Oliver v. Kahn Part II

An AI Death Match

Dear Friends,

I’m continuing with Part II of this exchange with Substack legend

Henry Oliver

on AI and literature.

Best,

Sam

HO:

I don't know how you could possibly justify this statement: “Writing isn’t just putting words together, and it isn’t just some high-quality technical achievement. It’s a mainline to the soul.” Since when was writing a mainline to the soul? What does that even mean?

You are not describing what writing is; you are describing how writing makes you feel. One swallow does not a summer make! Remember the Turing Test is dead. We don’t always know what is AI and what is human. However good we are at telling the difference now, we will be worse in the future.

Shakespeare said, “The truest art is the most feigning.” We know nothing of Shakespeare's soul. His writing is not a mainline to anything. These declarations about writing are never as universal as they sound. The temptations of aphorism are to be avoided. Unless they are written by Oscar Wilde: “Lying, the telling of beautiful untrue things, is the proper aim of Art.”

As for your proposal to draw a line in the sand. Impossible! The fact that you offered this dialogue, despite the policy¹ of The Metropolitan Review, is an example of that. If literature is going to deal with the world as it is, AI is unavoidable. Cinema changed how the modernists wrote novels. Radio very much did affect writers. Think of how different dialogue is in post-radio novels! Think of the stories by Kipling and Cheever with radios in! Larkin wrote a poem with a radio in, too. Maybe they hated the radio. But they did not ignore it. Fictional characters watched TV. Think of the boy in Equus always singing an advertising jingle.

Plenty of writers worked in radio and television and had good careers. They might not otherwise have been "literary" writers, but so what? The few examples of literati being defeated by the system are merely that: a few examples. The Mousetrap was a radio play before it was a stage play. Helene Hanff wrote scripts for TV dramas. Was that so wrong?

You are muddling the literary and the non-literary and reasoning from too small a set of examples. Even if it is true that no serious writer will produce anything with AI (and I simply don't believe it, but we cannot predict), there is no serious future for literature that doesn't involve AI.

First, it will become inevitable for the bureaucracy and administration of The Metropolitan Review, it will be an essential research tool (have you seen the recent announcements from OpenAI....), it will be able to proof-read, offer robust editorial feedback, provide new lines of thought, give you a reading list... I say will be, it already can. Writers who cannot use AI will be left behind in more ways than the purely literary.

Second, AI is going to be part of everything. A novel without the internet already feels stale. This will get worse. How will realistic novels remain appealing if they keep feeling like they are set, technologically, in my childhood or earlier? Yes, that's what Middlemarch was, and War & Peace, but it's not what Pride and Prejudice was, or Jane Eyre, or The Way We Live Now. The novel incorporates. Dickens loved writing about the 1830s. Still, the train arrives in Dombey and Son. Ronald Firbank incorporated radio, telephone, cinema ... Waugh took that influence and used it well. Fiction made such splendid use of the telegram!

Third, I don't agree, at all, with what you say about technology destroying art forms. Figurative art was exhausted. Watercolour landscapes were still being painted in the late nineteenth century, well after photography was invented. They were dull. Turner made abstract paintings in the 1840s! Impressionism wasn't a mere tactic against technology: it was a clearer vision of a changing world. Art is not so crabbed or defensive as you think. It is a constant struggle for excellence. After a period of great discovery, modes and mediums become tired and repetitive. Shakespeare's sonnets are largely anti-sonnets: he is trying to do something original in a form that had been worked to death in the 1590s. Figurative painting peaked in the Dutch golden age. The longer your view, the more you can see this. Art is always working with innovation.

AI may or may not produce something great. It might be a new radio or a new Hollywood. We don't know. I can well imagine someone producing an entirely new sort of book, one which integrates writing, images, videos, commentaries, footnotes, live search, voice recognition, audio narration ... someone will surely produce a digital novel that blends and merges several mediums. It will be a whole new experience, akin to the time when photographs began to move and voices came out of the air. We are so inured to these great changes, we forget. We forget that before Edison, hearing voices was a sign of ghosts and madness. Seeing pictures move was a hallucination before celluloid. Remember the moving portraits in Harry Potter? AI will make that for us now. Imagine an edition of Harry Potter² that is like the book combined with the film in some new VR experience. You might recoil from this. You might say this is not reading. This is not a mainline to the soul. Join the leagues of people who hated impressionism, hated the novel, hated the theatre. On and on they go.

Imagine if you can use the AI to create the world of the book as you imagined it. Won’t writers use it for world-building? Won't they want to see their words live? I am just as attached to the book as you, but this is the world we live in now.

Someone is going to try and use this new technology in a process of discovery. They are going to do what artists have always done. When we discovered perspective, we got the Renaissance. The financial crash of 1825 started us on the periodical novel. Who knows what's coming now? If the current writers don't do it, they might not be so current ...

Your view of art is a kind of degraded Romanticism. It is a particular thing from a particular time. It might not be true anymore. I doubt it was ever really true, not quite as true as we wanted it to be. And I say this as an individualist. You are sceptical of the market forces that gave us television. Market forces also gave us Mozart! It's true, there is no FirstFolio of HBO. I am a Netflix sceptic. (Yes, I watch, I just don't admire.) Still, the people who wrote those shows had good careers as writers. Not my sort of writer, but who cares? We never get given that many geniuses. When was the Golden Age for writers? They complain in every generation.

Whether AI produces great art or not depends on many other factors. Art relies on the innovation of ideas as well as technology. It needs a receptive and demanding audience, maybe a fastidious one. It requires a market, maybe a flourishing polis, or at least a lively one. It often does well in times of moral rearrangement. Because it is a response to the past, it is made in the crucible of present change. (I wrote about this here.)

AI is that change for us. It might be terrible, it might not be to our taste, it might not benefit us or people like us. It still might produce great art. It certainly will have to be incorporated into many, many aspects of literature.

I am only a critic, a self-proclaimed critic at that.³ It doesn't matter what I believe or what I want. What I shall do is pay attention. Jane Austen (1775-1817) lived through wars, revolutions, the first era of fast fashion, and a time of new ideas in moral philosophy and economics. She was part of a society that looked different, dressed differently, had more wealth, new machines, new shops, more social mobility, new ideas about economics and virtue. She could have stuck to the old ways. She was, after all, a conservative. But she read Adam Smith. She knew about the social upheavals of the war. She was interested in abolition. Her letters show a fascination with the new textiles. Her novels are all about how social life changes as the world changes.

She watched and she wrote. That's what we need to do with our own stirring world.

SK:

Hi Henry,

We’re a bit talking past each other — it’s a surprise to me how much we disagree! — so I’ll try to be a little more down-to-earth in my last post and to be clearer with my terms.

We seem to be talking about three different components of the AI revolution and it’s important to distinguish between them. There’s 1) the content of a story and whether or not it includes whatever the cutting-edge technology of the moment happens to be. At least from my point of view, this is a fairly uncontroversial subject. Insofar as artists have any kind of ethical obligation, it’s to render whatever the emotional content is of their time — and that can be ‘timeless’ themes (e.g. how does love play out at our moment in time) or can be specific to the technology they come across. That’s really up to the artist. If somebody writes something interesting about AI, in the way that Patricia Lockwood wrote something interesting about Twitter, then we’re⁴ happy to cover it and give it its due, but we (or I) don’t particularly think there’s any obligation among artists to be, like, technological early adopters and to zero in on whatever the current gizmo happens to be. If the dirty word you throw at me is “degraded Romanticism” (why ‘degraded’ by the way, maybe it’s just “Romanticism”😊?), the dirty word I’d throw back at you is “technodeterminist.” Were the best books of the ‘90s about AOL Instant Messenger? The best books of the ‘60s about fax machines? Artists can engage with whatever they find interesting. The technology of the time may make inroads into our emotional lives, but I don’t think it’s necessarily inevitable that we are shaped by it. We do have say, and choice, in the degree to which we allow technology to dominate our lives.

Then there’s 2) the use of AI as a research tool. This is also fairly uncontroversial. If it works for you, it works for you — and AI can be seen as just being part of a succession of other tools (Google or the microfilm library, or whatever). I don’t have any particular objection to employing AI as an assistant. (I myself have, in the time between my last post and this one, been talked into using some sort of AI tool to catch typos for The Metropolitan Review.) At a personal level, I’m just wary of it because, in my lifetime, I’ve seen what smartphones did to the society — the way they captured everybody’s attention spans — and I’m acutely aware of how a “tool” can go on to manipulate its user.

But the real point here is 3) the use of AI to actually generate the “main body” of written content. And this is, I believe, categorically different from any other use of AI or anything else that has ever existed in writing. The only real analogy is that AI may soon be to writing as the 19th century camera was to figurative imagery – with a machine taking over the core function of an activity (doing it with greater technical proficiency than a human being could possibly manage) and in so doing profoundly reshaping our understanding of that activity. If we allow AI to write for us — not help us in our research, not suggest directions but actually write (which is what all sorts of trade publications are in the process of doing, laying off their staff and “welcoming in” AI to generate the content) — then we are giving over our agency to a machine and in the most profound way we ourselves are no longer engaged in the activity, we become a spectator and consumer in our own lives and no longer derive the deep meaning and satisfaction that the activity can offer us. I regard the advent of AI in this sense to be an existential threat to the activity of writing, and I believe that it’s fairly easy to draw the line in the sand here. It’s the difference between somebody playing the game and watching the game. The person playing that game might not be the world’s greatest at it, but at least they get to play. Turning that agency over to AI — and there’s about to be abundant pressure to do so in all walks of life — is to actively participate in our own obsolescence.

As for what I mean about the “mainline to the soul,” I guess we first have to agree on what we mean by the word “soul.” I don’t mean anything explicitly “religious” or “mystical.” I just mean the interior domain of a person’s being that acts as an analogue to the external impressions they receive throughout their lives — their “inner life,” to put it more simply. When it comes to Shakespeare, we may not know that much about his biography, but we know an enormous amount about his soul (or the soul of whoever wrote the plays …. I’m very slightly a Baconian⁵ but we can have that mud fight some other time). We know his meditations on love, death, power, friendship, you name it, and know it — because he had a unique ability to put words to his reflections — better than we do the soul of just about anybody else. (Let’s not be hung up here, by the way, on questions of literalism — speaking imaginatively or fantastically is just as much, if not more, a reflection of the workings of the soul than the lived “realistic” experiences of one’s external life.)

When I sit down to read something, what I am interested in is the honest truth (however fantastical or whimsical or distorted) of that person’s experience of the world. Since AI has no experience of the world, and no individuality, I am categorically uninterested in what the AI has to say, however artful its magic trick of feigning individuality and experience. Where it gets more complicated, I suppose, is if we think about the AI not as a machine but as a collective intelligence — as the entirety of the Internet (all those pixels) combining together to speak in a single Frankensteinian voice. Is that interesting? Maybe at some level (DeLillo's "the future belongs to crowds" comes to mind) but we very much get a choice in who we extend a hearing to, and in whether we opt in to our own agency or surrender it. If we're listening to (or reading) AI, that means that we're at that given moment choosing not to listen to the experience of an individual human. If we turn over our own creative functions to AI, that means we're surrendering our own agency — and, from the perspective of the soul (i.e. from the perspective of our lifelong endeavor to express and to understand ourselves), that self-surrender is indefensible.

-Sam

HO:

I think you are right about a lot of this, and we probably disagree less than it appears. I am not, for example, a techno-determinist. Merely a pragmatist.

Let me put it like this. If, say, Catherine Lacey decides that she will never touch an AI and that it would be detrimental to her art, I think that is a fine and sensible decision that may well work-out very well for her. If the entire group of people who currently constitute the literati make that decision, it will not be so smart. One day soon, you will read something that moves you, and you will not know, or be able to know, if a human wrote it.

I was playing with an LLM recently trying to make it write a half decent poem. It gave me one line, in a poem set in the American Civil War, describing a letter box at the end of a lawn as a "sentinel of hope", which, honestly, is a good image. If you saw it out of context, you'd likely fall for it. That model is now out of date.

What we disagree on is this:"When I sit down to read something, what I am interested in is the honest truth (however fantastical or whimsical or distorted) of that person's experience of the world." Who ever said this before AI was invented? It became a meme in the last year or two. But it's not true, or not entirely true. When you read, you want your sympathies excited and your mind engaged. You want beauty, truth, wit, excitement, character, learning, feeling, plot, trope, inventive imagery, and so on. It is often the case that writers transmogrify their own experiences, and often the case that they do not. You often cannot tell. (This is one of the great lessons of literary biography.)

Great writing can be engaged in saying the same thing again and again in different ways (as I quoted Johnson saying earlier).⁶ Maybe we are going to witness a return of this form of writing, that prizes tradition? You do not read Agatha Christie for her "experience of the world" but for her ingenious plotting. You cannot know Shakespeare's experience of the world: he seems to have written about almost every experience of the world. Literature is a much higher art than this: it is a means of thinking, not just a means of reporting. It is an act of imagination, not just of experience. A novel like The Fountain Overflows does very much give you Rebecca West's experience of the world, but in a transformed way, so that she also gives you the experiences (or sympathies with those experiences) of other people. One does not read Proust merely to know what it was like to be Proust, and irrespective of his style.

This idea which is, I think, a defence against the possible problems of AI (which I am not as uncomprehending of as I perhaps appear) would be totally inadequate to explain the nineteenth century novel. The theories of Bakhtin would have to be thrown out if it were true! This is what I mean by a degraded Romanticism, (which was not intended as an insult!). To degrade is to put out of office (de-grade), and I think what has happened is that the core ideal of Romanticism has been put out of office in the current discourse that is looking to find a defence against AI ever being able to write a great poem or novel. It was never really the Romantic ideal only to make a self-expression. Indeed, the great Romantic image of the Aeolian harp might give you pause about the idea of "surrendering agency".

I happen to agree that AI's lack of experience of the world — a lack of personality — may be what is currently stopping it from writing great literature, but this is only speculation. The models are improving so much...

I have said more than enough, again, so let me end by thanking you for this discussion. It has been stimulating! I hope we can talk more as the models change and the reality we are discussing has become new again.

SK:

Enjoyed it Henry. Thanks for doing this!

1. Again, there’s no actual “policy,” but we did bash AI in our manifesto in addition to more generally bashing corporate slop.

2. I hate Harry Potter so this particular argument is lost on me.

3. You’re definitely a critic for me, Henry!

4. i.e. The Metropolitan Review

5. So sorry Julianne Werlin

6. Henry, you’re just too charming whenever you do this