who use AI to “help them edit” their work. Even Nobel Prize-winning author Olga Tokarczuk admits to using AI for fact-checking her new novel. If readers are happy, perhaps this is all fine. But my sense is that while many of us are fine with AI playing a supporting role in the editing process, overall we’re looking for something genuinely human-made.
My guess is that if you ask the AI chatbots about all this, they’d agree with me. At least that would be the intelligent answer. It would say, “While I can generate a poem for you, you know that the poem won’t be yours, right?”
***
For my on-again, off-again poetry group, we recently gathered a few AI- and tech-related poems to discuss. In looking around for such poems, I was surprised, at first, to find that the subject of technology is not one many poets focus on. But in retrospect, this makes me proud of the field of poets today. I imagine, in time, that human-generated poems about AI and its intrusions and benefits will be more available, but at the moment, poets are drawn less to the question of machines and more to matters of human behavior and experience.
Among those poems we found, a couple caught my eye. The first is “All Watched Over by Machines of Loving Grace,” by Richard Brautigan. It was published in 1967 and appeared in a book by the same name. What stands out in this poem is Brautigan’s wish for a better, more peaceful world. On one level, this wish for a harmonious world of human’s and machines is a whimsical, utopian wish — one in which AI-like computers are integrated into our lives to serve the goals of democracy and justice. In this “cybernetic meadow… mammals and computers/ live together in mutually/ programming harmony/ like pure water/ touching clear sky.” Brautigan also envisions a world that balances technology and environmental health. His daydreaming here is offered with a certain lightness of spirit. But there’s a kind of serious desperation here, too. His three parenthetical phrases (one in each stanza) are the real subject of the poem. They contain the urgent need that so many young people felt in 1967 for human and environmental justice in an age of war, injustice, and increasing environmental degradation. The parenthetical phrases Brautigan uses — the sooner the better; right now, please; it has to be — may seem to us now as naïve. But I think the whole push for technological advances we are experiencing today have the same desire at its core — a desperate wish that technology can somehow (right now, please) save us from ourselves and our more damaging impulses. That AI is not actually serving this wish at the moment — given that AI is being corralled for war and surveillance and corporate profit — shouldn’t really surprise us. But still, it’s hard not to dream. What else do we have?
What pains me about this poem is that, back in the day, I believed deeply in the counterculture. I believed the baby-boom generation would, in fact, rewrite the rules of American culture so that we’d become a more just, peaceful, and verdant place. It’s still a wish of mine. But clearly we aren’t going to solve our problems by turning over all solution-finding work to machines. It’s hard to tell from Brautigan’s poem, but it’s very possible, as some critics say, that he was also mocking this wish — given what our early foray into technology had been able to produce in his day — much of it in support of the military-industrial complex. Who knows? But we do know now that Silicon Valley’s early-year pledges to use technology “to make the world a better place” have left the building.
The other poem that stands out is Adrienne Rich’s “Artificial Intelligence,” which was published in 1961. The poem is dedicated to “G.P.S.” I thought that meant “global positioning system.” But, of course, our GPS system didn’t exist back then. I’m told Rich’s dedication is to “General Problem Solver,” an early computer program designed to solve human problems.
Rich’s poem is astonishing to read sixty-plus years hence and realize she was tapping into a feeling that many of us experience today. The poem pre-dates electronic chess, yet Rich clearly could see it, and more, coming. She could also sense the human vs. machine tension that has been on the rise ever since. The poem opens:
Over the chessboard now,
your Artificiality concludes
a final check; rests; broods—-
no— sorts and stacks a file of memories,
while I
concede the victory, bow,
and slouch among my free associations.
I love how Rich catches herself in the poem. For a moment, she attaches human emotion to the machine, but then realizes her mistake. The machine doesn’t “brood,” as she first imagines. It can’t . Instead, it “sorts and stacks a file of memories.” I love, too, that she contrasts the machine’s post-game behavior with her own — as she concedes the victory and slouches among her free associations.
The triumph of the poem is that while the machine can trounce us in chess, it’s still a machine. Astonishingly, Rich anticipates the question of AI and poetry here — as if she can see where the whole story of computers was heading back then. Rich makes the essential point that machines can (at some future date) write poetry with ease, but even then the poem is not coming from the heart. In fact, the computer is being “forced” to take on this task.
Still, when
they make you write your poems, later on,
who’d envy you, force-fed
on all those variorum
editions of our primitive endeavors,
those frozen pemmican language-rations
they’ll cram you with? denied
our luxury of nausea, you
forget nothing, have no dreams.
It makes me smile that Rich, considering the possible human emotions she could include here, settles on the “luxury of nausea.” But it feels appropriate, too. For a woman writing poetry back in the early years of the 1960s, nausea at our intensely patriarchal society must have been a daily occurrence. To think machines would change that our society would have been an act of magical thinking.
Poetry aside, t’would be nice to think that AI could be a helpful tool in creating and sustaining healthy democracies. But it’s hard to see how that’s going to happen when the powerful people behind AI appear to be working primarily for their own wealth and power. And clearly, I’m not alone in my reservations. One recent report found that only 18 percent of Gen Z-ers feel hopeful about A.I., and almost half say the risks associated with AI outweigh the benefits (see those reports on job losses to AI). Researchers at Stanford University found that out of people in 30 countries, Americans had the least faith in their leaders’ ability to regulate AI.
In truth, I don’t give much thought to AI and poetry most days. I’m just reading the poetry generated by remarkable, hard-working, deeply committed poets — the ones who and bleed. In my dream world, however, AI, if it has the kind of intelligence folks say it has, will understand that it is being used in a game to create wealth from a small cadre of real-life A-holes — and will somehow work to undermine such wishes. What would that take? Among other things, it would require AI to truly understand and embody key human emotions, including the horrors of nausea.