[00:00:14] Chris Freeland: For decades, we've turned to search engines and social platforms to understand ourselves and the world around us. We've typed in our questions, our curiosity, our grief, trusting that somewhere in the machine there might be clarity or connection. But those same technologies that promise, understanding are built by corporations with their own ambitions, incentives, and blind spots.

If our most human searches are fed into machines designed to study and monetize them, are we searching the technology or is the technology searching us? Hi everyone. I'm Chris Freeland. I'm a librarian at the Internet Archive. Welcome to today's book talk, in 'Searches: Selfhood in the Digital Age' tech journalist Vauhini Vara, draws from her own very personal experience using chatbots to write about her sister's death to explore how technology companies have both fulfilled and exploited, our desire for meaning and connection.

Joining Vauhini in our conversation today is Humanity Scholar Luca Messara, who wrote Vanishing Culture. So here to introduce our speaker is Dave Hansen, the Executive Director of Authors Alliance. 

[00:01:23] Dave Hansen: Thanks, Chris. Welcome everyone. I'm really excited. I think I say that every time we do a book talk that I'm excited about it.

But this one's really cool because you know, a lot of the book talks that we do, a lot of the issues that we focus on are really pretty heavily like legal policy, technology focused, and I think they're very helpful and they're interesting conversations to have. It's really easy, I think, to sort of lose the forest for the trees.

And what I mean by that is we get really kind of down into some of the naughty questions about who's making money where, and who's benefiting from what policy or another, and what does that mean for, you know, the future of technology. And, and I think it's really helpful to sort of take a step back and just sort of reflect on how we as humans interact with these different systems that are.

In some ways thrust into our faces. So I was really excited about this book talk in particular. So first, Luca Messarra, I think many of you may remember, Luca, for his work on vanishing culture. It's a really, really incredible report. Luca was the primary author serving in a role as a public humanities fellow at internet Archive was the primary author.

A very compelling initial section of that report that deals with media preservation and the production of public memory, so I highly recommend you check it out. It's super interesting. Luca is a PhD candidate at Stanford University in the English department, and he will be moderating today's discussion with Vauhini Vara.

Who you probably also have heard of, I think I have seen her mostly, or I first encountered her work through The Atlantic, but she's previously written and edited for the New Yorker and the New York Times Magazine as well, and she's been a contributing writer for business week. We're really, really excited to have her here.

She, of course, is the author of 'Searches: Selfhood in the Digital Age', which she's gonna talk through with us today, and this book has received a lot of attention. It was named the Best Book of the Year by Esquire and Kikis and Publishers Weekly, and received numerous awards, but she's also written some really interesting other books.

One is titled, This is Salvaged. Which I, I would encourage you to check out. And the other, The Immortal King Rao, which was actually a finalist for the Pulitzer Prize and several other awards. So with that, I will get out of the way. I will turn it over to Luca and ine and let you take it away. 

[00:03:49] Luca: Thank you Dave, for that incredibly kind introduction.

Hello Vauhini. Nice to see you. 

[00:03:53] Vauhini: Hi Luca.

[00:03:54] Luca: can we just start off a sort of higher order overview. What exactly is your book and what's it about and why did you write it? 

[00:04:01] Vauhini: Yeah, so. I have found as somebody who's written as a journalist about the tech industry and also somebody who is 43 years old and started using the internet when I was in middle school, that we've fallen into this kind of status quo where there's a somewhat binary discussion happening about technology.

So a lot of people, especially critics of big technology companies and their products will say. These companies are exploiting us and we have no choice in the matter. Full stop. They're so dominant, there's nothing we can do about it. And then the tech companies will say, no, no, no, no, no. You all are using our products because we're providing you with something you really want.

If you didn't really want them, you have the choice to stop using them. And then we've go outta business, right? Some version of. And I was interested in the fact that sort of both and neither of those statements can be true at the same time. And I wanted to explore what it feels like to be a person using technology in a world in which both of those things are true and untrue at the same time.

Right. And I wanted to do that in a couple different ways. I wanted to use the technology products themselves, my use of those products as a way to explore that. So for example. Sort of every other chapter of the book. As you know, Luca is made up of like a kind of archive of my use of these technology companies products.

So there's a chapter made up of my own Google searches over 10 years. There's a chapter made up of my own Amazon reviews over a period of months. Those are the even numbered chapters. The odd numbered chapters are sort of like more conventionally narrative chapters, talking about my own experience as like just a person in the world starting when I was 12, 13 years old.

As a journalist covering technology companies, and so it kind of proceeds chronologically through that and my changing understanding of these companies and the power dynamics in their products. Since then, I wrote all that and then when I finished it, I did this sort of, you know, unconventional, maybe controversial thing, which was that I fed it two chapters at a time into chat, GPT, and I told chat GPT.

I was writing this thing and I wanted to discuss it with chat, GPT, and so there are these sort of interstitials starting the book and ending kind of right before the end of the book in which you kind of see that dialogue with Chachi PT in real time. 

[00:06:20] Luca: That was a great introduction to the book. So I'm curious where you started when you began writing this project.

So you're working from, is it 2019 when you first started with your AI experiments up until 2022, is it? 

[00:06:33] Vauhini: Yeah, so the first thing that I wrote that ends up in the book is the chapter made up of my Google searches, and I didn't know that it was the beginning of the book. What I was thinking about at the time was the way in which.

The most complete archive probably of my life over the past 20 years now since Google has been keeping an archive of all my searches is maintained by Google. Say I die however many years from now and wanna like give my archive away my papers away to like some university archive. That university archive that has like my notes about the novels I wrote will be less complete than the archive of me that Google has.

Right. And that's interesting to me because it's valuable. Like that's a very valuable archive. It's valuable to me personally, like it contains all this intimate information about my life over the past 20 years. At the same time, it's valuable in a different sense to Google itself, right? It's financially valuable to Google because, and politically valuable because it gives Google all this information about me that it can profit from both financially and otherwise.

And I thought an essay made up of those searches could sort of help reveal that. I published that essay in the New York Times in 2019. As an opinion piece, sort of an unconventional opinion piece. But what I realized after that was that like I thought people needed some handholding to like have context for understanding what I thought the subtext was.

Publish that in 2019, about a year, two years later. I did my first sort of AI oriented experiment in which I wrote this essay about my sister's death 20 years earlier, and my grief over it, in which I asked this early AI model, one of the early GPT models underlying that, you know, the predecessor to the models underlying chat, GPT to ask it.

To sort of complete my thoughts in my effort to write about my sister, and that also became an essay. But with that too, I published it as an essay, but then felt later that it needed some context so that people would sort of have some philosophical and political grounding to understand it. So that's how the form kind of came into place.

[00:08:48] Luca: I wanna talk about ghosts actually, because that's how I came to understand you as a writer. That's what I was first exposed to when I was working through grief myself and somebody recommended that I check out ghosts and it was actually really powerful and moving for me in that moment. So thank you.

You expressed like a lot of ambivalence about the popularity of that story, and I don't mm-hmm. Dunno if you could share a little bit in this format why that is, that you feel a little strange about its popularity. 

[00:09:10] Vauhini: Yeah. So just to explain a little more for people who aren't familiar with it, what it is.

I didn't think it was gonna be an essay. I just was playing around. I got early access to GPT-3. That was the model I was using, opening AI GPT-3 model more than a year before chat, GBT came out, and some of you may be familiar with this, but the interface was like this box and you would type in something.

Hit a button and then the model would kind of complete the text in a way that was intended to match something about both the substance and the style of what you'd given it. Which is of course very different from how chat GBT now functions. And I was thinking, because I think about what technology companies are trying to sell us when they give us products, I was like, well, what would be like the product form of this?

And I thought to myself, oh, it would be something that would promise us. It could give us words when we're at a loss for words. And so kind of like it intellectually, but also emotionally. I was like, well, what are the things for me, I'm a writer. I'm a journalist. Like what for me is something that I don't have words for?

And of course, oftentimes like trauma is what we don't have words for. And one major traumatic experience in my life was the death of my sister. So I went into this thing, I typed a sentence, it was. When I was in my freshman year of high school and my sister was in her junior year, she was diagnosed with Ewing Sarcoma, which was the cancer that my sister was diagnosed with.

Hit the button and it completed it. It was all wrong, so I deleted it all. Went back to my original first sentence. I kept that. I wrote some more, then hit the button again. What it gave me was like somehow a little closer to saying something about death, about grief, about sisterhood, you know, and so I did that process over and over nine times.

And each time like. My own texts got longer and it's got shorter, but with each iteration, it seemed to me that there was like something more attuned to human experience of grief and death that the model was getting at. Which, you know, is probably controversial to say, because some of us don't like to say, oh, an AI model is like expressing something, quote unquote about the human experience.

But. That's what happened. But also, ultimately, of course, the AI model wasn't able to express what it was like to be me experiencing my grief about my real sister. And so when I published it, I thought of it as being about both of those things, about the way in which like this model was capturing something, but also about the way in which ultimately it couldn't.

And so I felt that there was like. A critique embedded within that. It seems to me from the response in the, like many months after the essay came out, it went viral. It was adapted for this American life. It was mythologized, but a lot of people who appreciated the essay were just like, wow, this is so cool.

You know? Like this writer is collaborating with the same eye model. The result is this really moving thing and interesting thing about grief. And I felt that like the critique that I was interested in. Was not something that a lot of readers found themselves interested in and that unsettled me.

[00:12:10] Luca: Something that you do that's really exciting to me in the book version of Ghosts, 'cause you republished it in the book as well, is that you changed the ending. Can you talk about that decision to do that? 

[00:12:19] Vauhini: Yes. So in the original essay, so there are these nine iterations, and in the last iteration, almost all of the writing is my own.

It's like me writing about my sister's death and my grief. Then there are a few sentences at the very end, which I give to the AI model, and so in, in a sense, the AI model, like it's much diminished in that last one, but it still has the last word. And it occurred to me that like on the level of rhetoric for people who are reading the essay, that essay has a kind of ambiguous ending, right?

Like if I'm giving the AI model the last word, maybe that's because. AI model deserves the last word because it's expressing itself better than I'm able to. Right. And I wanted to make clear rhetorically in this formal structure itself that that wasn't what I was interested in. I was interested in saying, listen, I'm the one who should have the last word.

And so in the version in the book, in the ninth iteration, I don't have any text from the AI model. It's all my own writing, and I've taken away that AI ending. 

[00:13:20] Luca: I think your emphasis on rhetoric throughout the entirety of the book is really interesting. So oftentimes when you ask chat GBT to evaluate the chapters that you had written, you ask it how it feels about your representations at the tech industry, which I thought was really fun.

And it produces, of course, these very saccharine responses that are just like, oh yeah, you should probably be a little bit nicer to Sam Altman, obviously, we're like, no. If that's okay. The thing that I want to kind of get at is there are people who might say that writing experiments with AI and LLMs are sort of contradictory to the humanities stuff that we sort of advocate for.

You're a novelist. You care about the craft of writing, so I'm wondering if we could think a little bit more in depth about the sort of higher order stakes of writing with LLMs. And the first thing that I wanna post to you is that I think your book is really good and that it operates as a kind of archive of large language models because you've written it over the course of so many years.

There was a point in Cheche Bee's history where the like rich tapestries were all the rave. Right. If somebody had tapestry in their paper and I was teaching at the time, I was like, well, that's a little skeptical. I'm not sure if this was written by this person or not, but the question that I'm kind of getting at here is, do you think that it's like important to archive these large language model responses, and what would that work look like?

[00:14:30] Vauhini: Yeah, your underlying question I think is also about like just the ethics of using AI in language production period, right? As a creative person, as a writer myself, and I think there too, there's been this kind of binary discussion, right? It's either large language models are useful tools that we can use to make writing easier for us, more efficient, cheaper, all those things, right?

And then the other side of the argument is. Writing and reading are fundamentally human activities and should therefore be emerged from our human psyches or in collaboration with other humans. These machines should not be involved and more sophisticatedly these machines built and sold by major, rich, and powerful technology companies by exploiting labor, exploiting the environment should not be involved, is the more sophisticated version of that argument.

And I think there's another way to think about it, which is the way I'm thinking about it in the book, which is can we use the output of these products that produce language to reveal something about the products and about the companies behind them and about the power dynamics embedded in our interactions with them.

I think an interesting analog here actually is research, like academic research that reveals. Power, dynamic biases, shortcomings in these products by using the products. Right? And also, academic research isn't typically going into the process of like. Investigating these products with a goal of like a gotcha, right, of saying like, look how horrible these are.

They're legitimately curious about understanding how they function and that's the spirit with it, which I'm approaching them too. I'm not approaching these products to say like, let me use them to show how biased they are. I'm just, I'm curious as a reporter, as a human. So that's what I'm doing there. And like with academics, like with journalists operating with the journalist hat on, I am interested in just as you say.

Having there be an archive of what these products are now and for future readers or future species who discovers these things, what these products were when they existed. 

[00:16:36] Luca: And I think that's why it's like so interesting to me that you inject your own personal narrative to get back to the research thing that you were saying.

Mm-hmm. Your personal voice and your personal narrative is so important in this book because as it grounds a lot of the juxtaposition between you as a writer and like the outputs of the LLMs. I have to admit, like I woke up this morning and I was rereading your book and I. Oh my gosh, what a bucki is just fooling all of us.

And this is entirely an AI book and like even the personal writing parts of it are just ai, which I don't think is true. Maybe that's true, but you could let me know. Something that I kind of wanna pivot to thinking about here then too is there are a lot of people who would I. Defend ai, and you talk about this in your book in the 13th chapter when you're talking about pseudo, right?

There are people who would say, this is like a direct quote, somebody who says, I am proud to be a criminal against the arts. If it means now everyone of all abilities can write the book that they've always dreamed of writing. That's a very hyperbolic argument, but it's like something that strikes as. A somewhat compelling argument in favor of LLMs in that it promotes accessibility to like the master discourse or the effective language that one would need in order to succeed in like a business environment or even to just create art that people are individually interested in.

Do you have like a response to that argument where, oh yeah, this ease is great 'cause it means that people can make whatever it is that they want to make now. 

[00:17:51] Vauhini: I do, I have a sort of like systems oriented response to that argument, which is to say like with a lot of arguments about the usefulness of technology products, there's this underlying assumption that it is a desirable thing for people to use products that will allow them to use the discourse used by the master or the dominant culture because those are the people in and institutions in power.

And if we want to. Succeed in that world, we need to use that language. Of course, another way of thinking about it is to say, let's address the underlying issue here, which is that assumption that there are certain ways of using language that are considered the desirable ways of using language, right?

Because people who study language, including people like you, Luca, know that there are all kinds of ways to think about what the role of language is, and one alternative way to think about language is to say. What makes language powerful is the fact that we all have access to it and can make of it whatever we want.

And it makes sense for language to be expressed in all these a multitude of different ways because we individually are different, our communities are different. And if we look at language that way, and if we exalt all those different forms of language, then we create a world in which like this sort of very standardized, specific.

American white, white collar office speak that is standardized across AI products actually doesn't become desirable. It becomes the opposite of that, and that's something that interests me. So those people who say, I can't write because I don't have talent, or maybe I have a disability that makes it impossible for me to write, I sometimes hear things like that.

Another way to think about it is, is to say, well, can we change the terms of the discussion? Like what if. You do write the way that you feel able to write the way that comes naturally to you. And as a culture we say yes, like that is an appropriate and interesting way to write Precisely because it is non-traditional.

Right? Precisely because it doesn't match with convention. 

[00:20:05] Luca: And this is where I think printing the LLM responses is like really valuable and important because to use like the literary studies word to me, it like really defamiliarizes our experience of language. So defamiliarization is like this formalist concept that says that like what makes.

Poetry, really, really wonderful is that it makes unfamiliar our experiences of everyday life. So like the classic example is the stone. So like poetry makes the stone stony essentially. So it makes us like rethink what a stone is. And I think your book makes us rethink like what? Personal voices in juxtaposition with the outputs of an LLM.

And I think that only comes out when it's printed in the format of the book Nevada. And I'm interested in, there's this common thing that whenever we reprint stuff from the internet, we put it in serf. And whenever it's like somebody else's voice, we'll put it in like SF voice and serve font. I'm worrying if there was ever a decision to actually make the entire thing just, or to like not so clearly delineate.

What is the AI and what is your voice? 

[00:21:00] Vauhini: That actually gets back to something I wanted to address earlier, and you were kind of saying it in jest, but like the fact that you woke up this morning and was like, oh no, what if this is all a joke against us? I do as a person, as well as as a writer and journalist specifically, feel that disclosure is really important.

Right. So there is a section in the back of the book where I talk about like the process that went into each of these and the extent to which like AI or other. AI generated language or other language generated by technology companies, products is used in the book, and the choice in different fonts had to do with that as well.

I wanted to clearly delineate in the book itself what was me and what was language generated by a product, by a big tech company's product. So that was part of it. But the reason I chose that sense was also somewhat political. Maybe satirical, I would say like it's a satirical choice in that I wanted to, I mean, partly it's, but people are familiar with that sort of design language, right?

So people see that and they kind of have a sense of it. So I thought it would be useful for clarity, but I also kind of wanted to poke fun at the fact that's the visual language, a big technology company's products, and the Serif font is traditionally the language of like human written work. So that was something I was thinking of.

[00:22:15] Luca: Awesome. As we're thinking about this kind of juxtaposition between like the human and the non-human, you as a writer, AI as a writer, I'm curious what word you would use to describe your relationship, your writing relationship with ai. Are you editing with ai? Are you like using AI as like a form of poetic production?

What word would you use to describe your relationship? 

[00:22:37] Vauhini: I would back up for a sec and say that I think oftentimes we say like human versus machine, human versus technology. And I am more interested, I would say in like this idea of like human as in people, you know, like people as individuals, people in communities as opposed to institutions, right?

That are invested in power and wealth. I'm sort of more interested in like this, like human versus. Powerful institution dichotomy. And the reason I bring that up is that I'm not that interested. Well, I am interested, but for the purposes of the book, I was less interested in like. Large language models or language models as technology period, right?

Which could be anything. It could be a university built language model. A person could do it in their basement. A really tiny model, right? It could be the internet archive could make a language model, but the reality is the language models that are dominant are built by the big technology companies. And so that's the relationship I wanted to explore.

So in that question, I think is like we all are familiar with some of the language people use to talk about ai, the use of AI in writing or in creating art. Oftentimes those words include things like collaboration, right? Or assistance or aid. I'm sure I'm forgetting some. And I really chafe against those words because.

The use of those words, I think absorb the rhetoric that we're being given by big technology companies, right? So the technology companies benefit from using language that kind of humanizes their products. So we collaborate. Collaboration is something that we do with each other, but if we use the word collaboration in relation to a product, like I would never, we don't say things like, oh, I went and collaborated with Amazon the other day to buy toothpaste.

Or I went and collaborated with Google to find out. Find an academic paper I was looking for, right. But they've created that word so that we feel like there's something human like about the products, which of course serves their interest in getting us to trust their products and use their products and then confide in them.

And so to go back to the language I use, there's no consistent language I use, but I tend to, you might have noticed in this conversation actually, it's kind of wordy and cumbersome, but I tend to like to say. Big technology companies, products rather than just AI or chatbots, because I wanna characterize them as products.

Who do have like humans behind them, but those humans are people working for the companies, investors in the companies that are sort of setting those company's motivations. And so I talk about those products that way. Another thing I do is I will set use words like use. I use the products I. Put language from the products, or I put rhetoric from the products in the book.

And so I'm trying to kind of actively defamiliarize our relationship with these products in the way that I use language as well. 

[00:25:24] Luca: And I think the word collaboration is really interesting here. 'cause towards the end of the book, the very last chapter, which is my favorite chapter of the entire book. 

Could you maybe like summarize for the audience who maybe hasn't read your book, what's going on in your last chapter and like why you decided to conclude your book with that chapter? 

[00:25:38] Vauhini: Yeah, so I made this survey in a Google form and it was deliberately in a Google form, a product from Google. About what it's like to be alive.

And I posted it in all these different places and I asked people to fill it out, and I said, your responses might be used in something that I'm working on. I didn't know at that time that it would be for the book. So I disclosed that and I posted this thing all over the place. I actually also put it on Amazon's mechanical Turk service deliberately to see how that would change the way people responded.

And so it includes questions like, tell me something, anything you want about yourself, and what was happening in your life in the time, seconds or minutes or hours or days or months or years before you started filling out the survey. Stuff like that. And then I ask women, people who identify as women to fill it out for personal reasons and also reasons that like felt political and connected to some of the subject matter of the book.

Then I collected all these responses and I created a chapter made up of the questions, followed by a selection of responses. And what I love about the chapter may be part of what you love too, Luca, is the way, just how different it is from the way chat GBT answers things, right? Like people are so.

Original people are so creative, the way people use language is so poignant, basically. Like there are little stories and these anonymous answers people provided that are so much more meaningful and beautiful than anything. Chat, GPT or Claude would produce if I said, tell me a little story. Just beautiful stuff and I love the power of that, the emotional power, also the political power of showing that on the page.

[00:27:16] Luca: I think like I would love to sit and theorize more about what's going on in that chapter, but maybe what we should do if you're open to it, is to like do some like improvisatory readings of excerpts from that chapter. So I think what we'll do, I'd like to just read some of these responses that we've gotten from women presumably around the world, and their responses to these questions.

So what is it like to be alive? 

[00:27:37] Vauhini: I can whistle, but I can't snap my fingers. 

[00:27:40] Luca: I broke a glass coffee pot on a kitchen countertop. 

[00:27:44] Vauhini: My parents let me play in a wok in their kitchen. 

[00:27:47] Luca: My father taking me to the desert to see camels. 

[00:27:51] Vauhini: My mom's mom grew up in Nazi Germany and her dad grew up in Kentucky, so there are some words she pronounces very strangely.

[00:27:57] Luca: I cried over a man. 

[00:28:00] Vauhini: I know they survived. 

[00:28:02] Luca: I'm a cloud, 

[00:28:03] Vauhini: everyone I love inside this house is sleeping in different places, 

[00:28:07] Luca: scraping my heel on a photo album. 

[00:28:10] Vauhini: Taylor Swifts Powerful Thighs. 

[00:28:13] Luca: I love that one. Probably some of them had lactose intolerance. 

[00:28:18] Vauhini: It would have to have, oh, I just wanna give context here.

This is in response to a question I have about inventing a technology that essentially allows us to make humans better. Okay. I'm gonna read too. It would have to have a huge mechanism to deal with shame and then. It has a part where you flate into a little steampunk nozzle and it turns into fragrant oxygen.

[00:28:42] Luca: I have a secret desire to just get a one-way ticket and go anywhere in the world, all by myself. 

[00:28:48] Vauhini: Lots of men were mean to my mother, including my father. She decided she would rather be alone. She went to college, taught public school, raised two girls. Neither of us are mean. 

[00:28:58] Luca: Humans are social creatures who have managed to knit themselves into a massive, intricate web.

Isn't that a beautiful existence? To be human is to never be truly alone. 

[00:29:09] Vauhini: We end there 

[00:29:10] Luca: bigger. It'd be a good way to end it. Yeah, that was great. So there's a, maybe I could say one more thing about this. You have like a kind of note where you mentioned you made some sort of light stylistic edits to the piece.

Could you maybe talk through about, I know you said like they weren't that intensive, but 

[00:29:24] Vauhini: Yeah. It was interesting, Luca, you asked me about this in, in like an earlier chat and it reminded me that at the time of the copy editing, I have a lot of back and forth with a copy editor about how to approach edits to this section.

And so this morning I sent you a couple of examples of how it worked. The way I thought about it was that because people communicate and express themselves in like such diverse, weird, interesting ways as much as possible, I wanted to preserve that. But there were a few places where it was clear that somebody used like.

A spelling of a word and probably intended to use a different spelling, like for example, using then THEN instead of then THAN or leaving out a word like of, or something like that, that like clearly belongs there. So there were a small handful of instances where we did put those in, in the editing, but other like quirks and how people use language, we left in.

[00:30:20] Dave Hansen: I was gonna tell you, there's a rich tapestry of questioning, a lot of it has to do with how you think AI is sort of fundamentally changing just the future of human writing, even when we're not using these tools. 

[00:30:33] Vauhini: Oh, 

[00:30:33] Dave Hansen: yeah. And you're a novelist. You've written lots of things outside of this context, and so maybe you could start there, just like, how do you think it has influenced.

Your overall approach to writing in big ways or small ways. For instance, I teach, and I know all my students tell me they're now terrified to use M dashes because the accused of using ai. So maybe you have something more profound than that. But 

[00:30:54] Vauhini: I mean, what I'll say is that I think, again, thinking of ai, not as some mysterious, profound new thing, but as something that's embedded in products by companies, what I can say is that it's been true for all of history, right?

That like. The way we use language as individuals and communities is influenced both by our own personal interest in how language should function, the interests of the collectives that we're part of, and then the interests of the powerful institutions that decide how language should be used. Right? And so I write in the book, for example, about colonialism and how.

Colonial powers used language, particularly Spanish and English obviously, to express and exert their power. So I think that's true of AI generated language as well. But I'm interested in thinking about that less as like this crisis that's particular to this moment and more in thinking about it as part of a continuum.

Because if that's the case, if it belongs to this continuum, that means we have experience in thinking about how to exercise it. And I think for example, actually about like. Indian writers who write in the English language, the language of the colonizer in specific ways, using their own version of the language and also using their language itself, their own rhetorical power to say anti colonialist things, to say things that are, that run counter to the interest of the British Empire.

Right? And so we can think about that in the same way when we're thinking about AI generated language. 

[00:32:26] Luca: I just wanna pull out really quickly one colonial thread too, 'cause it's in your book Vauhini and it's really quite important. You talk about the Spanish requirement, which I always feel I need to mention to people.

It's this crazy text that the Spanish would read whenever they were about to go and pillage like a group of indigenous people in the Americas. They would read it out loud that explained to the indigenous peoples why exactly they were under the control of the Spanish. And it's an absurd document.

Because if you're not from that epistemology, if you're not from that mindset, it's fundamentally absurd. And like even the people who are reading it necessarily, they might not fully understand the language of what's going on in there. So language has always been this kind of tool that like we borrow and use without really knowing where the origins of it are from.

And we use it to exert power in strange ways. So this isn't just something that's limited to LLMs, it's something that is like been since the beginning of time. 

[00:33:16] Vauhini: And then to take that thread and pull it, like think about Bad Bunny at the Super Bowl, right? What a great example of how we then in turn can counter exert power using that same language.

[00:33:27] Luca: Exactly. 

[00:33:28] Dave Hansen: So that's a good segue into the question in here about discerning and critically considering chatbot's responses. So you write about technology. I would probably characterize you as like a pretty well-informed user of these tools, but a lot of people aren't. And so the question is really about AI literacy essentially.

Yeah. And maybe in particular you could talk about that from the angle of new writers. I don't know if you know this, but I was searching before this, doing a little research and there are whole lesson guides for like creative writer programs to try to do what you did with your kind of back and forth with the AI system.

And it's interesting to me to see people teaching that and seeing what students pull from it. So I'd love to, to hear your thoughts on that. 

[00:34:11] Vauhini: Yeah, I mean part of what I'm trying to do with the book, obviously is like, I'm hoping people see the chatbot conversations in the book. Like I hope it's a form of literacy education, right?

Because you see in real time how this product is trying to convince a writer to write a book that's more positive about the tech industry than the book she's actually writing. You know, there was a study out of Cornell a couple years ago, the same researches at Cornell have some more recent related research as well.

That showed, they took these three groups of students and they told them to write essays about social media, and one of them wrote the essays themselves. Another group was given this, like this fine tuned chat bot that was predisposed to express positive opinions about social media. And then a third group was given a chat bot that was predisposed to express negative opinions about ai.

And they found, not surprisingly, that the students who use the chat bot. Were far more likely than would be typical to express opinions aligned with what their chatbot expressed. And so you think about the broader implications of that, right? So if it turns out that chatbots from companies like OpenAI and Google and so on, maybe themselves predisposed to express certain biases about their own industry, for example, or racially or otherwise, the research suggests that.

Regular folks who are using these products may be quite inclined to like absorb that rhetoric into their own language also. So we talk about like these different dimensions I think of like why it's problematic to use these products to write. But I think like for me, like the rhetorical influence of these companies through their products is one of the scariest pieces.

[00:35:56] Dave Hansen: Luca, I wanted to open that to you too, 'cause you teach, and I'd be interested in your perspective on how this is all shaking out, especially in that context. 

[00:36:03] Luca: For my teaching, what I usually lean on is project-based assignments, actually, rather than an essay. So like to me, like there's a lot of crisis in the humanities about what does it mean to write papers.

I tend to not actually assign papers a lot of the time in my class, and I'm more interested in assignments that engage students in like the lived communities that they're in. So for example. When I was at City College in New York, I was teaching a Black American Poetry and poetics class, and the main assignment that I had them do was we were reading like all these old archival sources of poetry, many of which are no longer published or available today.

So there's like an interesting copyright and like Orphan Works element to this. What I had them do is assemble their favorite selections from the poems that we had read. We were reading like entire collections in whole. They had to pick a specific order to the poetry that they were doing, and then they needed to provide some sort of introduction about like why they chose the poems that they did.

And I guess they could use AI to do that, but like, that'd be kind of sad if they've spent all this time like crafting some intimate poetry collection about love, which is what most of them were about. Where like, this is like how I experience and think about love. And then if they used AI to write that, that'd be kind of sad.

I'm pretty sure most of my students didn't use AI to write those introductions. And I think that that sort of like project based humanities is like a good way to go for like the humanities going forward. Like I think we've spent a lot of time thinking about essays. To be clear, I still love essays. I think they're wonderful and great, but the humanities as like a field does things beyond the essay and beyond writing, and I think the crisis of LLMs is a good opportunity to rethink like what it is that we're doing and what is the value of what we're doing.

[00:37:34] Dave Hansen: I love that. Yeah, I mean, I think my big concern there. Is that there are probably a lot of people I know not everyone is like this, but for me writing is sort of therapeutic and like I don't actually know what I think until I write it down. And I know not everybody is that way, but I worry that there is this segment of the population that gets sort of sucked into using these tools and maybe doesn't realize like how valuable the self-reflection tool.

Writing can be because you just are able to sort of skip a lot of those mental steps to generate it. So I have another question about dangers of using ai, which we've kind of jumped around already, but I wanna take it out of the creative writing fiction, writing context. Well, Heini, you're a journalist.

You write for a number of journalistic outlets, and what do you see as the dangers there? That's a really different kind of writing. 

[00:38:23] Vauhini: I think we can separate. Some different dangers, but they also are interconnected. So there's the danger, of course, if news organizations are increasingly controlled by a small number of media conglomerates, and those companies want to be as financially successful as they can be, and so it follows that.

Some of them may think of text on a page as like a commodity that. They hope can be as cheap as possible to produce. And if that's the way in which they think about language and communication, then it would be natural for them to start turning to AI models on a very basic level, like that's dangerous to journalism as a profession because if there are fewer of us, we don't have individual journalists working in a newsroom don't have the resources to do the important work they need to do, which connects to the broader sort of societal implications.

So just like if it's true that. 20 students in an in a classroom are all writing the same pro ai, pro social media essay because of the AI thing they were using. That's bad for them, but it's also bad for anybody who would come across those essays. Right. It's also the case that if newsrooms are using those kinds of products, including as like chat bots to tell you what was in the news today or whatever, right?

That has implications for the journalists themselves, but also for society. 

[00:39:46] Luca: Yeah, I mean a lot of this is an issue of scale. So we're talking about like journalism at scale and descaling the work of being a journalist and the fact that like something that is going to be distributed to like millions of people might have been written by a chat bot.

That's not great 'cause it's gonna reproduce structural inequities to me. And I don't wanna have like a head in the sand approach to it. 'cause I don't believe in a head in a sand approach to it. But like to me, it calls me to like participate more in like local. Journalism, be more interested in mutual aid networks.

Get on the ground and like know the people that you're living with and being around and pay attention to, like your hyper-local politics. And maybe that's one way to get out of the kind of like big scale, big data approaches to journalism. Not, I mean, obviously we should be attempted to them too, but there's also something we said about like, I don't know, starting a little zine with your neighborhood and distributing it around.

[00:40:33] Dave Hansen: I think I have time for one last one, and this actually touches on a topic we were chatting about before we started the book talk. You're using AI in this book and AI causes a lot of feelings, including, there's like a lot of negative reactions sometimes from certain communities. And I wondered, have you experienced that, a kind of negative reaction from broader communities about your use in this book?

What's your reaction been to that? How have you just like navigated all of that? 'cause it's sort of a complicated space right now. 

[00:41:00] Vauhini: I really appreciate that response to the book actually, because part of what I think of the book as a kind of provocation, right? Like I want it to incite feelings and political, personal feelings.

Political feelings in readers, and if a strong negative reaction to the use of AI in the book is one of them, I think it's a very legitimate reaction whether people understand or not, like what my goal is in writing the book, people get to encounter the book on their own terms. So if somebody picks this up.

Sees that it's made up partly of chat, GBT generated language and says something like, I don't wanna pay, I don't even know how much the book, 20 whatever dollars for this thing that's like made up of this worthless text. I think that's a really interesting response. It's kind of a response I like, right?

In some ways because people are making a political statement as readers, as consumers, as members of society about like what they value or not, and I think that's really meaningful. If people say, listen, I understand that this is a critique, but I think you can effectively critique these companies and their products without using the AI tools themselves and putting the output in your book.

I think that's a legitimate critique. So of course, I wrote the book the way I did because I felt personally that it was the only way to express what I wanted to express. In the way I wanted to express it. But like, that's one opinion, you know? So I really welcome all those critiques. I think they're good critiques.

[00:42:31] Dave Hansen: Well, I love it. I, I'm a big believer in showing, not telling and I think it really like illustrated in such a powerful way. Lemme just thank you both. This was such a wonderful conversation. I really enjoyed it. So thank you for this and I'm gonna turn it over to Chris. 

[00:42:48] Chris Freeland: Thanks Dave. I wanna echo my thanks to you Vauhini and to Luca.

Thank you so much for your conversation today, for writing the work and for being here to, to talk with us all about it. Thanks everyone. Have a great day.

[00:43:00] Chris Freeland: Thanks for joining us on this journey into the future of knowledge. Be sure to follow the show. New episodes, drop every other Wednesday with bold ideas, fresh insights, and the voices shaping tomorrow.