← New Books Network
Loading summary
Transcript198 lines
- [00:01]
A
Welcome to the New Books Network.
- [00:05]
B
Hello everyone, I'm Paul Starabin and welcome to America and the New Books Network. I'm thrilled to have as my guest today Mark McGurl. He is the author of the book, the program Post War Fiction and the Rise of Creative Writing, also of Everything and Less the Novel in the Age of Amazon and he is a literary. A professor at Stanford now still. Is that right, Mark? Yes, and he is. We are going to discuss AI and fiction, which is a topic that can go in so many different directions. I'm really, I'm welcome, Mark. And I just, I'll throw it open. I'm happy to hear whatever you have to say on this sort of ripe subject.
- [00:58]
A
Great. Well, it's so good to be speaking with you about this. The first thing I should say is that, you know, we're, we're standing on really an uneven ground. I mean it's just things are moving so quickly. It's hard, it's almost nerve wracking to say anything because from day to day the new evidence we get for what the situation is changes. And I'm just dispositionally, you know, averse to sort of simple orientations really to anything. And AI is very, it's very tempting to sort of stake out, to get on the barricades and stake out a strongly negative, you know, relation to it. But I um, but like anyone, I sort of begin with the, with by noticing how you. That the, the political economy of AI is a disaster. It's an, it's a social inequality machine. It's a disaster for sustain this, sustaining our relation to the environment. All of these things are like the basic realities that are kind of undeniable. And then, and then you get to the subset of the matter which is the question of like oh, what is this going to mean for creativity and literature and questions like that. It's a small subset of the question, but it's pretty fascinating. One, one about which I try to, you know, I try to I don't want to say have an open mind in the sense like I'm receptive to it being all great. I just have an open mind to at least sort of observing what is kind of interesting about this moment as well as maybe troubling about it.
- [02:59]
B
What strikes you as interesting?
- [03:01]
A
Oh well, I mean start with, I mean start with the technology at the center of this generative AI LLMs as a technological fact. Anyone who has recently spent time with a chatbot, with Claude in particular. But any of them, how one can't find that A bit fascinating is beyond me. And like. And the reflections that it inspires about what it is possibly teaching us, not about the specific thing we're asking about, but about language itself and about human language use and creativity and all of these things that we may have harbored, it turns out, possibly too simple, attitudes toward. And so just start with the texture of the experience of interacting with an AI. I find it fascinating enough so that I'm not merely repelled. I want to. I find myself wanting to engage and to try to understand this, even as I know that out there on another level, there are so many troubling things about what it, you know, what its implications are economically, socially, et cetera.
- [04:37]
B
Right. Well, does it interest you as a writer?
- [04:42]
A
I haven't used it as a writer and don't know if I will. It's interesting. We were talking a little bit before we got online about Olga Takaruk, a Nobel Prize winning novelist who recently in an interview was talking about how she uses AI and the outrage that she has been met with is, Is, is. Is. Is astounding.
- [05:11]
B
Just remind us, so how, so how does she, how does she.
- [05:14]
A
She uses it, as I gathered from a interview translated from the Polish, the. The key bits are. So she was writing a novel, it's a historical novel by a few decades. And she asked a chatbot, what would my characters, what songs would they have been listening to? Oh, and it's interesting because that seems like fairly close to Google. Like, what were the big songs in 1978?
- [05:41]
B
I think so.
- [05:42]
A
But the chatbot presumably knew her characters to some degree. So your characters would be listening to these songs, not these other ones. So it's a little bit different from Google. And I think most, a lot of people would not be deeply troubled by that use of the technology. But then she goes on and I think she says that she asked it to edit some of her writing. And then that sets off like, wait a minute, you are. Now you've crossed the line. You are ceding creativity to the machine. And there are plenty of people in my, in my feed who are like, I will never read another word of this author again because she has betrayed
- [06:26]
B
humanity in a sound way. Did you have like a visceral reaction yourself?
- [06:34]
A
Certainly not to that degree. More like, it was more like, wow, okay, how do I think about that move? Yeah, I have trouble getting outraged about using a chatbot as super Google or super Wikipedia. It just seems kind of interesting and it just makes the process of search a little bit more efficient. Right. Having it edit Your prose. I don't know. It's pretty tricky. I mean, I guess what I'm assuming she would say, although I'd have to read the whole interview closely, is that the author, her authority has sort of stepped back, and she's obviously going to assess the editing job that the AI does. And so she's a little bit like an author editor figure. And I guess that's an interesting sort of wrinkle and our conception of what an editor might be going forward.
- [07:32]
B
Yeah. When you said the word edit, I was thinking in my mind, like, what. You know, what exactly does it mean to edit or to have an editor? I mean, it can be quite a number of different things.
- [07:48]
A
No, absolutely.
- [07:50]
B
Yeah. And in fiction, I think we know that. I mean, you talked quite a bit about this in the program area. The whole thing of the workshop experience. I think it was. Ken Kesey. You had an interesting example there of how, despite what people might think about how he wrote his novels, in fact, there was quite a bit of input from the group.
- [08:18]
A
Oh, yeah.
- [08:19]
B
So is that not an editing function of some? I mean, what if you had five people around the table and one of them was an AI chat? But, you know, I mean, I don't know. We're now into philosophy, maybe, but.
- [08:35]
A
Yeah, you know, but it's tricky. I mean, in a way, you know, this issue has been with us for much longer than the rise of AI. So for. Just to take a classic, canonical example, Pound's editing of the Way of Eliot's Wasteland.
- [08:50]
B
Okay.
- [08:51]
A
The great poem of 1922. Ezra Pound did a incredibly aggressive editing job on the manuscript of that poem.
- [09:00]
B
Really? I didn't know it was aggressive.
- [09:02]
A
Oh, yeah. No. Like, one section of the poem, he cut. He cut one section of the poem in half. Wow. He made changes throughout. And, you know, so. But. And people have just sort of made their peace with that kind of like. But nonetheless, T.S. eliot is still the author of the Wasteland because he agreed to what Pound had done to his poem. So you go from there to. You're right. A lot of the anxiety surrounding the writing program is that it'll make writers unoriginal because they're sitting there listening to their peers, imitating their peers, and not becoming these sort of individual creative voices that they should be. I think that that's too simplistic. I think that writers in that situation. It's not that. That's not true, but they're also differentiating themselves from their peers at the same time. That's how the game is played. You can't simply do exactly what somebody else does. You've got to look for your own thing. You've got to find your own voice, as they say. But certainly the anxieties are the same. And you know, the anxieties are the same surrounding our thoughts about genre fiction, so called formulaic genre fiction, which make. Which you know, some people find depressingly mechanical in the way that it sort of reduces creativity to following a recipe. And then now AI is sort of the magnified version of these anxieties that have been with us for quite a while, really with the rise of industrial modernity and its capacities for. Incredible capacities for repetition and giving us more of something we want. And then AI is really just the shifting of all of that into the domain of language use in a really profound and in many ways profoundly disturbing way.
- [10:57]
B
People seem to worry that they might not be able to detect the voice of AI and that that would reveal something maybe deficient in them. Since, you know, we readers kind of pride ourselves on, oh yeah, that's Dickens, nobody sounds like him. Or the, you know, magisterial voice of a Tolstoy or something like that. And then, you know, and what if the AI bot can fools us? Yeah. What does that say about. I don't know.
- [11:36]
A
No, it's interesting. I've noticed. I've noticed that ability to detect AI written things has become kind of like a point of pride, actually.
- [11:46]
B
Yeah, there you go.
- [11:47]
A
And people talk about this sort of earmarks of LLM speak, at least as of this moment. Of course, a year from now, things may have evolved. The technology might have evolved, but, you know, it uses EM dashes more than most writers do. It has certain kinds of linguistic constructions and people at least purport to be able to sort of see it, know it when they see it. Whether that's true for most people or whether that'll be true for people going forward is an open question, I think.
- [12:18]
B
Well, I think, I mean, Dunk college professors are trying to.
- [12:25]
A
Yeah, no, that's our daily reality is trying to rethink the entire structure of assignments, writing assignments in particular, in the face of this technology, which can do most of what students have done from time immemorial when they write an essay. And we can't be satisfied with them just punching out an essay in five minutes with an AI. So we have to rethink what we're actually going to ask them to do.
- [12:53]
B
Yeah, I mean, I suppose there's no turning back. I mean, if you kind of put on your hat as the sort of how the like structural sort of slash economic trends proceed in the publishing business. I mean, is there anything that could possibly work against the, you know, further inroads by AI?
- [13:18]
A
It's tricky. I mean, on the one hand, my first thought is skepticism of pure refusal. You know, you just have to look at the example of, say, consumerism. It's incredibly difficult to convince people not to buy that next thing, that they might not actually need. That desire for the next purchase is incredibly powerful on the one hand. So you almost have to be like, okay, if we can convince people that AI is sort of like smoking, smoking cigarettes, then you're like, okay. At least over the last, mind you, it took 50 years. But over 50 years, sure enough, smoking rates have precipitously declined because enough people have gotten the message, particularly young people, never to start smoking in the first place. And those who did find ways to quit because they know it's killing them. Is AI smoking? Is it cigarettes?
- [14:23]
B
Or
- [14:25]
A
is it the thing that makes your life easier by, by, you know, by, by speeding up routine processes or work that you don't really want to do? It's really not clear to me whether we're going to convince people that it's act that this is a cigarette. This is destroying your mind. In some cases, though, you surely want to try, especially in the easy cases like young kids, you know, who, I think we, we will be able to convince ourselves that that's a bad idea. To let children educate themselves or spend too much time with this technology, which sort of just offloads too much of the task of thinking for oneself.
- [15:12]
B
Yeah, and possibly like imagination.
- [15:16]
A
Yeah, that too. That too, absolutely. Yeah. So that's, that's, that when you look at it in those terms, it's pretty troubling. But you know, I have a healthy respect. I just, I wrote a book on Amazon which is founded on the idea that, you know, the consumer desire is insatiable on some level. Our, our standards for what having enough is, have risen steadily for the last few hundred years. And, and Amazon rides on that fact. We, we, we want and want and want and want more and more. Enough. It never seems to occur to us that we have enough. And so that dynamic. Is AI going to participate in that dynamic as we sort of, you know, use it and it makes certain aspects of life easier? I don't know. Or are we going to be like, no, this stuff is poison. I'm not going to touch it. I don't know which of those. You know, it's, it's an interesting, it's an interesting moment because I think there was a lot of uncertainty about where we're going. Certainly all of the very powerful economic forces of our time and political forces of our time are in the direction of. You must use this in the corporate. And in corporate workplaces, they're mandating.
- [16:37]
B
Oh, yeah.
- [16:38]
A
That employees use it. And so if you start there, then it's pretty easy to wind yourself into a pretty pessimistic position pretty quickly.
- [16:49]
B
Yeah, yeah. Well, when you mentioned Amazon, I mean, my wife and I just took a road trip from Cape Cod, Massachusetts to down to the Philadelphia area and back. And I would say it seemed like every other truck was this Amazon Prime. It was just like, you know, I hadn't done one of these trips in a while and it was, we were both like, really, you know, they have their own fleet and. No, they're everywhere, you know, they're everywhere.
- [17:14]
A
They are. Yeah. It's an extraordinary phenomenon. And it didn't exist 30 years ago. It's incredible.
- [17:20]
B
Didn't exist 30 years ago. So, Yeah, I try not to be myself too cynical about it, but, but I hear what you're saying about this, you know, never enough so.
- [17:32]
A
Yeah. And like, you know, the, the, you know, the, the college kid who. It's, it's, it's 1:00am the paper's due.
- [17:40]
B
Yeah.
- [17:41]
A
You know what I mean? Like, you know, it's just like people are finding themselves in situations where they're, you know, where their ideals that they may or may not even hold because they may be generationally different, dispositionally different, but even if they do hold the ideal of like, I'm not, I'm not for this stuff, I want to boo this stuff. You put them in a certain situation where they're under duress and this tool can help you not fail the class. You know what I mean? It's a really tough.
- [18:09]
B
Yeah.
- [18:10]
A
Thing to ponder.
- [18:11]
B
Well, when it comes to fiction, some of the standards are set by the little literary journals that seem to. They get a lot of stuff right and publish very little of it. And so you wonder if they could make a stand.
- [18:30]
A
Well, I mean, so AI and literature is everywhere in the news these days. And so there was the Nobel Prize winner who uses AI, but then also a smaller scale thing that I sort of noticed in my feed was Granta, this sort of very, very well known and respected literary. English literary magazine.
- [18:55]
B
Yeah. British literary magazine.
- [18:58]
A
And then scandalously, they published a short story. Lots of readers are quite sure that this story was partially, at least written by a And so there's a huge kerfuffle surrounding that. And that's a. You know, again, that's not the Nobel Prize level, but it's pretty darn. You know, it's like certainly the upper reaches of advanced culture. That's not self publishing paranormal romance on Amazon. That's the different level where. Where indeed that was the. You know, where scandals have arisen. So an example that kind of I can't help but find kind of amusing. The self published writer of a paranormal romance. She left the prompt that she gave to the AI Somehow remained in the text that she published.
- [19:52]
B
Oh my goodness.
- [19:55]
A
And it's since been retracted from Amazon, but. So you would be reading this, you know, quite generic novel.
- [20:05]
B
Yeah.
- [20:05]
A
And you stumble upon. I've rewritten the following paragraph to match the style of this other more successful writer, lending it tones of whatever. And. And the. That was left in the text of the novel. So she literally did not bother. So talk about the editing function. She didn't even bother to edit her text well enough to notice that it was still. That prompt was still in the text of the novel before she published the novel.
- [20:34]
B
Yeah, well, that seems kind of. Yeah.
- [20:36]
A
You know, so like, whoa, man, that's.
- [20:39]
B
Well, shouldn't the AI Bot have.
- [20:42]
A
No. I know. Yeah. Right.
- [20:43]
B
Gotten. You know, you don't really. This isn't good for our.
- [20:47]
A
Right. So you got to say, hey, AI, please make sure this doesn't.
- [20:50]
B
Yeah.
- [20:50]
A
I haven't left any embarrassing marks of my use of A.I.
- [20:55]
B
no fingerprints. Well, I don't think. I mean Amazon's. They don't. Do they have any standards when it comes to like self publishing with.
- [21:05]
A
I haven't checked just lately, but I'm pretty sure they do. I'm pretty sure they at least have a boilerplate statement that it's not. You're not supposed to do that. A lot of entities like that. Yeah. Journals, et cetera. Will at least state that you're not supposed to do it. And I know I. Too much to sort of fanfare. I think it was the New York Times that came out with a very explicit set of policies about what their reporters are allowed to do with AI and whatnot. In that case, of course they're worried about the hallucination problem just because generative AI just produces so much and it really can't be fully cured ever of this, as I understand it just produces so much fiction, ironically, that. That a newspaper, the paper of record, of course, can't. Can't go down that road.
- [22:03]
B
Yeah. Well, that wouldn't I mean, you know, the New York Times prides itself on having, you know, really tough standards, and that doesn't surprise me. But I wonder, you know, I mean, I'm. Substack has become this big kind of alternative media realm. I recently started a substack newsletter, and so far, I mean, there doesn't appear to be anyone telling me what that you can't, can't publish anything. I mean, every, There must be somewhere. I don't know how it exactly works,
- [22:39]
A
but yeah, I don't know the back end of substack well enough to, to, to be able to say I, Yeah, but there is a lot, presumably. There really are. There aren't any hard guard rails. It's really like, do people encounter what they feel like is inauthentic writing?
- [22:57]
B
I haven't detected any. Yeah, I, I.
- [23:00]
A
And who knows if you would. So we're in a great age of paranoia, basically.
- [23:04]
B
Yeah, well, there's that, you know, it's basically.
- [23:07]
A
Are these pod people or these, are these, you know, have you been taken over by aliens who look a lot like us?
- [23:15]
B
Well, yeah, but, you know the line about how even the paranoids are sometimes, you know, on target here.
- [23:21]
A
Very true.
- [23:22]
B
So, so, yeah, and also about substack. You know, there's, there's, there are a lot of people, you know, short stories and things like that on substack. I mean, I, you know, I have no idea. I usually, I sometimes scroll them quickly. I, I don't know what, you know, exactly where they're coming from, and the barriers to entry are practically nothing.
- [23:46]
A
No, I know, I know. And I guess, I mean, if one sort of tries to be not depressed about this reality, I mean, you could say, well, the proof will be in the pudding. If you liked the story, why do you care that it was AI assisted or even AI written? I mean, on the other hand, I understand, I respect deeply that when people read fiction, most of them, it's important to them that that's an act of communication from a human author to themselves. This is deeply. We don't often talk about this, but this deeply underlies the experience of reading a fiction as you're getting the vision of a fellow human being. And to remove that can be kind of weird. On the other hand, there are people, you know, it's called the Eliza effect,
- [24:41]
B
when
- [24:43]
A
you treat a chatbot as though it's a real person. And there are, you know, I'm sure you've seen the news, there are people who are going off the deep end and taking their chatbots as real people. Yeah. And, you know, and so that could happen, too. And that just seems a little bit disturbing, like how you could forget that this isn't a real person. They don't. This thing doesn't care, really. So there's so many angles on this. It's very.
- [25:12]
B
Yeah. Well, I think what you said before, I mean, there's a certain. I mean, just speaking as a reader myself, I mean, the intimacy that you can find sometimes in connecting with an author, and particularly when it's a novel, maybe sometimes when it's a philosopher, but I think the novel kind of foremost. So I've, you know, I've certainly felt that way. I mean, when I was in my 20s, I was reading a lot of Dickens or my first time reading Middlemarch, I felt like the author is Georgia. She's kind of speaking.
- [25:48]
A
I. Similarly, Middlemarch is my go to for novels as some of the greatest human creations. That would be my candidate. And yes, it has a lot to do with encountering a sense of a human sensibility, of such profound sensitivity and mercy for the human condition. The thing is that, you know, like that, I guess that in theory can be simulated, and the evidence of people who fall in love with their chatbot suggests that, at least for some people, they're going to take the voice that they're. That they're encountering as a real human voice and sort of project humanity onto it. And so I don't know what to do with that.
- [26:33]
B
Yeah, it's hard not to feel a little bit judgmental about it. I mean, it feels kind of superficial. I mean.
- [26:39]
A
Well, I mean, it's one thing. It's one thing if you're fooled. Like you. Like you were told that this is a person and you thought you're interacting with a person, and then it turns out you're not. What I can't get are those scenarios where people are fully aware that this is a chatbot and they fall in love with it anyway. Like. Well, how did you forget how. Like, how did that happen? But I guess, you know. Yeah, it seems like human psychology is strange.
- [27:06]
B
Yeah. Or possibly a wider cultural failure of some sort. I don't know. I mean, I think maybe I don't know how to explain that one either.
- [27:15]
A
I don't know. I mean, it's no doubt some sort of failure. On the other hand, it's a profoundly human thing to project meaning into the world and to find things that are, strictly speaking, not meaningful. Unless you're religious, like Arak. Some. We. We find ourselves Projecting human, deep human meanings into the non human environment in lots of ways. And so we're primed to find life meaningful. And so this is kind of like pathological instance of that very basic human propensity.
- [27:51]
B
Yeah, yeah. I mean, right now I'm thinking, I mean, you know, I guess if we wanted to say this has been going on forever. I mean, the whole Pygmalion myth, and when I read some of the versions of it are really quite disturbing. I mean, the sculpt, I mean, it's almost like. I mean, there's a very sensual, sexual, almost aspect, maybe more than almost. You know, the, the creator is, is running, you know, his hands all over the body and this and that. And it's. I don't know, is that, is that, you know, were these, do these myths, you know, come from that? Are they tapping a similar desire or impulse?
- [28:39]
A
No doubt. I mean, and you could also look at Frankenstein, which, you know, which is obvious. Super relevant.
- [28:45]
B
Yeah.
- [28:46]
A
I mean, in those cases in Frankenstein and, and Pygmalion, the creator of the creator is interacting with their creation and they're having to ask themselves, you know, what, what have I done? The.
- [29:01]
B
What.
- [29:02]
A
The, the thing that's happened now is that that creation has been partially, at least displaced to distant corporate entities which own the situation.
- [29:12]
B
Yeah.
- [29:12]
A
And so in any given instance, it's not you interacting with your creation in Pygmalion or your monster in Frankenstein. It's corporate sponsored and owned. You know what I mean? And so there's a, one further level of possible paranoia about what's actually going on in the situation. You know, LLMs, it's intentionless language on some level. There is no human intention behind it. And LLM does not have any sense of reference to the real world. It's a lot of language. On the other hand, of course, there is a couple layers of intention behind it. There's the prompt, first of all, your intention and getting something from it, getting it to do something for you. And then behind that, of course, is the ultimate intention, which is the intention to sell you this experience and to acquire, you know, vast amounts of power by way of assembling all of the world's information and essentially controlling it. So, you know, the, the, the, you know, Elon Musk or Sam Altman as a. Yeah. As the author of your is. Is. Should. Should disturb people deeply.
- [30:28]
B
Yes, yes, I'm glad you said that. I feel like, I feel like we've kind of moved on from the being disturbed by the Facebook stuff and now we're I know.
- [30:38]
A
Can we have that? Can we have that?
- [30:40]
B
That seems innocent and bit benign.
- [30:43]
A
Oh, exactly.
- [30:45]
B
Yeah. Elon Musk and kind of in control. I mean. Yeah. Musk might be a side topic because he's. I mean, he's very interesting because he's also the one who points out all of the dark things that I find him in some ways more interesting than. I don't know, any of these people, really.
- [31:03]
A
Because he's unfiltered. Well, I mean, it's.
- [31:05]
B
Tim Altman seems. I don't know. I mean, the depiction of him, fairly or not, is like. He's more kind of like a hustler.
- [31:12]
A
Yeah.
- [31:13]
B
And Elon Musk isn't only a hustler.
- [31:15]
A
Right. He's in the weird space where he's an incidental truth teller. I mean, he's such a liar, but an incidental truth teller in some ways, because he just doesn't care on some level.
- [31:28]
B
Yeah. But he blurts it out like, yes, AI could make us extinct.
- [31:35]
A
Right.
- [31:36]
B
Think about that.
- [31:38]
A
Yeah. And then those admonitions are always partly like marketing devices. That's what's so strange about this moment. It's like, worry about AI and then the stock goes up.
- [31:51]
B
Well, yeah. Right. It just seems so much less. Right. We're being manipulated and we kind of know we're being manipulated. And there are going to be a million Netflix streaming series about. In fact, I stumbled into one, this bizarre thing when I was on the plane coming back from Italy. I was kind of running out of things to think about or read or what. So I just. I thought it was going to be a stupid, like, romance comedy. And it turned out it was. I think it was called the Companion. It was. They were all like bots that they were being ordered. And I was thinking, wow, this is like, they're going to be like a million movies like this.
- [32:28]
A
Yeah, no, this is going to be a theme. And it's funny, you know, it's. You want. You want to come back to literature.
- [32:35]
B
Yeah.
- [32:36]
A
And would like to think that literature, the novel, and certainly lyric poetry as a refuge from that, from this kind of stuff. And that's what makes the. The AI insofar as it applies to literature and literary experience. So troubling is because we are trained, in some ways, rightly, to think about literary experience as, you know, a potentially profoundly authentic medium of communication between human beings. And so to the extent that that gets undermined, it seems deeply troubling.
- [33:12]
B
Yeah. It also just seems a lot flatter. I mean, one of the going back into my 20s, I was just. For some reason, I was just like swallowing all these Victorian novels. I mean, I really was drilling Mrs. Gaskell and the Brontes and every. But then I was reading also some of the biographies about them and something of the social and political conditions that, like in the part of England where the, you know, the Brontes were, you know, the more living, you know, it was all kind of fascinating to me and it. But it made sense. So, you know, I. Connecting their imaginative, you know, works to, you know, something that they were in some way grounded in.
- [33:59]
A
Right.
- [33:59]
B
But that just seems impossible if we're talking about AI because as you just said earlier, it's as. They don't seem to make any of these kind of value judgments or there's no content. They don't have historical consciousness or context.
- [34:14]
A
No, I mean, what. They. Yeah, they're. They. They are, to use the term you just said you just used. They're ungrounded. Yeah, they're ungrounded in any relation between words and the world that we walk through as embodied people. It's for the. It's all pattern recognition. Oh, this word tends to occur near this word. That's literally the. The essence of it times a. Times a zillion in terms of iterations, et cetera. And. But it never needs to, as it were, touch down, touch grass with. With embodied reality. And so it, you know, it is this, in a way, profoundly different thing. The irony, though, is that, you know, in some ways. In some ways, you know, generative textuality is, in that sense, purely autonomous, which has been one of the great themes of high art since the 19th century. And so. So just to add another wrinkle that should sort of philosophically trouble you a bit, in some ways, the ideal for art in some. In some places has been precisely the sort of transcendence of mere utility and of mere groundedness in some particular reality in favor of the timeless wisdom of, you know, and that's why you get this, why Shakespeare can be so uncannily relevant hundreds and hundreds of years later. And that's by virtue of Shakespeare having transcended his historical moment. And speaking to us now, LLMs are this sort of uncanny kind of version of language freed from that kind of groundedness. But we shouldn't necessarily want that.
- [36:18]
B
Yeah, I come back to. Yeah, I mean, I'm trying. I have a little reading group now. We're making our way through Finnegan's Wake. Oh, wow.
- [36:27]
A
Good for you.
- [36:27]
B
Yeah, well, we're 144 pages in. And I don't think it is not unreadable. This is what I say.
- [36:39]
A
Yeah, no, it's. Yeah.
- [36:41]
B
But a lot of our little discussions we have in the group on Zoom every other week are about. Have to do with Joyce and what was he thinking and this and that. And I think it would be fascinating to have somebody like Joyce, James Joyce, live in the era of AI and what.
- [37:00]
A
Yeah, like, what would you. Yeah, how would he do with it?
- [37:03]
B
I mean, he was such a magpie. I mean, he just. You could spend hours just trying to figure out his. His illusions, his references and all this stuff. So I. You know, I don't know what the answer to that is.
- [37:16]
A
Yeah, no, it's a very good question. And I don't know if anyone has asked an AI to read Finnegan's Wake Up.
- [37:24]
B
That would be.
- [37:25]
A
Yeah, you could let Claude assist your reading group. No, I'm just kidding. Don't.
- [37:30]
B
Don't do it. Well, we throw in things, but. No, I mean, we've been using the. The Joseph Campbell in the.
- [37:38]
A
Oh, the introduction.
- [37:39]
B
Yeah, he has, like, the skeleton key. Anyway, it's a side topic.
- [37:43]
A
I have to.
- [37:43]
B
No, no. But I haven't felt. I've been a little bit disappointed given. Given Campbell's reputation, because I feel like he's winging it also, like the rest of us are.
- [37:54]
A
No, it's his first book, and he's just trying to. Yeah, no, you're right. But people. It's an interesting thing about that kind of difficult, experimental writing, especially from that era. Everyone reads it with help. Nobody. I tell this to my students all the time when I'm teaching them difficult modernist works of literature. It's like if you think somebody picked up the Wasteland in 1922 and, like, read it as though they were reading the newspaper, think again. Everyone was completely confused. Everyone found Henry James prose difficult to deal with. The sentences were so long and complex. That's not just our experience.
- [38:32]
B
They're long.
- [38:34]
A
That was the experience of a lot of readers in his time. And so when we're dealing with cultural objects like that, we all seek help from either real teachers or the teachers that are embodied in the handbook that help us read that stuff.
- [38:52]
B
Well, right. I would never read the Wake on my own. I mean, it would just be.
- [38:56]
A
It would just be like. It would be impossible.
- [38:57]
B
It'd be ridiculous. You know, I read, you know, our sample couple of pages, you know, every two weeks, and then I myself. And then I. And then we talk about it, but without the talking about it. Part which. Which, you know, I don't know. Maybe the AI would. Would be helpful. It's hard.
- [39:14]
A
Yeah. I mean, there's. There's some continuum between a kind of guidebook to a difficult work of literature and what AI might do. I mean, and that's the thing. I mean, we don't. At least I don't think. And we shouldn't consider, like, using something to help you read Finnegan's Wake as deauthenticating your experience of the work. I think it makes your experience of the work possible in the most basic sense. And so when we get to AI assisted comprehension of things, how are we supposed to think about that? Can we just make our peace with that? Or is there just something so undermining about using our own minds in that context with that tool that we should really just not do it?
- [40:01]
B
I don't know. Yeah, I don't know either. I mean, it. I mean, the one thing AI was. Doesn't seem it will ever do is try to shame you. It doesn't want to sit in judgment over you. In fact, I have to remind my Allison, just stop with this, like, ridiculous, you know, flattery or just, you know, just give it. I'm training her to just give, you
- [40:23]
A
know, to be, like, a little bit more. Yeah, yeah, yeah, yeah. Well.
- [40:27]
B
Right. It's also more the way, you know, my friends are or something. We don't.
- [40:32]
A
Yeah. Mine.
- [40:32]
B
So I don't call people up so just they could, you know.
- [40:35]
A
You're so, like. Wow, that's so fascinating.
- [40:39]
B
Right, exactly.
- [40:40]
A
Stop. Wow. You've. You've. That's really sharp the way you've made that distinction between blah, blah, blah. Yeah, no, I. I don't want a safe.
- [40:47]
B
I don't. I'm. You know, maybe it's a function of my aid, not looking for the safe space, you know.
- [40:51]
A
Right, right, right.
- [40:52]
B
I can handle it. Just, you know, give it to me straight. So that would be my ideal kind of AI experience in that sense is just.
- [41:02]
A
I wonder if they're going to come up with acerbic AI personality. Presumably they will.
- [41:07]
B
Sarcasm is one of those things that's really hard. I've seen this in social media, that a lot of people who are pretty sharp people, the sarcasm can go right by them.
- [41:19]
A
Yeah. No, fair enough.
- [41:21]
B
So I don't know, you know.
- [41:22]
A
Yeah, no, that's a point. It's an interesting question. I don't even know the history of why AI is so ingratiating. I'm not sure how that happened, but it is an interesting Fact about it
- [41:35]
B
must be some market research or something that.
- [41:38]
A
Yeah. Or the training set was. I don't know. Yeah. Who knows? Who knows?
- [41:43]
B
I don't know. I don't know. I mean, as an author, I had, you know, non fiction. I have reason to be pretty cynical. I mean, you know, anthropic is this huge class action suit where they basically fed, you know, all these pirated books. And I've got, you know, I think three of you know, I've got some money. They're coming right away. Yeah, you probably do too.
- [42:04]
A
Yeah, I'm waiting for my check.
- [42:06]
B
I'm waiting for my check. I'm literally waiting for my check. Will it ever come? But.
- [42:10]
A
But even though. Even though. Even there, like. And you know, my lefty friends will hate me for saying this, you know, like, it's not. AI does not cut and paste from your work and feed it to people. That's not how the technology.
- [42:26]
B
Oh, no. Why would they think so?
- [42:28]
A
There's a level of mediation here in the plagiarism, in the quote unquote plagiarism, which is kind of, again, I have a capacity just sort of be fascinated intellectually about it before I get to the. To the stage of condemnation. Yeah. It's kind of an interesting question. Like what, How. Like what is the nature of that derivation from everyone's work such that it can be understood as theft, even though it's not cutting and pasting from your work. That's not how it works.
- [43:01]
B
No. Right. Yeah. I mean, I suppose I'd feel somewhat disappointed if I wasn't among those who were stolen from.
- [43:08]
A
Right.
- [43:08]
B
What would that mean? My books were so, you know, inconsequential, peripheral that they never made it to the, you know, to the theft. Right.
- [43:18]
A
This is our. This is our immortality here.
- [43:20]
B
Exactly. This is our immortality. So right there. Well, you know, so there's so many ways to slice it and probably a lot of really, you know, bad metaphors that can be applied, which we can avoid, but we just don't know. I think, you know, maybe that's not a great place to end up in the conversation, but did you have any predictions? I mean, I don't know if you're in the prognosis business, but do you think there will be a novel that, I don't know, book reviewers will someday hail as magnificent that comes out of the AI machine in some fashion?
- [44:00]
A
It's a really interesting question. I was reading this is not quite that, but it's close enough. I was reading, you know, fellowship Applications just Part of the grunt work of being a professor.
- [44:13]
B
Yeah.
- [44:14]
A
And there was this really interesting project where this young scholar was observing a certain kind of avant garde artist, one of whom what this artist does is construct, not LLMs, but SMM, small SLM, small language models.
- [44:33]
B
Okay.
- [44:34]
A
And they use these small language models, which are, of course, highly particular, so only certain kinds of things are allowed to be the training set. And then uses those small language models to sort of generate texts that have a different quality than ones generated by the LLMs, the everything set. And that just got me thinking, like, maybe there's some hope, like human ingenuity will be like, well, if I can tweak this and tweak that. And actually, it's a scale problem. If I'm going to take a subset of particular kinds of works and work with them such that I, as an artist can sort of re. Achieve some sense of overarching agency over what is happening even as I'm remaining conversant with the technology. I found that kind of intriguing. I was like, and is that maybe a hopeful sign that if what everyone is saying about human creativity is true, maybe we can count on figures figuring out how to take the next step in an interesting way such that we feel like we're using this technology, not being used by it. I don't know even in that one case if that's how one should ultimately understand it. But it was intriguing enough. It was like, oh, wow, that's pretty aggressive. Take an LLM and make it an slm, a small language model, make it bespoke and start to do interesting things artistically with it. I don't know if that was like, okay, if that can be the model, maybe we'll figure out how to remain humans.
- [46:25]
B
More agency, individual agency. Yeah. Well, maybe there's a parallel in art. I mean, in art, like in. I mean, the use of digital tools to create art, like abstract art, you know, I mean, which is. I know, you know, many artists who do that, you know, and I mean, Apple, I think, you know, seems to have like a universe of computers that are kind of set up for people in that kind of creative.
- [46:54]
A
Oh, yeah.
- [46:55]
B
Work. And that doesn't seem to draw. I guess people think that. Oh. Because the person doing it is, you know, fully in control. But I don't know, maybe there are suggestions being, you know, if you're talking about constructing a painting or something. I don't know.
- [47:13]
A
Yeah, well, I mean, it's, It's. It's fundamental philosophically, the capacities of your tool, the tool you're using become, you know, your, your limitations. They sort of defines the parameter, the affordances of any tool you're using. And then so you're ultimately always going to be asking are you using the tool or is the tool using you? That's sort of inherent in the, in the, in the, in the proposition. It's just that this moment with this technology AI is such an exaggerated version of that. It's one thing, it's one thing to say that, you know, if you only own a hammer, everything looks like a nail. We get that. It's something else. When all of the imaginable tools in the world have been absorbed and offered back to you for your use in a way that's possibly fundamentally limits you as well as enables you. It's really tough to know how to sort that out.
- [48:14]
B
It could be that literature also is a unique situation. I mean, we're always still, we're talking about words in the end. I mean that's, you know, it's literature. There are words, whether it's a poem or a short story, a novel. But if we talk about like you know, quote unquote, painting, I mean, people might say, well, it was really more authentic when, you know, Manet was working with you know, oil paint and a brush and on the canvas. And then, you know, even Jackson Pollock was, okay, he was like dripping cans of paint on the canvas. And now we have artists who are, who are using their computers, you know, to generate these things. But with literature you're still, it's still. It has to be about words.
- [49:02]
A
It does. It, I mean, does. But those, yeah, but those words are they, they can now be algorithmically generated in ways that convince a lot of readers that they make sense and that they do the job. And so really the jury is out on whether human linguistic creativity will establish a beachhead which AI cannot assault. I really don't know.
- [49:34]
B
Yeah, or less, less high mindedly. It could be. I don't know. I have to admit I haven't read a lot of romance genre novels or formulaic. But I assume isn't the goal to get the reader to cry at some point or something like that?
- [49:51]
A
Yeah, but the evidence of the ELIZA effect where people take their chatbots as real people suggests that that's not gonna be hard.
- [49:59]
B
That's what I'm thinking. That part will. Never mind about George Eliot giving us a new, you know, conception of the universe or something of the social system in the universe.
- [50:10]
A
Exactly.
- [50:11]
B
Can the AI, you know, can they produce tear jerkers or, or horror, You Know, whatever the genre is. Doodlers, horror, tearjerker.
- [50:20]
A
Yeah, no, there's a deep relation. This is what I'm writing on currently. There's a deep relation between genre and generativity. What, what, what. What AIs are doing with language. In a way, an AI is a very generic phenomenon. If you prompt it to write something with castles and rattling chains and whatnot, it immediately gets the generic environment that you're referring to, and everything that it'll produces will look like a gothic novel. And so that's deeply embedded in language as a system is sort of genres of discourse and we use different genres. And that extends to real life too. We walk into a room and, you know, we. We engage in a certain genre of disc.
- [51:09]
B
Yeah, we can.
- [51:10]
A
That's appropriate to the context.
- [51:11]
B
Where does the word genre come from?
- [51:15]
A
Well, I mean, it's. It, you know, it's. It's. It's related to. It means kind. Right. And so it's related to genetic. It's related to, you know, Generic. Generic. Yeah, genericity, Generativity. You know, and. And that. What that also points out is not that like one generic thing is similar to another generic thing, but that there's a generativity. So, like, it's not just that that things are. Are like other things, it's also that they're propulsively moving forward, making new things. That's the genetic part. That's the sort of making new beings. And that's the aspect of AI that's also kind of interesting is just how quickly it makes words.
- [52:06]
B
Yeah. Yeah.
- [52:07]
A
Well, yeah, I mean, you know what I mean? And that's the generative part.
- [52:11]
B
Yeah.
- [52:12]
A
Faster than I can type one token after another, very quickly. So genre also has that meaning of more and more and more.
- [52:22]
B
Yeah. I was. I was listening, trying to think where the. You know, who. Where the etymology of the. I don't know the answer with the etymology of the word. It's not a word that you seem. It feels like I hear it a lot more in the last, like 20 years than I heard. Like, you know, like 40 years ago or something maybe that. I don't know.
- [52:40]
A
Possibly. Yeah.
- [52:41]
B
Yeah. Is that your next book, do you think? I don't want to, you know.
- [52:44]
A
Oh, yeah. Well, it's something I'm working on, trying to keep myself sane by doing it. I'm ready to say that that tries to make the link between literary genre and gener. Textuality and like, what it. How can we begin to think about the longer history of generative AI by thinking about the history of genre.
- [53:06]
B
Yeah. Well, that's interesting. Okay, well, I'll wait for that essay to come out.
- [53:11]
A
Yeah, I'm waiting to. And alas, I can't. An AI will not be writing it. That's why it took me months. It's going to be, as always, it's going to be the usual suffering.
- [53:23]
B
That's it. Well, there you go. Thank God for that. I mean, there's suffering still bleeding onto the.
- [53:31]
A
Oh, no, no longer. Sweet suffering. Yeah.
- [53:34]
B
Such sweet suffering. There you go. All right. Well, I'll stop torturing you. But I've enjoyed our conversation.
- [53:40]
A
Me, too.
- [53:41]
B
Hope you have. And so this is Paul Starabin signing off for America and beyond.