“The college essay is dead,” declares an Atlantic headline, and what killed it is a new, artificially intelligent chatbot created by an outfit that Elon Musk — the media’s now-dearest whipping boy — co-founded in 2015. The program can respond to complex questions superficially or encyclopedically or, if the user so instructs it, in sophomoric five-paragraph write-ups. Compared with previous iterations, gushes the New York Times, the new chatbot is “smarter. Weirder. More flexible. It can write jokes (some of which are actually funny).” (They aren’t.)
But set aside its imitations of poetry, the accidental humor, its Seinfeld scripts and the confidently asserted rubbish — these prompted by over a million users’ playing around with the program since it launched last week. The news confirms what we ought to have suspected by now: We’ve miscalculated AI’s potential. Many activities that perhaps we never imagined would lend themselves to automation — protein-structure predicting, refugee resettlement, designing covers for the Economist — have recently done just that. There’s no reason for creative enterprises, such as art and writing, to be AI-proof so long as we conceive them merely as tasks.
Before you wonder — yes, my prematurely aged, reactionary brow did wrinkle at the prospect of change, but it regained its stoic repose at the realization that the birth of creative AI would be the demise of a previous, particularly unfortunate novelty: the content writer.
Content marketing has been around probably since John Deere first published the Furrow to envelop its farming-equipment ads in prose. (There’s something visually tranquilizing about verticals of script.) Once every product was given a website in the late 1990s onwards, the practice became ubiquitous. Now, decently capable AI that can handle copy, website boilerplate, and social-media piffle might free humans of professionalized insipid communication.
Are robots, then, asks Paul Krugman, coming after so-called knowledge workers in highly specialized jobs? He’s right to be concerned. His own scrawling might as well be replaced by mechanically generated text. (This is probably the sort of joke that AI can also learn to make, raising the possibility of our soon being able to let the algorithm produce Krugman columns and Krugman-column “clapbacks.”) Many mass-manufactured lines that appear in modern publications — especially online, where space is infinitely stretched and time compressed — can be technologized because they have more in common with copywriting than with writing: They’re predictable, bland, anodyne, styleless, and trite.
By contrast, writing conceived as a practice that constitutes intellectual, literary, and aesthetic traditions is how we participate in an endless conversation with our contemporaries, with those long gone (before we join them), and with those who will follow (before they join us) — with all of whom we share the tragedies of human experience. To robotize this conversation would be to remove the conversationalists: This is where AI’s potential ends by definition.
You read a short story by Jorge Luis Borges and witness how this author, in Argentina in 1947, plays with solipsism or how he sees villainy and salvation. AI can mimic Borgesian writing, but it cannot write Borges. Similarly, I’m interested in what Dominic, specifically, has to say about tracking truckers, or Luther about Americans’ love of gambling, and not what information can be pooled from the digital depths about these subjects and distilled to a page. Even a machine-learning algorithm trained on everything Luther has ever read, said, and written — a computerized facsimile of the old boy — wouldn’t be able to replicate the organic mess of personhood (physical decline and the fear of death, moral dissonances, a sense of the absurd, visceral impulses, epiphanies and insecurities): the chaos that continually and uniquely shades one’s opinions.
What of entire genres of literature whose function is purely to entertain? In these cases, because the author’s voice and tone and character are secondary to helping the reader pleasantly kill time, such work can conceivably be mechanized without diluting the (here’s that word again) content. Hence the romance or young-adult fantasy novelist might follow the corporate writer as he is politely shown the door.
Artificial intelligence can be more intelligent than we thought. To be proven wrong, or unimaginative, is often exhilarating; I’ll choose to greet this as one such moment. If cliché is the aid of the trade — a regurgitative college assignment, bloggy filler, self-regarding or background art, or any creative product that is valuable solely as content, and not as an extension of the individual psyche that produced it — AI will prove lethal. These are trends rather than traditions, so mourn them not.