Oliver v. Kahn Part I

An AI Death Match

Dear Friends,

I’m pleased to share the following exchange on AI. The maximally-British

Henry Oliver

is one of my favorite Substackers — sort of the crown prince of the Substack literary community. If you told me that Henry — whom I will only ever picture sitting in front of a fireplace in Herefordshire and stroking his cat Arthur as he reads Middlemarch for the fifth time — didn’t own a cell phone, I wouldn’t have been at all surprised. But Henry leapt to the defense of AI when it was (somewhat obliquely) castigated in the founding manifesto of

The Metropolitan Review

and we ended up having this exchange, which I’m running in two parts. I’m very happy to think this through, since there’s (pretty obviously) no greater question in writing, or anything else, at the moment than the advent of AI.

SK:

Hi Henry,

Nice to have this exchange!

We can start with AI, and start with the exchange we’ve already had, and branch out from there. I imagine the terrain we'll really want to get to is the sort of moral obligation of literature to keep pace with the times (which is something that I certainly have mixed feelings about).

So — for readers joining us — the jumping off point for this is this fiery manifesto that the editors wrote up for The Metropolitan Review, in which we denounced AI (in addition to denouncing many other things). You wrote on your Substack that you liked the general idea of the publication but that something struck you as willfully atavistic about our stance here.¹ “But the most significant thing happening to writing and culture right now is Artificial Intelligence,” you wrote. “It is changing everything. Suddenly [by rejecting AI], they [The Metropolitan Review] are not racing to the future anymore.”

So, first of all, I think of you as such a classicist² that I was surprised to see you leap to AI's defense! Do you want to lay out how you think the literary community should view AI and where there can be some sort of harmonious interaction between literature and whatever is coming out of AI?

All best!

- Sam

HO:

Hello Sam,

Many thanks for suggesting this! I am always interested to have these discussions. Let me start by reiterating my enthusiasm for your attitude at The Metropolitan Review, which is not only one of opposition to the dreariness of much modern culture, but a willingness to do something new in response.

I will try and answer your question, but I want to fill in a few relevant points that come prior to what you asked.

I am a classicist in the sense that I promote the value of old books, but not, alas, in the sense that I can read Latin and Greek. So in the longer view, I am a modern. (Obviously, I venerate the ancients.) The old debates, from the time of Shakespeare and running through Swift and Johnson, were about the ancient classics versus the modern classics. It's a debate we still have: traditional literature or modernism and its inheritors? Like Swift, I want to take an ambivalent position. I like both. I want both. I choose abundance. I love Shakespeare and Helen DeWitt.

One reason why we are able to have the luxury of this discussion is technology. Without the printing press, there would be no First Folio, no Paradise Lost. As John Pistelli so lucidly discussed (I would give my fingernails to be able to write like that), many other technologies, such as the typewriter, have been instrumental to the production of literature. Of course, yes, there would still be literature without these technologies, but not the same sort of literature.

Art is often the result of technology: Shakespeare and the indoor theatre, Hollywood and celluloid. The essence of poetry, said Samuel Johnson, is invention. He was describing the way poets found new ways to say old things, but he knew full well that in the modern world (the world of innovation in technology) that has meant new ways of saying new things as well. He saw the rise of the novel, which is still shifting and changing as a form.

The biggest change technology has inflicted upon literature was the invention of the radio. That produced a break with the past unlike anything else. Voices in the air! From there on, entertainment technology became an alternative to print: film, television, videogames, computers, the internet, social media, and smartphones. AI is the latest in that line.

Television was clearly the worst of these inventions. I cannot request anything from television: no Shakespeare, no Mozart, no Rembrandt. Not on demand. But now? I am listening to Mitsuko Uchida as I type this! When I write about Shakespeare, I can summon up the texts, in various formats and editions. I have been released from the grim mono-cultural dross of the 1990s into a world of art. I saw Uchida play live last night. And why? Because I saw the details on Twitter. No social media, no Metropolitan Review, in form or in ideology.

AI is a much broader, more general technology than any of these, but one of its most important functions is to re-organise the internet. I asked GPT o1 recently (not pro, alas, I am too poor) whether the 1980s were an especially good time for Mozart recordings. In my follow up question, I told it two of my favorite recordings from that period and asked for others. Both answers were splendid. With other models (including o3, Perplexity, and r1, though I have stopped using that for now), I can get links to supporting sites. Rather than thinking of AI as some glib predictive word-arranger, I see it as a super-Google, a much more advanced means of calling up information from the internet. I listened to several excellent Mozart recordings as a result of that one query. Ask it about Bakhtin: you get a decent result.

Now, this is just one example. AI is going to do (is already doing) so much more. It is being used to help scientists understand how animals communicate with each other. It is helping historians read old documents. It is increasingly able to mark homework accurately and give helpful feedback. (Daisy Christodoulou is changing her mind.) It makes management consultants more productive. It is being used to make medical research quicker, too. We are going to see the development of more drugs that save and improve our lives. The quicker we get those drugs, the more people we help. AI is also an excellent coder, and it helps us repair potholes. Students can learn twice as much with an AI tutor than in some classes.

Now, AI is not all good news. As I wrote, every technology is a Faustian pact. The printing press unleashed all sorts of disorder onto the world! But when I see people say that AI is all nonsense, I assume they are not reading the right sources. If you read opinion columnists, including here on Substack, you'll see a lot about their opinions on what the media are saying about what the tech companies are saying, but a bit less on whether AI will help us cure dementia. The idea that the primary use is "scams" is obtuse. Read Rohit instead!

So what I want is to see literary people take AI seriously. Writing it off, ignoring it, assuming it is a slop machine and little more, being a lot of Cassandras about it, is a narrow, badly informed, and false view of what is going on. Many people still don't like the internet, but the world is what it is. We can only live in the times we have. Why is it that when I see pieces like this one (recommended: George Eliot and tech) they are not often coming from the literati?

The world is going to change. The way we work will change. The way we manage our health will change. Education, politics, war: everything. How can literature ignore this?

I have been a teaching assistant, law-firm blogger, Parliamentary research (bag carrier), and marketing consultant with large corporate clients. In any of those jobs, I would find AI hugely useful, interesting, and unignorable. One of the first things I tested it on when GPT3.5 came out were the sorts of questions I would have used in my old consulting role, and, you know, it gave good answers!

In the literary world, AI is not so obviously applicable. There is understandable hostility to the idea of an AI novel. (Though I think people are quite confused about those issues.) But we still have to think about it! I will say again: it's going to be part of everything. Hollis Robbins has just written an excellent piece about how we ought to re-think the way we teach literature in an AI world. Hollis has been thinking about this for a while, as have I. I had been interested in AI for several years before GPT 3.5 arrived. People back then would say, “oh but chatbots are useless”. They were right. I would tell them that car insurance claims were processed by AI, or that half of the work for Goldman Sachs to bring a company to IPO was done by AI. They were reluctant to follow the implications. But here we are. (I have previously written about AI and criticism, though even in this short while it has improved,

So my piece was simply trying to say that ignoring AI, as you plan to do at The Metropolitan Review, isn't just misguided — I suspect it isn’t possible. It reminds me of people telling me chatbots were useless. AI might produce new art, or it might simply become so pervasive that for literature to ignore it would be like trying to write novels without trains, cars, or electric lights.

As for how AI and literature can interact... well, I feel like I have already gone on for long enough!

SK:

Nice points Henry. As stylish as I would expect from you!

So it’s a little interesting for me to be cast as the atavist in one of these conversations (and this against someone who can fluently quote Samuel Johnson!) since, in other skirmishes elsewhere on the platform, I get accused more often of being a techno-optimist (which mostly has to do with my being so irritatingly bullish about Substack).

I fundamentally agree with you that literature has some obligation to keep pace with the times, and for its borders to be permeable enough to allow in both the cutting-edge technology of the moment and (more crucially) to take into account how that technology is reshaping the culture. No argument there.

There are two ways in which this gets more complicated.

One way is that sometimes technologies come along that are of no particular use to literature. You named radio and television, which are both perfect examples of that. Both would seem to be conducive to good writing, but the market realities of radio and television at their peak (the need for near-constant ad breaks, control by corporations endlessly looking for profits, the low attention-span of most consumers, who were, after all, driving or cooking or in the middle of doing something else) combined to make their value for writers close to nil. There were people who could adapt to the form — Larry David figured out how to make network television work for him, for instance — and there were the moments when the form adapted to let talented writers operate with freedom, as with Garrison Keillor or Jean Shepherd’s radio shows for instance, or as happened in a big way once streaming came in and effectively replaced the old commercial-driven television model. But for close to a full century, in which radio and television utterly dominated mass media, serious writers were nonetheless well-advised to give the new forms a wide berth, to stick to mediums that allowed them creative control and artistic freedom. For writers who went the other way and tried to wholeheartedly embrace the new forms under the principle that new had to be good, they very often ended up as cautionary tales. That’s, for instance, what happened to Clifford Odets, who dropped his career as America’s leading playwright to be buried alive in a Hollywood studio. That story is memorialized in the Coen Brothers’ Barton Fink, but the real-life version of it is even more macabre with Odets shouting from his deathbed that he might still write some play that would “redeem the last sixteen wasted years.”

The other way this gets more complicated is that there are some technologies that — I believe — represent existential threats to a discipline. AI may well be one of these. Photography is a perfect example of a technology that represented an existential threat to figurative painting — and, in a very real sense, painting has never recovered from photography’s advent (certainly, painting, and visual art in general, have nowhere close to the cultural centrality that they did before the camera’s invention). Faced with an existential threat like photography, the art world — I’m talking about figures like Van Gogh and Picasso — did the intelligent thing. They figured out how to draw a line in the sand, to cede to photography everything that photography was capable of doing (for a long time it was passé for painters to engage in representational art at all) and to identify the real core of their activity, which was self-expression, which was something that no camera, however advanced, could replicate. Theater (and to some extent the traditional novel) faced a similar existential threat with the arrival of film, and theater responded in the best way it could — by emphasizing the live-ness of the performance, by moving away from the kind of garish spectacles that were very popular in the 19th century and that were soon transported over to the new medium. Writing was less affected by the invention of the camera, but, now, a new machine is showing up that can put together words with intelligence, that can replicate most of the core functions of a writer. When a beast like that appears, the intelligent thing, usually, is to RUN — to draw the line in the sand, to concede various kinds of technical writing to the machine, but to really try to delineate what the core vital qualities of writing are (e.g. soulful expression, mementoes of lived experience) and to keep the machine as far away from them as possible.

The next question is (as we watch warily from the shrubbery) what the nature of the new beast is. I think you’re right that ultimately — or at least in its current incarnation — it’s basically a super-google, just presenting web searches in more easily digestible format. That’s not so dangerous, and writers may find themselves using it as we currently use google — as an assistant, as a way of quickly looking up relevant information. But, as I think we all know, AI’s impact is soon going to be greater than that. I think the real disruption is going to occur in film, where there are so many barriers to entry, and AI will slash through many of them. AI will eliminate most of the costs associated with making films and give complete outsiders, intelligently using prompts, the ability to make fully functional films that can compete with what the studios are putting out. From a democratic perspective, it’s hard to argue with that — since it allows people who otherwise wouldn’t have the capacity to express themselves to find an outlet for their expression. The case is different, however, with writing. There, the only barrier to entry is a paper and pencil (or word processor), so it’s difficult to think of AI in writing as anything other than cheating, i.e. as moving away from self-expression and letting the automatic pilot fly the plane.

Now if we think about where AI is likely to enter into writing, the obvious place, I suppose, is to help writers through difficult patches. Like you, I’m an Eisenhower man. I was writing a novel about Eisenhower and that meant that I had to write out a bunch of battle scenes, and that meant that I had to spend months doing a lot of pretty technical research to feel like I was able to inhabit Eisenhower’s world. I could easily imagine — this may be possible already — that I could have spared myself all of that by telling the AI to write the battle scenes for me and then to concentrate on the rest of the narrative, which was in any case more important. I imagine that that sort of temptation will soon be overwhelming for writers, but I think it’s very important that we draw the line in the sand somewhere around here. I wasn’t writing about Eisenhower because I wanted to accurately render Eisenhower’s life or even to write a great novel about Eisenhower, it was because there was something about Eisenhower that, strange as it sounds, was important for me and I wanted to explore that. To let AI directly into the process would be to, fundamentally, corrode what the process is about, which is self-exploration. In other activities – activities that are directly affected by AI –- we are able to draw the line in the sand without too much trouble. Chess players, for instance, can play against computers or study with computers to their heart’s content (and, on the whole, it makes them better chess players), but access to computers is strictly banned in live play — for the perfectly good reason that chess is, ultimately, a game, a test of skill and will against an opponent, and the second a computer shows up with its answer key of the best move in a particular position, the game stops being fun. Writing doesn’t have rules in the same way that chess does, but I think that we in the literary community (you should picture me sniffing haughtily as I say this) are able to agree on some parameters, and that AI actually helps to think us about what writing really is, in the way that the camera helped early 20th century painters to think about art. 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 AI doesn’t have a soul (unless DeepSeek, or whatever the fuck, is working on that), AI, however good it gets, isn’t capable of the crucial work of writing, and it falls to writers and (sniff, sniff) the literary community to draw the line in the sand here, to rule AI out of bounds and to focus on what is really important, which is our own passion and our own enrichment.

I’ll run Part Two on Sunday.

1. I should emphasize that it’s not totally clear The Metropolitan Review actually has a “stance” on AI. The line in question was about comparing AI with a more general dose of “cultural slop.”

2. Clearly a bunch of projections are going on vis-à-vis Henry.