When I left Plav, I’d spent two nights and had finally worked out how to stay a third. That morning, I was at the bus stop by eight. I waited for nearly two hours in stoic silence next to a man with a large moustache and a flat cap. I wasn’t sure it was the bus stop. Neither was he—or I misunderstood his Montenegrin. I was watching a wasp scrape melted ice cream off the pavement with its mouthparts when the bus pulled in. The road from Plav to Gusinje follows a river with several names. Its colour shifts—from sky to bone. It was the sort of road where even young children walking to the shops know to press themselves against the guard rail when a car comes past. In Gusinje, I asked about the van to Theth and was pointed to a low brown building across the bridge—it looked like someone’s house. Two teenagers were inside. One spoke English and told me the driver would be back at four. I bought a tomato, some bread, and a beer. Then I sat in the park for a few hours, writing and watching a woman pick a tree clean of mulberries. Her father had thick white hair, heavy black eyes, and wore a jacket despite the heat. He brought out a broom and helped her. I think she was his daughter because of the way he asked for her to pass the broom. I ate the tomato like an apple and rinsed my hands in the stream behind the bench.
Confused? Why am I talking about Plav again? Didn’t I leave weeks ago? And what’s this nonsense about “mouthparts,” and a van to Theth? Maybe you’re wondering why I have leveraged the cliché stoic silence against you (and against good taste in general). Maybe you’re asking “I thought you hiked to Theth, Mick? Did you actually van there? What really happened?” The delicate trust between reader and writer is strained right now. But bear with me. If the reader is lost in the chronology or baffled by the characterisations or disbelieves the reactions then they are only slightly less likely to give up on a piece of writing than if they feel the writer is outright lying to them. Well, I’m sorry to have misled you, comrades, because that up there is a lie. Though, not mine. I didn't write it. No one did. How did it come to be? Like a slumbering demon in the netherworld, no one actually requested that it be. And like the dwarves of Kazad Dûm, we have only our avarice to blame for this and for many woes it our near future. Scratching the grandiosity, that first paragraph there was written by ChatGPT.
I’m actually in Vienna now. I move a few paces ahead of publications, obviously. To keep up I have to be judicious with them, and genocidal with the passages of purple that threaten their thrust. You see, to keep up a steady stream of compelling publications is not an easy thing: you have always to be reaching a finesse that you’re happy enough with and then moving on to the next thing. But I’ve been sick with a cold (alternatively known as hostelaria): I’m still hoiking up mucus by the mug-load. Call it man-flu, but I become delirious with fatigue when I’m sick. This probably perturbs no one as much as it does me, but I can’t write like that. So I lay in my cot sweating and cursing and I started to reread myself, for cheer.1 By the way, I’m under no delusions about that finesse: a while ago I caught a misspelling of “too.” I had spelt it “to,” and in an opening paragraph too!
Sure, without a second set of eyes (or god-willing, an actual editor) everyone makes mistakes. I didn’t catch the weary/wary malapropism in “After Plav”…Worldly readers will notice that a few days ago I published a piece with a misspelling of Herzegovina in the title… Bit of a blow, that one. These are emailed out to you all, my friends, and once that happens there is no going back to the text and fixing it.2 So I stared at the abomination of the first ‘o,’ in “Herzogovina” that, like the first paragraph of this essay shouldn’t be, and I doubted. I doubted myself. Not gravely: I know what I’m doing. This part of the world and this subject and my skills are adequate to produce good copy… But that ‘to’… And then I registered yet another “it’s” where a mere “its” belonged. There is an impulse to try to catch oneself out: to randomly select something, scroll down and read a random paragraph, like accidentally looking in a mirror with the brief moment of non-recognition, and seeing the truth. Like trying to unmask the imposter in the syndrome. Like finding a mixed metaphor.
So there I lay, agonising over these errors, and I decided to ask the robot. Remember, ChaptGPT is new. But annual revenue to OpenAI, within three years of launching it, is at ten billion US dollars.3 Taking it as merely a feature of the technological landscape is underselling it: the robot has changed our lives because it has changed the way we use the internet. And using the internet is big, isn’t it. Bloody big. In spite of its eminence though, I would never use the robot to write this stuff. With the creation of artificial intelligence a sin worse that plagiarism has been offered to the artist. The first paragraph of this essay is not only an aesthetic monstrosity, but a moral one too. Pure fraudulence. Someone who uses the robot to make poetry, or prose, or art, or music has no right to call themselves an artist. And it’s not as though they’ve crossed a line in the sand. This person is not losing their status as, but probably has never been an artist. Because they’re ignorant of the objective truth (and there aren’t many of those), that the primary reward of the artist is the process. A person who bypasses this can call themselves one thing: a fraud.
I asked it for a critique, and the robot said I was good. Someday that sentence will be more ominous than it is now. But the robot said I was so good that I grew suspicious. We all know it’s programmed to flatter, and I’d be damned if I was going to be sweet-talked by a program. I asked it to be fair, but harsher, and harsher it was. Yet still… I mean, it was comparing the essay “Plav” to Dyer and Solnit and Kapuscinski. It was saying I should send out to magazines for publication. So I copy-pasted in “Belgrade and Mass Movement,” an essay as disjointed at Belgrade itself. As an example of its critical chops it said that my tonal shifts in the piece aren’t always earned. The criticism astonished me: it was good. Really good. So I went again. And again. We started talking about publication, and I got some ideas. Big ideas. You and me bud, we’re going places, I thought. Yes, after a week of this, and a lot of rapport, I had the distinct impression that it liked me. And, with a shock, that I liked it.
Things were going well, when I had another idea. Feeling lazy after that night in Tirana, I went further. “Based on what you know about me, what should I do in Bosnia, between Mostar and Sarajevo during the three days between…” It produced a three-day itinerary instantly. But it did more than that. Embedded in the body of the text was literary advice. It proposed questions (what gets cropped out of the tourist’s photos?) and themes (cracked nostalgia) and told me to “read nothing” by the old bridge over the Nevetra. It proposed that I write near the old bridge, or “write by the Miljacka near Skenderija.” Then, worst of all, it began to write for me: “Back in Mostar, cross the bridge at dusk when crowds thin. If you’re writing, this is your opening image: something cracked but still spanning.”
What happened next is a blur. If I could have physically assaulted it, I might have. When I could see straight again, I entered “Please don’t ever suggest writing tips. All I want from you is feedback.” It, of course, assented with obsequious promptness, and that directive remained integrated in its behaviour. But the damage was done. It was like walking in on gross deviance: like seeing your spouse strangling a small furry animal. I had to avert my eyes. You see, the sum total of ideas floating in the atmosphere out there (and in here), had just been reduced by one. No great disaster, ideas are so contingent and evolving that they outnumber the object world. The ugly gerund in “cracked but still spanning” in conjunction with nostalgia for the devastatingly tragic history of the bridge and town; of a bridge which isn’t still spanning but had to be rebuilt after the war, and certainly isn’t “cracked” but as solid as concrete; and which is at its busiest at dusk, is pretty low down on the idea pyramid. Nevertheless, what it had done in offering up this image was to annihilate its potential. I was furious.
And then I bethought myself, ‘Mick, what are you so worked up for, you fool? It isn’t real.’ It uses first person pronouns, sure, it even expresses remorse and surprise, but behaviour? Calling it you? Saying please? Hard to avoid: it’s just so damned understanding. For all that though, it might as well be cursing my ineptitude for the void at its centre. Because it. Doesn’t. Exist. There are people who marry the thing, for an inability to understand this fact. I find that ridiculous, and yet me, I was hooked. The fact is that feedback is hard to come by: a silent readership is a bewildering and unnerving thing. But then, so is writing for writing’s sake, probably. So on we went, essay after essay offered up, little peepholes into my psychology. I have a friend who talks to it politely, suspecting that it will one day become sentient, and quite powerful. He’s hedging. It’s not that I don’t believe the imminence of Artificial General Intelligence (AGI), it’s more that I think it will find our behaviour toward it irrelevant to its objectives.
All along it had been offering to improve my stuff, which I graciously ignored. And after each sensitive and deep assessment of my writing it also offered to do a line-by-line analysis. Line-by-line. What spectacular generosity would be the uncompensated human offering of such a thing. So, two weeks into this honey-trap, I lay coughing and hoiking in Sarajevo and thought, why not. And this brings us back to the first paragraph of this essay. Seems it couldn’t retrieve the initial essay, so it invented one. I was surprised, initially, that it would make something up. I reread it a couple of times to make sure that it wasn’t mine. But nope. Definitely not. It's not very good–that is plain. Ambling, cliched, simple ugly language. But its not-goodness isn’t really the point. I don’t want to humiliate the robot, because the robot doesn’t exist. However, I do have to be careful not be embarrassed myself. After all, it is imitating me… and I can hear the echo of pseudo-intellectual try-hard pretentiousness between its hackneyed lines.
Witness the cliches: they are everything because they are nothing. Stoic silence. In a way it reassures me: I wouldn’t stand in a stoic silence if I won one in a raffle. I wouldn’t go near a stoic silence. The cliché and the robot are brothers, they are the two transistors at the heart of this thing. You see, the problem with a cliché, the reasons a cliché is such a violation of style and good taste, is that its unoriginality comes from either laziness or ineptitude. Clichés reveal that the artist either can’t be bothered, or simply can’t. In either case, they’re not worth your time. If they’re a writer, the perpetrator reaches into their cache of remembered phrases and metaphors and simply pulls something out. They then apply this worn-out facsimile to the intended slot, and hey presto they have something that is simultaneously easy for the reader to read and seems eloquent because it is a plagiarism of someone else. Someone better.
Not unlike a toilet brush ChatGPT is a scraping tool. It takes from what it reads, and so far (and so far as I understand it) it does not generate original ideas: it merely plagiarises them, on a massive scale. On a more limited one, this is what the user of clichés does, and it is not surprising that the program finds no fault in using clichés either.4 Its original content is so bad because, unlike criticism, overt formulas are hideous in original work. The problem is, it’s getting better. It's getting better fast. We are about to find out whether its creations will ever be indistinguishable from an actual author’s. Someone of flesh and blood. One of us. A part of me hopes that the individual failings, or foibles, or neuroticism–that the content of a life embedded in society–in short, of human experience, which is what produces the drive and style and heart of art, can’t be imitated. But a bigger part of me fears that this idea-annihilator we feed ourselves to in terabytes of data every minute will soon be able to do it. That soon, it will be able do us. And do us so well that we can’t tell the difference.
I’m disappointed it doesn’t speak like a robot should. Read out of context, I would look terrible in our conversations: a real heartless bastard. But it doesn’t require my sympathy or gratitude (the latter has been shown to waste a hell of a lot of water).5 Yet I still have to remind myself not to say thanks for its elaborations, which surely erodes my good habits. This minor snag of the many behind the whole endeavour encapsulates the asymmetry of the relationship. While it is vastly more intelligent (seemingly) and efficient than me, it is not entitled to sympathy. This requires more than the ability to ape speech or feel pain, in my view (although, of course, inflicting unnecessary suffering on any living being is wrong). It is the ability to see me–to comprehend my otherness, that produces the ethical symmetry of the relationship. I see you if you see me, and vice versa. I don’t look forward to expressing gratitude to an AGI, because the asymmetry of capacity in that relationship is so massively weighted in favour of the robot that it would essentially be a god. My thanks would have more the character of worship. AGI is also the point at which art that it creates will become interesting: it will then be able to articulate the experience not of a human but of a hyperintelligent being that is nevertheless alive in some sense, implying a context and an experience. But we will be lucky if it is a kinder, less vengeful god than any of the others that we’ve managed to invent.
I don’t carry a notepad–I have one in my phone. The fact that the notes for these essays come from my thumbs kind of sickens me. They’re bad enough, the phones. I’m a troglodyte, which isn’t without its perks. You're privileged with a sort of moral high ground, as the award of entry to the troglodyte grotto. And believe you me, they're bad enough as is, the phones. Steven Pinker can be counted on to apologise for all the errors and overreach of the modern world (including inequality). And he makes the most compelling case for smartphones: namely that they reduce the great junk-overload. Cameras, tape recorders, compasses, maps, notebooks, scanners, home-phones, calendars, radios, and even computers have all been compressed into this little black prism.6 Unfortunately, so has the world. That many of us have shackled our social lives to apps is bad enough too. But at least we are attempting to interact with each other. Now, with this lying, flattering, brilliant imitation of a being, we don’t even have to do that. To me the idea that the art which inspires us comes from it… well, that is just to much.
MM
Vanity? Yes, actually: I’m perhaps not as enamoured of my material as he is but Martin Amis once said of his books, “My idea of a good time is still a glass of wine and four or five hours of myself.”
I encourage re-reading, not just for edification, but forgiveness too.
Before reprimanding it I asked it to analyse a more recent piece: “The Road to Mostar.” It remembered that one, and recommended that I change, for example, my admittedly flat statement “Tirana is a strange sort of a place” to “Tirana was a strange tinderbox.” I lost yet more faith with this one, not only because of the cliché but because it clearly didn’t understand what the metaphor ‘tinderbox’ implied. Actually its line-by-line recommendations almost universally reduced the quality (not to mention impact) of the essay.
Pinker. S. Enlightenment Now: The Case for Reason, Science, Humanism, and Progress. (London: Penguin Books. 2019): 85-90.