Can AI write real poems?

Plus: A 9,000-year-old shrine, why Christopher Alexander matters and more

A rebuild of a British Bombe located at Bletchley Park museum. Tom Yates/CC BY-SA 3.0
Share
Text
Text Size
Small
Medium
Large
Line Spacing
Small
Normal
Large

In the New Criterion, Carmine Starnino writes about the newest AI tool, GPT-3, and its uncanny ability to imitate human expression. What does this mean for poetry and the other literary arts?
Dozens of websites, with names like Poetry Ninja or Bored Human, can now generate poems with a click of a key. One tool is able to free-associate images and ideas from any word “donated” to it. Another uses GPS to learn your whereabouts and returns with a haiku incorporating local details and weather conditions (Montreal on December 8, 2021, at 9:32 a.m.: “Thinking of you/ Cold remains/…

In the New Criterion, Carmine Starnino writes about the newest AI tool, GPT-3, and its uncanny ability to imitate human expression. What does this mean for poetry and the other literary arts?

Dozens of websites, with names like Poetry Ninja or Bored Human, can now generate poems with a click of a key. One tool is able to free-associate images and ideas from any word “donated” to it. Another uses GPS to learn your whereabouts and returns with a haiku incorporating local details and weather conditions (Montreal on December 8, 2021, at 9:32 a.m.: “Thinking of you/ Cold remains/ On Rue Cardinal”). Twitter teems with robot verse: a bot that mines the platform for tweets in iambic pentameter it then turns into rhyming couplets; a bot that blurts out Ashberyesque questions (“Why are coins kept in changes?”); a bot that constructs tiny odes to trending topics. Many of these poetry generators are DIY projects that operate on rented servers and follow pre-set instructions not unlike the fill-in-the-blanks algorithm that powered Racter. But in recent years, artificial-intelligence labs have unveiled automated bards that emulate, with sometimes eerie results, the more conscious, reflective aspects of the creative process. Microsoft’s “empathetic” AI system, Xiaoice, designed to explore emotion in language, has composed millions of impassioned poems in response to images submitted by users. Deep-speare, the brainchild of Australian and Canadian researchers, caused a stir when it taught itself to write Shakespearean sonnets.

These initiatives have now been dwarfed by Racter’s newest descendant. Released in 2020 by OpenAI, a San Francisco start-up, GPT-3 is an AI tool that was force-fed a vast portion of the internet (the entirety of English-language Wikipedia adds up to only a fraction of the billions of words ingested). Endowed with algorithms that help it make sense of all that data—”neural” algorithms modeled after the circuitry of the human brain — GPT-3 can produce, from a simple prompt, astoundingly human-like writing of any kind: recipes, actuarial reports, film scripts, real-estate descriptions, technical manuals. Of course, there is buzz for GPT-3’s poetic chops too. In one example, the American poet Andrew Brown asked the software to take the perspective of a cloud gazing down on two warring cities. GPT-3 delivered a rhyming poem that began, not uncharmingly, with “I think I’ll start to rain.” Stephen Marche, writing an article for the New Yorker, assigned GPT-3 maybe the trippiest poem in the English canon: Samuel Taylor Coleridge’s fifty-four-line “Kubla Khan” — an opium dream interrupted in the middle of its composition in 1797 and never completed. GPT-3’s mission? Finish the fragment. What the program fantasized was so sophisticated (“The tumult ceased, the clouds were torn,/The moon resumed her solemn course”) that readers unfamiliar with the original poem might have had a hard time discerning where Coleridge ended and the computer began. With an estimated billion dollars in backing, GPT-3 isn’t a better Racter. It’s a godlike Racter. Forbes named it the AI “Person” of the Year. Anyone who believed AI to be “nothing like intelligence,” said one expert, “has to have had their faith shaken to see how far it has come.”

Does GPT-3 show creativity to be “fundamentally algorithmic, a product of mere procedures,” Starnino asks? No. Poetry is an art, Starnino writes, “of brilliant accuracies, of reality re-described in ways that bind sound to perception. And here AI’s deficiencies are brutally exposed. Because to compete at this imitation game, a machine has to show that, by micro-adjustments of effect, it can draw our senses to the highest pitch of expression.”

In other news

Speaking of computers and literature: Maria Dahvana Headley’s 2020 translation of Beowulf starts with “Bro,” which makes it sound up to date, but the work fails to capture the nuance of the original text to such a degree that it’s hard to call it a translation at all. A visit to the University of Kentucky’s Electronic Beowulf, Maryann Corbett writes, provides a “striking visual lesson” of the poem’s complexity:

While I waited for Headley’s book, I raided my bookshelves and was taken aback by certain copyright dates. Frederick Klaeber’s third edition of the text—the basis of my Beowulf education — is dated 1950. Bad enough, but the first edition it was built on is from 1922, a rather alarming hundred years ago. Supplements at the back of the book tack on updated critical and textual notes that are too brief and too easy to miss; they barely mention, for example, the discovery in 1939 of the Sutton Hoo ship burial, and they say too little about the way it corroborates the funeral stories in the poem. There must be better tools for the modern student. There are. The most impressive is the Electronic Beowulf, a collaboration of the British Library and the University of Kentucky. Edited by Kevin Kiernan and programmed by Ionut Emil Iacob, it has a dazzling assortment of lookup tools. These tools let students and scholars see every page of the manuscript called the Nowell Codex, the sole surviving manuscript of Beowulf (which is part of the larger book called Cotton Vitellius A. xv), and they let us see it in bright light or ultraviolet, next to its edited version. They also let us zero in on each individual problem-spot with enlarged images of the unclear words, and they let us open with a click windows of commentary about those words so that we can dig into the layers of past scholarly emendations and conjectures. Even at a glance, the Electronic Beowulf (which I’ll call EB from here on) is a striking visual lesson.

The 1619 Project’s “junk history”: “Nikole Hannah-Jones’ new book sidesteps scholarly critics while quietly deleting previous factual errors.”

The architect Christopher Alexander has died. He was eighty-five. Michael Mehaffy explains why he still matters.

The philosophy of Alexander Dugin: “Russian philosopher Alexander Dugin may or may not have Vladimir Putin’s ear, but his philosophy of Eurasianism is worth understanding.”

Writers have all sorts of hobbies. T.S. Eliot loved to sail, Eugene O’Neill liked to drink, and Wallace Stevens loved to shop.

The selected letters of John Le Carré to be published in November: “Edited by the author’s son, Tim Cornwell, the book includes correspondence with Ralph Fiennes, Hugh Laurie and Alec Guinness, the actor famed for playing le Carré’s fictional spy, George Smiley, in adaptations of ‘Tinker Tailor Soldier Spy’ and other classic thrillers.”

Is surrealism back? Alex Ulam thinks so and argues it is a good thing: “When one considers the cultural politics and agendas of many contemporary Surrealists, it is striking how much they echo those of the original vanguard. Indeed, Surrealism was ardently ‘woke,’ long before the term even existed.” Oh, brother.

A 9,000-year-old shrine has been discovered in Jordan: “Researchers uncovered two standing stones featuring anthropomorphic carvings and a model of a ‘desert kite’ used to trap wild gazelles.”