Carmen Starnino

Reading “Poetry & digital personhood: On artificial intelligence and creativity” (The New Criterion, Vol. 40, No. 9 / May 2022) brought to mind a famous essay from two hundred years ago. In his “Defence of Poetry” (1821), Percy Bysshe Shelley takes issue with friend Thomas Love Peacock’s suggesting that poetry was no longer necessary given science and technology. Carmen Starnino’s article struck me as a contemporary reworking of the case for art versus our digital age’s utilitarian focus on data.

Starnino begins by describing attempts since the 1960s to build artificial intelligence (AI) systems that might actually “think” for themselves. These ranged across the algorithms of Racter and Microsoft’s “empathetic” Xiaoice, to Deep-speare that taught itself to write Shakespearean sonnets, and their many imitators and offshoots. Particularly astonishing was Generative Pre-trained Transformer (gpt-3) “endowed with ‘neural’ algorithms modeled after the circuitry of the human brain”, which could produce “astoundingly human-like writing of any kind: recipes, actuarial reports, film scripts, real-estate descriptions, technical manuals”, even what sounded like fresh and striking poetry. An intriguing example Starnino quotes is: “I must have/ Grey thoughts and blue thoughts walk with me/ If I am to go away at all.”

But was that poetry? If greater and greater amounts of data and operational patterning were fed into AI, one day could it actually leap the gap from machine to human? A disturbing thought indeed, especially for poets. Starnino writes: “Much in the way certain skills and industries are in danger of being replaced, artists are being asked to contemplate their own obsolescence. ‘Clearly, AI is going to win,’ one developer has said. ‘It’s not even close.’ ” Yet how ironic that developers chose poetry itself as the ultimate measure for AI’s hoped-for success.

Even so, their bravado left me uneasy. How would Starnino defend our art, especially in light of some of the startling compositions AI could produce? Where was the line drawn? I read faster, anxious for a cogent rebuttal.

In the end, Starnino does not disappoint, and his eloquent reassurance is well worth contemplating to appreciate what poets do at our best. That’s why rather than trying to paraphrase his conclusion, I urge you to read it in his own words, and of course, the entirety of his thought-provoking article:

In the meantime, Starnino’s conclusion also opens a larger question. Besides structural elements such as metre, rhyme, line length, syntax, etc. and a vast bank of words programmable by categories such as colours, number of syllables, vowel or consonant combinations, etc., what is it that makes a poem more than a linguistic LEGO AI construct? Conversely, humanly generated words run in short lines down a page may appear superficially like a poem, yet feel no more than prose cut up. What is the je ne sais quoi that makes the difference? Does poetry lie more in the synthesis itself of various elements, that complex balancing act, or is there something akin to the spiritual that is needed? Or perhaps the answer is not in the words we see or hear, but in their impact. Do they awaken? Delight?  Inspire? Soothe? And so on. Have the words been fashioned to communicate, to build a connection with another human being, as in the first two letters “co”, to share? Surely that must imply a human creator.