I was interviewed on Substack Live by Walker Larson the other day. Here’s a brief section (I edited the Substack transcript a little for clarity) about AI and literature, which develops some of the thoughts I wrote up recently.
Walker Larson: How do you think the the rise of AI is going to change or influence our relationship with great books? Obviously it causes problems in schools, you know if you’re going to try to get students to write about books, then you have to contend with artificial intelligence. But even just how do you see it affecting the literary world? Will we see a resurgence of interest in classic books because they’re permanent, because they are not generated by artificial intelligence, but by real human beings with hearts and minds? Or will it just more and more obscure those works?
Henry Oliver For the first time in a long time, there’s real competition for the book people. And it’s really good. It’s getting better every day. And very soon, we’re going to experience the kind of disruption that has been relatively unknown. A lot of people will sort of disagree with that and think the internet was very disrupting to books and publishing and education and all the institutions around books. I suspect that once the AI revolution really gets underway, we will come to realize that the real competition is starting now and we’re going to be much more shaken up than we were before.
Ultimately, I think it’s going to be a good thing because AI will be much better than you at doing taxes or whatever, but there’s no point in having it read Tolstoy for you. Tolstoy cannot be summarized. Tolstoy can barely be explained. I mean, you know, it’s like saying I watched the weather report and believing that you have somehow experienced the storm that’s happening outside. Either you have been out and stood in the rain and heard the thunder and seen the lightning hit the tree or you haven’t. And that’s Tolstoy.
So one of the things that will be left to us entire and whole is the reading of literature. Because it is irreducible, it is irreplaceable. But one of the things that AI will be able to do is interact with scholarship and criticism. And some of the things that literary scholars currently do, AI is getting much better at doing. And some people will use AI. Some people will just sort of refuse on principle. Some people will use it. The people who use it will have, you know, the equivalent of dozens of other people working for them. It’s like getting research assistants. Go and tell me everything that the scholarship says about X. Once you’ve done that, go and look up such and such a thing in the archive and then find me this and then relate it to this. And the work is just going like that all the time. Once that starts happening, the way in which scholarship and criticism works will just have to change, right?
If we think we can sustain a model whereby it’s, you know, organic, it’s 100% human-made, no AI was consulted in the process, probably enough people will pay for those magazines that the old model will just carry on to some extent. But... We just have to accept that some incredibly bright young kid in Newcastle or Ohio or wherever is going to start doing things the new way.
So that that’s coming and that will change all the surrounding things about literature. And a lot of the other young people who are coming up may prefer that new way to our way. Or they may want a bit of both a bit like how a lot of people use AI for their therapy. Going to see a human and talking about your problems, you would think is pretty irreplaceable. Like, why confess to someone if it’s not a priest? But people do. So I just think it’s going to change. If it means that overall... There’s more humanities, there’s more discussion of literature, there’s more discovery of what’s actually in the historical record, what’s in the archives, there’s more suggestive information, like stylometric analysis can be done, all sorts of, like, if it means we end up with a lot more stuff like that, that can only be a good thing, right?
It may be that the people currently practicing those things are the ones who lose out but that the overall output is different and better. At the level of the school pupil, I’m not sure that AI is a problem because you’re no longer, I mean, a lot of people will not like the way I’m going to phrase this, but you’re no longer stuck with your teacher. You can have a different teacher if you want. And that, for many people, Depending on which part of the education system they’re in and how well it’s working and how well it suits them and whatever else. That may be very, very liberating. And it may be what gives them what they need. It may be the difference between low expectations and high expectations. So overall, I expect AI to do good things, but it will also be used in bad ways. And the final... The final equilibrium is some way away, and God knows what it really looks like.
Agreed on Tolstoy. I would slow down on the therapy aside, because it is the same case. A person can be handed an accurate summary of their own life and nothing happens. I did that more than once early on, and watched a correct account of someone slide straight off them. What changed anything was slower and had another person's attention in it. The weather report fails for the storm in a book and for the one in a life, and I expect people will find that out in the same order.
As accountants are already seeing dreadful errors made by AI when people are using them for their tax returns I think we need to be more circumspect about both its capabilities and applications. An accountant I know recently saw just such an example and said "this is going to keep us very busy for the next 10 to 15 years".