It seems to be difficult for my fellows in the commentariat to remember that things can get worse without threatening life on earth as we know it. I remain skeptical that artificial intelligence is going to bust loose of its own accord and start wiping out continents, or that it is going to cause a total death of white-collar work. Of course, like any technology, it will be harnessed to human wickedness and cause plenty of problems, but in this it is not distinct from the internet, or the internal combustion engine, or blacksmithing, or bipedalism, or any of the other human technical triumphs stretching back to old Adam and/or the first ape loping around the savanna with opposable thumbs. (Throw on some Wagner and cue the Heideggerian frowns.) And, as in all those other spasms of progress, some of these ills will be distinctive to the technology itself.
One such ill of the AI age has been obsessing the old editorial noggin lately, namely, the fact that quite a lot of would-be contributors have been chucking copy at me that they didn’t actually write. I’m a gentleman of modern outlook, and the teacher’s-pet outrage at people “cheating” doesn’t much move me. To my mind, there are three obvious, disqualifying problems with using the AI to generate your writing, and moralism isn’t one of them.
The first is purely economic. Let’s say we all collectively decide that pushing AI prose on our readers is unproblematic. Then why am I going to pay you to push buttons on the machine? I could just get my own subscription to the toy of choice and feed in my coverage priorities. Believe it or not, I mostly do not enjoy corresponding with contributors and, all things being equal, would happily liberate the ten or so hours a week I spend reading and writing emails. The writers who are dabbling in this stuff could use a bigger dose of longshoreman-style self-preservation.
The second is technical, and explains why I do not abandon myself to the heady temptation of never interacting with a writer again. A truism repeated does not become less true: The word “writing” describes a process as well as a product. The mustering on the page of facts extracted by research and reporting is the tangible artifact of thought; when you have to walk through your own arguments and do your own fact-finding, soup to nuts, you find things you haven’t considered and you refine your argument. The way AI is used by a lot of the slop-slingers in my inbox is to write out the boring but formally obligatory background for whatever allegedly novel arguments they are making—the donkey work where, if you are doing it yourself, you will really come to understand your subject (and occasionally reevaluate your priors). Use the thing for research if it’s useful, by all means, but if you’re not sitting down and forcing the facts to play with each other, you’re not actually thinking through the piece—and, if you’re putting your name on it, do you want to be responsible for arguments you haven’t thought through? This is the intellectual corollary to the old, sagacious IBM dictum: “A computer can never be held accountable, therefore a computer must never make a management decision.”
The final reason is aesthetic, which sounds pretty insubstantial but is in fact a natural complement to the other two reasons. AI tools have become much better at writing technically correct English prose; the early excesses of ChatGPT style have been successfully dialed down. (Or perhaps I should say, “These new tools aren’t just writing like ChatGPT—they’re smarter, cleaner, and more impactful.”) As of writing, however, they are still stylistically monotonous and weirdly slick or greasy; if they’re indistinguishable from human writing, it’s bad human writing. This conceivably is a problem that can be developed away. Perhaps in six months or so you’ll spin the dial on the Writing Box to the prose stylist of your choice, and it will spit out scintillating imitations of Ambrose Bierce and A.J. Liebling. But then we’re back to Problematic #1, which is to say, why am I going to pay you for something the box can do?
The first answer is, of course, that a unique style is itself a marketable good. No matter how good the Writing Box gets, it will never be able to develop a distinctive style for you; even if you would like to make it write the way you do, “the way you do” can only be developed by doing it yourself. The second answer is that, for all properly good writers, the style arises out of the thought, that is to say, the internal dialectic we described at Problematic #2. If you are shortchanging thought, you will have little sense for effective presentation; it is far from clear to me that any amount of computing power toiling over any number of literary corpora will be able to generate this aesthetic judgment, which is so intimate with the working of a writer’s mind.
I have written elsewhere that I do not think the technological annihilation of writers is a foregone or even the most likely conclusion; in an uncertain world, even if all the other virtues of the written word are somehow obviated by the machine, I think there will always be a market value for being able to hold the clown who wrote this responsible. The slop-slinging is, I hope, a passing fad. Of course, if I’m wrong, maybe it’ll be time to hang up the editorial spurs and turn myself over to the Quevedan translation of Thomas Browne’s Urne-Buriall. But at least then I won’t have to answer your emails.
Read the full article here
