AI will match human creativity for sure

Most of the academic writers I know acknowledge that artificial intelligence is beginning to encroach on their field, including the ability to write a cogent essay demonstrating critical thought. If it’s a bit bland or generic, well, so are a lot of student essays that we’ve read over the years. But we still claim as uniquely human the ability to write creatively. Only humans can combine feelings and words and their own associations to form art. That’s not true. AI is, among other things, a web scraper, a gatherer of what’s out there, a synthesizer. But so are writers, and what we value is their idiosyncratic skill at putting together what’s out there. As poet T.S. Eliot wrote in his essay “The Metaphysical Poets”: “When a poet’s mind is perfectly equipped for its work, it is constantly amalgamating disparate experiences; the ordinary man’s experience is chaotic, irregular, fragmentary. The latter falls in love, or reads Spinoza, and these two experiences have nothing to do with each other, or with the noise of the typewriter or the smell of cooking; in the mind of the poet these experiences are always forming new wholes.”

But AI can assemble something akin to that, even if the formation currently involves more fumbled tries in its blind workings toward complexity than a human writer might attempt. In fact, the processes of a large language model, or LLM, are slightly opaque to even computer scientists. And the models are trained on huge sets of data, as are humans. The students in my writing workshops strive for something original, a high bar to pass. Definitions of originality vary widely: presumably something we haven’t seen before. In fact, a lot of us need to read more because what we think of as original has been done and done before we were born. “There is nothing new under the sun,” reads a famous line from Ecclesiastes, undoubtedly not the first time that sentiment has been expressed in a similar way.

So if originality, per se, doesn’t exist, what are we looking for? This: Take a known element, such as a toddler’s birthday party, and join it to an element you usually don’t see associated with it, such as a funeral. The occasion becomes decidedly mixed. I just made up that scenario, but you can too. It’s called recombinant. The video player is currently playing an ad. Or take an ordinary object, such as a pen, and in your story, use it to dig a hole. This is repurposing. And lest you scoff at such an easy method to produce something new, understand that we certainly do applaud that as something original, from innovative ad campaigns to a short story that includes five languages.

Those who judge creative writing in the classroom are so interested in process — “This poem went through five drafts” — that we forget that in the end, the finished poem is what matters most. Does it make a difference whether a short story is the result of months of painful choices and revising or simply randomized selections that finally yield a masterpiece (six monkeys at typewriters eventually coming up with Hamlet)?

If so, why? A creative writer produces a work of art. Are the “how” and “why” relevant? How does one know, anyway? Think of the Turing test: If an observer can’t tell whether a response is human-generated or machine-generated, then the machine can be said to exhibit intelligence. And if some discerning critics still fondly believe that the novel they admire could have been created only by some demented genius in a garret or basement or castle, that’s romantic delusion. Perhaps it was written by the AI inhabiting a data server in Texas.

Last year, a poet in a nearby English department thundered that AI would never write poetry like Emily Dickinson. So another professor typed into ChatGPT, “Write a quatrain in the manner of Emily Dickinson that uses black-and-white imagery and the idea of drowning, yet with a sense of elation.” In a matter of seconds, the AI had executed a pretty darned good quatrain, which I won’t quote because you can try it yourself, and anyway, ChatGPT will come up with a new one every time.

The kicker? The thundering poet termed the result uninspired and banal, so the other professor went into Dickinson’s 1,800-poem corpus, retrieved an obscure quatrain and presented it to the poet, who called the result banal and uninspired. A surprisingly high number of professors I’ve talked to have strong opinions on AI but have never tried it out.