Subscribe here: Apple Podcasts | Spotify | YouTube
AI-infused writing is suddenly everywhere: not just in automated email responses or meeting summaries, but in books and even the pages of mainstream newspapers. When an editor at The Wall Street Journal recently defended an op-ed that was 100 percent written by AI, it felt like a possible turning point, at which people are no longer apologizing. So is AI-assisted writing in its post-shame era? What does that mean for the way we process, communicate, and think? Host Charlie Warzel talks to the journalists Elizabeth Spiers and Teddy Brown about being a writer when the robots are coming for you, and how technology does not have to be inevitable.
The following is a transcript of the episode:
Elizabeth Spiers: I just worry that AI is making everybody dumber. I think when you outsource the fundamental cognitive arsenal that you’re using to actually construct a good argument or to make a piece of art, what happens to us as a society if everybody’s outsourcing that to AI models?
Teddy Brown: When I worked in tech, I was writing 3,000-page papers about really dense B2B-enterprise marketing strategies, or whatever. Yeah, exactly. And I would get to the end, and it would go onto the website: I have no idea what I wrote about. I could not tell you one thing about that, because it was just all flowing through Claude.
Charlie Warzel: I’m Charlie Warzel, and this is Galaxy Brain, a show where today we’re gonna talk about artificial intelligence and the thing that I do for a living, which is writing. We’re almost four years into the ChatGPT era, and AI writing is everywhere. It’s flooded the internet. We’ve got SEO spam, slop—all that synthetic pablum that’s just immediately recognizable. A Pew Research analysis from July estimated that “signs of AI authorship can be found in over one-third of web pages published after ChatGPT was released.” One-third of web pages. It’s come for schools; it’s come for universities, which have been struggling for years to get students to write papers without chatbot assistance. And in professional writing, we’ve watched this slow and steady creep of AI show up in book publishing, in mainstream publications, in op-eds. And for a while, the reaction was all the same.
As my colleague Will Oremus wrote recently: “Up to now, those accused of passing off a computer’s writing as their own have tended to either apologize or deny it.”
In April, James Taranto, who is The Wall Street Journal’s editorial features editor, wrote that “those of us who practice writing as a professional craft have an ethical duty and sometimes a contractual obligation to ensure that our work is original.”
Now, a few months later, in the same paper, billionaire Stanley Druckenmiller published an op-ed which was identified as 100 percent written by AI. Only this time, there was no deflecting; there was no apologizing. Druckenmiller and the head of The Wall Street Journal’s opinion pages admitted to and defended the usage of this chatbot to write the op-ed. The head of the opinion page responded: “AI is a fact of modern life.”
You hear that a lot now.
And I want to be clear: I get it. Writing is hard. It can be solitary. It can be frustrating. There’s a lot of people out there who are not professional writers and who don’t want to have to write certain things, like a “just looping back” email to a boss or to a coworker. Last week I had to write an email to a pest-control company about being overcharged for a mousetrap. It’s not sexy stuff. There’s a lot of writing out there that is drudgery. It is the exact stuff that artificial intelligence automates really, really well. So listen; I get it.
And there’s also the issue here that, as a writer, I may not be the most impartial messenger. Of course, I’m gonna be defensive about this. I don’t want something to come out there and take the one marketable skill that I have and just make it irrelevant. But I also think that there’s something else going on here. Something that goes well beyond what precious writerly types like myself think. Because writing isn’t just a boring, overly formal thing that we have to do to get through school. It’s an essential form of communication.
There’s a lot of debate over whether writing is synonymous with thinking, and well, I’m not sure that it’s one-to-one here. Writing, and especially learning to write effectively, can be a powerful skill that helps advance or sharpen our thinking. Writing is a way to organize our thoughts. It’s a way to set aside quiet, focused time to understand what we really believe about something.
It is the scaffolding on which we build arguments, which means that it is, in part, a way to crystallize how we see the world.
What I noticed most in the response to this Wall Street Journal op-ed and all the backlash was this group of people arguing that this is where things are going. The toothpaste can’t be put back in the tube, and so we might as well embrace it. It’s a “resistance is futile” type of argument. But the thing about new technologies is that they’re never inevitable. Nor is it written in stone that professional writers and news organizations and people who rely on the trust of their audience to sell a product that is quite literally the craft of written argumentation need to roll over completely without negotiating their own boundaries in this moment.
To automate our drudgery is understandable. But to automate such a broad form of communication—to outsource an important tool for working past the roadblocks that exist in our own minds—that’s something different than embracing a new gadget.
And so when this op-ed came out, I saw all the backlash—all the “This is akin to using a calculator,” “This is like spellcheck,” “This is like Grammarly,” “Please stop all the hand-wringing.” I wanted to go deeper. I could imagine this grim future where these tools take over almost completely, and where writing becomes this strange, somewhat vestigial skill. In that future, I think humanity is worse off. Now, I can also imagine a future where AI-assisted writing is more akin to a tool. Maybe it’s something that someday breaks open a new hybrid form of the craft. Is there a world where chatbots can help people to learn how to structure arguments better, to enhance their vocabularies, to learn how to spot good writing from bad, and make us better communicators in the stuff that matters, while also automating all the little stuff? I’m skeptical, but it doesn’t mean that it’s not possible.
So, what does the future of AI writing look like? I wanted to bring on some other writers who are living through this moment and who crucially have different experiences doing different kinds of professional writing. Teddy Brown is a reporter at CNN, and in a previous life he worked in the tech industry as a technical writer, doing a lot of that for-hire, less sexy writing that’s under threat from Claude and ChatGPT. He wrote an essay titled “The End of the Working Writer” that gets at a lot of this stuff.
Elizabeth Spiers is a contributing opinion writer at The New York Times, the host of Slate’s Money podcast, and a digital strategist who also teaches writing. You may also know her as the founding editor of Gawker. She recently wrote a column with the evocative title “Why You Shouldn’t Use AI to Write a Professional Op-Ed, by Me, a Professional Op-Ed Writer Who Has Less Real Competition If You Do.”
So how should we think about all this? Teddy and Elizabeth join me now to talk about all of it.
Teddy, Elizabeth—thank you for coming on Galaxy Brain. Welcome.
Spiers: Thank you for having us.
Brown: Thanks for having us here.
Warzel: So I want to start here with a baseline. You’re both professional writers, so walk me through a little bit your relationship to generative AI and chatbots as it relates to your life and work here. Writing, drafting, research, cheap Google assistant. Teddy, I’ll start with you. Just how are you using this in your life, or not using it?
Brown: Yeah, so I actually don’t use it at all for any of my professional stuff. It’s funny; I came from tech before. I’ve become staff at CNN now, but that was sort of only five months ago. But before that, all my day jobs were in tech, when I used AI quite a lot. Just because it was sort of coming down from a lot of leadership that this is the tool we’re gonna use. And then I’ve shifted over during my tenure at CNN to not using them at all. So it’s become sort of—you see the differences emerge pretty quickly about how your process works. And I was a freelancer for a long time as well, so I always kept sort of a partition up between my journalistic work and, you know, my day-job work, when I would use AI tools in my day job but not use them for my journalistic work.
Warzel: Elizabeth, what about you?
Spiers: Yeah, so I’m a writer, and I co-host Slate Money, which is Slate’s finance-news podcast of the week. So I have to—you know, I write about AI. We sometimes podcast about AI. So I use the tools partly to understand them. But also, I work in a lot of newsrooms, as a consultant, and so I’ve been responsible in some cases for introducing AI to help automate workflows, or to do certain types of data journalism. So I’m not anti-AI. But I do not use it to write. It’s not very helpful for research, I found, because it’s sort of worse than baseline Google in terms of what it produces. So I’m happy to leave it entirely out of my writing process.
Warzel: Inside newsrooms, what are some of the ways that you’ve seen this stuff come into a newsroom and really help the journalistic process?
Spiers: Where I’ve used it more directly, it’s been for automating workflows. If you work in any kind of large newsroom, you know that we’re dealing with a bunch of different pieces of software that very often are cobbled together by processes that are heavily, you know, human oriented, and they can be tedious. And a lot of it’s just using things that are badly connected to other things. And you can use AI to automate some of that stuff, which takes the pressure off of, I think, journalists to keep doing things that are tedious and administrative, as opposed to actually going out and reporting or creating stories and columns and, you know—the things that are actually hard to do.
Warzel: Right. And there is that that difference between the research tool—the cheap, you know, Google replacement—and the, like, “I have this corpus of information; it’s so massive.” Or it’s, you know, scattered in all these different places. The ability to go and and scan through that type of stuff, or collate or organize, like those daisy-chaining, really tedious tasks together to make something else, is I think a sort of an underlooked version of that. Which, I think a lot of journalists tend to be industry-wide very split on AI stuff. And that’s sort of what I want to get into here, especially as we just talk about writing.
So we’re basically—November will be four years into the ChatGPT era. AI writing’s obviously flooded the internet. We have the book-publishing industry, which, according to a lot of reporting, is in utter chaos. Book deals falling through. I had someone in the book industry tell me recently, “Don’t write one now.” Like, don’t pitch one, don’t—just, like, stop on books for like six months, please. Because the industry has no idea what the hell to do, and they’re just terrified that they’re gonna be getting AI-generated work.
And then you have, further in the world of professional writing, something like the billionaire Stanley Druckenmiller, who recently published an op-ed in The Wall Street Journal identified as 100 percent written by AI. And both him and the head of the Journal’s opinion page admitted to and defended the usage of chatbots. The head of the opinion page [Paul Gigot] wrote in his own op-ed—who knows if he wrote that himself—that “AI is a fact of modern life.”
Right after that, Semafor did some reporting, showed that an analysis of 310 guest-writer op-eds in the last six months in The Wall Street Journal, Washington Post, and The New York Times—one in six used AI, according to Pangram, the AI detector. Now, Pangram is in its own way, you know—we’re not a hundred percent sure about the accuracy there. But it seems like there’s a lot here. And I normally start all these conversations that work toward the big thing, but let’s do the big thing first. Like, help me understand what’s going on here. Like, are we cooked?
Spiers: I think it depends on what kind of writing we’re talking about. I wrote a little ranty thing on Medium about the Gigot/Druckenmiller situation, and it was really directed specifically to people who wanna publish professional op-eds. And that’s actually not where, you know, I think the big problem is. Because, for the most part, the Druckenmiller kind of situation, part of that’s coming too from the way that these op-ed pages work. Where the problem is not about AI. But I do see people using AI for what I think of as “commodity content” a lot. You know, you go on LinkedIn, and you see the same kind of style of post that’s, you know, “Here’s what my divorce taught me about B2B sales,” or something of that variety. And, you know, that’s people sort of learning to write with a formula. But that’s also—you can sort of tell when they’re using AI. And I think AI can replicate formulaic, unoriginal stuff really well. So if that’s what you’re doing, then you probably are cooked.
But I also think that there is still a premium for original ideas, and I don’t think that AI is really capable of coming up with that. And the writing is that it produces is very cliché-ridden and formulaic, because it’s being trained on a corpus of data that for the most part is, you know, mostly mediocre. And there’s just no way to kind of prompt an AI into the sort of creativity or sort of holistic understanding of what it’s outputting the way that you can with a human and human editors.
Brown: I always tell people that it’s not the tics that these things have. It’s not use of cliché, it’s not lists of three. It’s not, you know, it’s not X or Y. It’s that these words don’t mean anything. Like, you read the entire piece, and like—I didn’t learn anything from here. This is all hollow. That’s what AI writing actually is.
Spiers: Exactly. Yeah.
Warzel: Right. So a feature of the dustup with the Wall Street Journal thing, which I actually think is much more important than the op-ed itself, is that lack of shame that we’re talking about here. Right? I think it’s the reason that so many people paid attention to it. If there was, you know, a tail-between-the-legs kind of thing, it would be like, Yeah, you know; somebody else got caught. We don’t do this. Do you think we’re creeping toward, like, a post-shame AI-writing era here? Does this mark for you all an actual departure, or is this just a couple of guys who were, you know, probably going in this direction anyway?
Spiers: I think there’s already a kind of post-shame stance from a lot of the writing community that’s not really institutionalized. And I don’t think that matters as much. But I think if you’re talking about a news organization—where trust matters, the relationship with the reader matters—I think at the very least you owe the reader a disclosure that AI was used in the piece. And you know, I’m a contributing writer for The New York Times, and my contract explicitly states that I cannot use AI. Even for, you know, putting it in to copy edit anything; not allowed to do that. In fact, if even if I’m gonna write about AI, I’m supposed to get permission to use the tools first. So the Times is sort of at the exact opposite end of the, I think, restrictive side of what people are and aren’t supposed to do.
But until Paul Gigot wrote this editorial, everyone I know at The Wall Street Journal was under the impression that you could not do your work with AI that way, either. So I think it may have some backfiring effect on them—that first of all. they’re going to get deluged with AI-written submissions. But also they’re going to have to clarify where that line is. And so far, they haven’t.
Brown: Yeah. I think that there’s a few arguments I’ve seen emerging. And one I’m not entirely unsympathetic to is that if the writing is good, who cares? And like, I understand where that comes from. Like, I think it’s actually beholden on readers and critics and things to be like, That writing sucks. and I don’t care if a human wrote it or if an AI wrote it. And now it just so happens that the balance of the vast majority of AI writing is awful, and that you gotta be able to identify it. And I know that it puts a lot of pressure on readers, who are just inundated with all of this slop, basically. But it is a little bit like, we should be a little more critical of what we’re putting in front of our faces and absorbing.
Warzel: I understand what you’re saying about having some—you know, you gotta put a little bit of an onus, too, on the reader to care about it. To vote with your eyeballs, etcetera. But to me, it feels like this is like a frustration that I have with the conversation around this op-ed. and around the feeling, too, from certain people who are just like: This is inevitable. This is where it’s gonna go. The internet is gonna flood with this stuff. So, like, we’re gonna look back on the hand-wringing about this and kind of laugh one day, right? It’s gonna feel super-quaint, all you people who are doing this.
And that may be true. Like, that I may be on the losing end of this thing, right? Like everyone’s just gonna be doing slop stuff, and it’s gonna get more sophisticated, and we’ll figure out some way forward. But for me, there’s also—in any era of technological change, you actually need people who are going to protest it. Who are going to stand and thwart it and stop it and object to it. Because that’s also how we create the new norms for this stuff, right? If we don’t spend some time thinking about that and having some people actually push back, if nobody pushes back, then it just happens. And it happens however these companies want. Or whatever the, you know, the lowest common denominator of slop that can get published. I find I’m really frustrated by the idea that some of these institutions, like the whole point of being the institution is to stand up for, you know, institutional things.
Spiers: Yeah, what’s crazy about this is that, first of all, it’s The Wall Street Journal doing it. Where you have one of the most, what business-side people would call “premium products in the space.” And you’re actively enshittifying your own product when what you’re selling is, you know, “the original thoughts of Stanley Druckenmiller”—not what he couldn’t be bothered to slap into a, you know, Google Doc himself, but produced via AI. Your value proposition, even on the business side, is eroded by that.
So I think there’s a lot of short-term thinking about headcount and cost. That really, people are using AI in ways that are not necessarily additive at the expense of the long-term value proposition of their own products.
Brown: I do wonder about this sometimes, because we’re in the media. And we exist in sort of a filter bubble of our own making, right? I’m like, do people in general audiences care that Druckenmiller uses AI? Does the 85 percent of people who do not care about writing and actually find writing a chore—that they don’t want to write a note about why they’re gonna be five minutes late; they’d rather press a button and have that happen for them—actually care that much? And I worry about that, because I care so much about writing.
Spiers: I think you’re right, but I also think with newsrooms, we have a trust problem already. And I think the readers care if they think they’re being deceived. I don’t think they really care about, you know, the process of writing, or writing for its own sake.
Warzel: I also think alongside the trust part of it, too, there’s an effort thing that I think is actually very real, that comes with the part of the trust. Like if people, you know, can’t be arsed to, like, do the work, and everyone else is like, That’s fine, yeah; just send that sucker through, right? There’s this idea of like—so where else do they cut corners?
Teddy, you wrote a good piece earlier this summer that was titled “The End of the Working Writer.” And I want to get to that, but the piece was kind of kicked off or inspired, perhaps, by a post by Bloomberg’s Joe Weisenthal, who’s constantly posting on X. This post set off, you know, a debate online. And he wrote: “Unfortunately, I think in the near future, not using LLMs to write for you will be like someone refusing to use Google Maps for directions in a new city. A bizarre, idiosyncratic choice that’s just completely incomprehensible to the vast majority of people.”
So first I just wanted to ask: Do you think that that prediction is right, here? Do you think that’s kind of where we’re heading?
Brown: I think “the vast majority of people” might be a little bit of an understatement. I think that as adoption picks up for a lot of these tools, I do think those people who, again, don’t really care about writing that much in the process of writing, and view it as a chore they can replace with a few clicks or a few keystrokes. Yeah—I think that’s gonna be quite a lot of people are going to start acting that way. And it’s just ease of doing these sort of things. We’re interested in the process of writing for its own sake, and in a way to sort of unpack our own thoughts and think about the problems we’re trying to solve. But yeah—I unfortunately see this becoming a little bit more acceptable to just use these chats for writing anything you need.
Warzel: Teddy, you mentioned earlier that you—and that’s part of this Substack post that you did—is that you used to do this kind of work. And a good bit of your argument in “The End of the Working Writer” is that the AI-and-writing conversation can get caught up in this, like, craft/artistic corridor—but that a lot of writing’s very utilitarian. It’s very technical, right? It’s this stuff: Press the button and say you’re gonna be five minutes late. Or it is this, you know, copywriting; you’re writing white papers, or product descriptions. Copy that’s gonna be used for search-engine optimization. Corporate speak. LinkedIn posts, Elizabeth, as you said. And since you did this kind of work, you argue maybe this conversation is not talking enough about how AI is changing that brand of writing. So tell me about the working writer a little bit—define it better for me, and what AI is doing to that category of writing.
Brown: Yeah. I mean, it’s all writing that no one cares about, is really what it comes down to. But I did this work because it was well compensated, and I had to afford rent in New York. And this is my career for a long time.
I think that what happens is that AI is very good at writing that doesn’t matter. All the things you were talking about—all those pieces of little pieces of internet copy that are not meant to actually be critically engaged with—AI is extremely good at writing those things. And so there are two forces happening. One is that people recognize that this writing is actually not worth that much anymore. And so they’re like, How can we do this for the smallest cost possible? And you have these enormous costs of AI integrations into technology companies generally, where a lot of the better-paid opportunities were.
Daniel Kolitz had a great piece in The Baffler today that’s talking exactly about this. Like it is there is this sort of squeezing-out of what I compared to session musicians in music. It’s like—Well, if I don’t need to pay these four people to play the backing track for this commercial thing; I can just plug into Suno and get the same thing, basically. It’s like, Why do I need to pay them $10,000? And like, right. That was paying those four people’s rent for the next couple months. And so I really worry about this hollowing out a generation of writers who are, you know, sort of 10 years behind me, who don’t have these opportunities to work on writing that is not necessarily the most wonderful artistic, literary pursuit, but is a way to at least make a respectable living in a very expensive city. And now those jobs are just gonna be gone.
Warzel: Let me just play devil’s advocate here. There are probably people who don’t live in New York City who are, you know, gonna listen in and say: Well, if it’s work that fundamentally doesn’t have any value, and it’s a good thing—like, what do we lose if we lose this?
Brown: I think it’s a big question. Because I was thinking about this before I got on, and sort of the idea of like—why, if you’re gonna take the job away from like 50 farmers with using mechanical tools or something. It’s like, Well, should we not do that? Should we not advance past this in order to sort of create more? Why would we keep paying people to do this writing no one actually cares about? And I understand that argument, sort of, in the medium and long term. But in the short term, it’s doing actual material damage to people who are trying to feed themselves.
Spiers: I just worry that AI is making everybody dumber.
Brown: That too.
Spiers: I think when you outsource the fundamental cognitive arsenal that you’re using to actually construct a good argument, or to make a piece of art, what happens to us as a society if everybody’s outsourcing that to AI models? I see this a little bit with, you know, I have an 11-year-old son in school. They’ve sort of—most of he and his peers all know how to use ChatGPT to ask a question. And it’s increasingly difficult for people in academic settings to get their students to do the rigorous work of sitting down and trying to think through an argument without these aids. And I think it may badly prepare students for the real world, where they may not have that assistance all the time.
Or they may have to learn how to think through more complex things that are just gonna be far more difficult, because they haven’t developed that intellectual rigor that comes from having to work your way through things from scratch.
Brown: I feel that as an adult. When I worked in tech, I was writing 3,000-page papers about really dense B2B-enterprise marketing strategies, or whatever.
Warzel: The dream.
Brown: Yeah, exactly. And I would get to the end, and it would go onto the website: I have no idea what I wrote about. I could not tell you one thing about that, because it was just all flowing through Claude. And all I was doing was editing line by line, and making sure it sounded good. But I could not tell you anything about what I actually wrote, which is weird. It’s a weird sensation.
Warzel: This gets to—I wanted to get here, which is that the broader “we” doesn’t think of writing, I think, as a technology. But it is. It’s an amazing technology that allows us to put our ideas in other people’s heads, in this intimate, transformative way. And then there’s this question of what it does for us, right? So I wanted to read something from the publication Nature about what might be going on when we write, that I enjoyed, here. Which is that:
“Writing compels us to think not in the chaotic, nonlinear way that our minds typically wander, but in a structured, intentional manner. By writing it down, we can sort years of research, data, and analysis into an actual story, thereby identifying our main message and the influence of our work. This is not merely a philosophical observation; it is backed by scientific evidence.”
And so I wanted to ask you all, as writers: Is the act of writing thinking? Or is it something different? Is it just something that forces us to think so that we may write? And I wanna get your thoughts on this, because I think it’s necessary and useful to put a point on what we are losing cognitively when we outsource it.
Spiers: I think it is thinking. It’s not the only form of thinking, but there’s plenty of research that kind of indicates that there are parts of your brain that are activated when you’re constructing something as a written sentence, versus other types of, you know, communicative expressions. There’s also a lot of anthropological history that indicates that when humans started developing language, it changed the way that they were cognitively structured. So I think if you remove that, I just think the additive benefits of it probably just go away, and their systemic effects on your cognition altogether. You know, maybe it gets replaced by something else. I don’t know.
But, you know, I wouldn’t rule out the possibility that whatever we’re losing can’t be replaced in the future. But I think for right now, we know that it serves a purpose. And we don’t have anything to replace it. So that does concern me.
Brown: I mean, of course writing is thinking. It’s thinking through issues, and trying to sort of figure out what you’re actually trying to say, or trying to write down. I mean, it is like it’s actively trying to entangle problems that you’re dealing with. Instead of being like, you can just use incredibly seductive tools like Claude and ChatGPT, ’cause you can just put in a problem and it gives you a solution. And I know that people will say you can have it challenge you, and you can turn things and press these knobs, and all this kind of stuff. But it is also outsourcing that path for you.
Warzel: I’ve thought a lot about this, because I’ve had some people challenge me on the “writing is thinking” thing. Like, are you actually doing that? And I would say if I had to, you know, really be precise about it—maybe it’s not just the act of thinking. But there’s no other time when I can get myself alone in a space where I am forced to think in order to write on a micro, you know, sentence-by-sentence, piece-of-logic-by-piece-of-logic basis, right? Like, I am having to have an extremely honest conversation with myself. Of like: Did that make sense? Does that connect to this in a way? And I think when people say, like, “Yes, man, it’s such a good sparring partner.” Claude, ChatGPT—it’s so good at, like you said, coming up with a counterargument or, you know, taking down your straw-man arguments, whatever it is.
I think, you know, maybe, that is true in some way. But I don’t think it’s as good as having to deal with the honesty of yourself—where you’re like, I have this idea in the back of my head. I need to communicate that on the page. Does what’s on the page match that abstract idea in my brain? And like frankly, I think the act of writing is reconciling that with yourself. And that ability to be alone, like it forces you actually to do a lot of—you know, people find writing so agonizing, sometimes, on the page. Because it’s like, you have to be honest with yourself. Like, Is that my actual idea? Is that, you know, is that any good? Is that worth anything? And that conversation, to me—that’s what I think we’re losing.
Brown: Well, and the models themselves, I mean the learning libraries force any problem that you’re trying to interrogate through one of the chats to look only backward. They can only look at things that have already happened. And I think it limits the amount of novelty you can have in an argument. It limits the paths you can take with it. And I know that Silicon Valley’s very antagonistic toward interiority and introspection right now, but the only way you can get there is thinking about: How do I feel about this? What does this do to me? What am I curious about, and how does this make me feel? And that doesn’t exist in a learning library; that doesn’t exist in a chatbot.
Spiers: Also just at a macro level, now that you mentioned that, I think Silicon Valley has a general anti-intellectualism orientation that sort of gets built into the tech. It’s: I don’t wanna have to, you know, use my precious brain cells on going through rigorous material that might be difficult. I’m gonna build a piece of software that’s gonna do it for me.
And I think the long-term ramifications of that are just sort of bad for society. There’s a cult in Silicon Valley of trying to learn things very quickly, but without friction, you know? And I think if you take those things away, what you’re doing is something that kind of looks surface-level like intellectual development—but isn’t when you scratch the surface. It lacks the rigor, the complexity, all of that.
Warzel: Ezra Klein had a phrase that I liked with this. Which is “the Matrix theory of the mind.” Right? With the idea that, like, some people see information as the Matrix thing, right? You just—you plug, you jack into the, you know, the Borg. And you upload the information, and like: Boom; I understand everything about Kant. Right? Like, I know how to, you know, fly this F-15. Whatever it is. And it’s like the actual process of learning tends to be extremely different. Which is, it’s just about how you, your brain grapple with material, with information, and you think. And I think it’s the disagreement that Silicon Valley has between one style of learning and the other.
I wanted to bring up, though, the argument with all of this with AI-assisted writing. There are folks that have argued that there’s an accessibility issue here. That not all writers either could, you know, can afford to have human editing, or that some brains are requiring outside help or accommodations. Chatbots assist with this. And another way of saying this—without getting into a conversation just about ability or disability—is that not everyone can write well. And not everyone wants to write well. And certainly, most people aren’t professional writers. So how should we be thinking about these tools for a crowd of people that aren’t pros?
Brown: I think there’s a lot of people who see writing as an intellectual costume they can sort of put on, and chat like these chatbots allow them to do that in a very seamless, frictionless way. I’ve been on the record saying this. I think there should be fewer writers. I think there are way too many people who call themselves writers, and should not be in the industry generally. I think there’s sort of—it’s flooded with that. It might be good in the long run, because people who are actually good at this might be able to separate themselves.
Spiers: I think, also, it’s sort of—I think people have this simplistic idea of writing. That, you know, anyone can do it. Just because anybody can sit down and type and do a document, or most people can. But I can’t think of—there’s a whole range of other professional skills that if I said, I’m really bad at this, but I wanna be able to do it, people would look at me like I had three heads.
Like, if I wanted to go play WNBA basketball right now, I’m 5’1” and 49. That’s not gonna happen, and there’s no number of tools that I could add to myself that would get me there. And so I do think, also, people do need to understand that writing is a specialized skill.
You know, it doesn’t matter. I think if you want to do it professionally, that you’re good at it. If you don’t care about doing it professionally, no one’s gonna stop you from writing for your own enjoyment or writing on a blog. I mean that the internet has completely democratized that. You can publish at almost zero cost.
Warzel: You expressed it perfectly—that there are so many other genres of things where if you’re like, I want to do this professionally, and I’m not good at it, the people are tough, right? But what’s interesting and frustrating about the conversation with writing is that we always seem to, culturally, regard it differently. Like clearly—writing is a craft. It’s an art form, but it’s also this thing that everyone does, right? Often in these quotidian ways as we’re talking about. And so I think it’s treated differently than other crafts.
And what I was thinking about in this space was Hollywood and, you know, AI video-generation tools. Like the guilds in Hollywood go out on strike, in part because they want protections from the studios from AI, and standards. And for the most part, people don’t say, like, “In a couple of years, of the all the hand-wringing”—you know, like all these people are being precious. “They’re gonna look back on this and cringe at what they were doing.”
And I think that’s because most people aren’t out there every day making movies and sending them to their boss, right? Like, the way that they write an email or a LinkedIn post. And so I’m wondering: Do you guys think that writing is more at risk than other creative crafts to AI because of its universal nature?
Brown: I certainly think the large middle class of it is, for sure. In the same way that I think the large middle class of moviemaking or the large middle class of graphic designers are at risk. I do think there is a huge contingent of this sort of middle 80 percent of writers who have to make ends meet, not through prestige, legacy media outlets. And those are at huge amount of risk.
Spiers: I think some of it, too, is just that the skill set that you use in writing—and all of the tools available to you, and all of the things that you learn—live entirely in your head. You know, if you’re a filmmaker and you have a wide set of skills, this also involves using things that people see externally as being highly technical. If you’re a writer, you’re using a laptop, which everybody has. And so the idea that there’s a barrier to entry to writing—people don’t kind of understand that. Because they’re like, Well, I could write something and put it on the internet, and what’s the difference? So I think anything that has a kind of zero barrier to entry, people tend to overestimate the ease of doing it.
Warzel: I think we’re all trying to be mindful of, you know, we are in our bubble of “this matters to us.” So as writers, are there opportunities here? You know, like there’s a few excerpts of media that I’ve read where people just start like, you know, talking to the chatbot in the middle of the thing as part of it; it’s almost like a creative-fiction-y thing. I’m just curious if you all see the possibility for opportunities here that aren’t just, like, I’m kinda gonna give a very bland prompt and then get out, you know, a bunch of Claude lorem ipsum.
Brown: There’s certainly—it’s an edge case for me. The sort of avant-garde being like, I think there’s people wanting to read your prompts about how you got something. I think that’s actually like a fascinating artistic case. Like, I don’t care about your writing. Show me what your prompt looks like—because that’s more interesting to me, because it actually shows what your thinking is. I think that what the most likely scenario for me. Or—I don’t know if this is most likely or sort of what I would prefer to happen, is that professional writing becomes more akin to vinyl collecting now. And that this is sort of—not a curiosity, but something for people who really care about it. And that they actually seek out, you know, good writing done by humans. Just because the flood sadly is already here.
I don’t want to be too fatalistic, but it’s just gonna be the state of things where it’s just like: This is too easy. You know it’s too big. The tools are too entrenched. And this is sort of—we have to figure out ways to live with that.
Warzel: Elizabeth, where do you fall on that?
Spiers: I mean I think that you know, if we do end up in a scenario where we’re flooded with AI writing and it’s all crap, that actually makes people who can do things that AI can’t more valuable. Like, for example, AI can’t write jokes. It’s terrible at comedy. And partly because it doesn’t have a kind of meta-sensibility, it’s incapable of understanding the sort of intersection between absurdity and novelty, which is where humor comes from. So I’m not
terribly worried in the short term. I think there is a possibility that we eventually end up with models that are good enough that they can produce novelty, and that they do produce interesting output. And then, I don’t know what happens. You know, I think we’ll all probably adapt to it in different ways. Right now I’m not thinking about that far ahead, about any of it.
Warzel: I feel sort of both sides of this. Where, to me—I said this on a different podcast, but I’m a really bad emailer. In the sense that, like, no punctuation or caps or things like that. And I’ve always had—
Spiers: A Jeffrey Epstein–style emailer?
Warzel: Haha, yeah; I guess, yeah. Just a wonderful comparison to make there. But that’s, like, my email style, and it’s embarrassing, and it’s always been. And I’ve had people comment recently, like: “You know what, man. I know you sent that email. Because it’s hot garbage.” Like, Thank you. Thank you so much for not taking the time to—
Spiers: It’s your stamp of authenticity.
Warzel: Yeah, exactly. It’s authentic. And I see that, you know, more in responses to just my work, personally, and things like that. Like, people feeling that that level of craft as like, as a distinction, right. And so I feel like there is that way which, in the near term, it can really distinguish.
But I just wanted to ask both of you to close out here. If you were made writing czar today and you could set some kind of lists of standards that you would like to see codified for professional writing. In terms of this technology disclosure, all that stuff. Like—what would be just a couple of your first steps? Just things that like, boilerplate, “This is the norm that we wanna impose around AI and writing.”
Spiers: At minimum, I think disclosure has to be a norm. Like, you know, even if newsrooms get to the point where they decide every single contributor can use AI—including to construct an argument—not informing the reader I think is a mistake. Because then readers will easily assume that all of the work of the newsroom is outsourced to AI, and therefore it’s sort of commoditized and not that original. And I think it’s a trust issue.
Brown: I think Elizabeth brings up a really good point, and the journalism thing is a perfect example of this. That there are those cracks in foundational trust for institutions. Listen, there’s a lot of trust issues generally, but putting yet another one into the dam is—it’s really difficult to sort of navigate ourselves away from. On an individual level, for writers, I mean, I hate to say it, but it just doesn’t have a place in my sort of toolkit as a journalist. By sort of filter, you know—Google searches. Sure, Gemini is in my life every day because it’s the top of the search bar. But just, after using it so much in my previous life, I have just sort of been like: This is actually quite freeing to do all this work myself again. And maybe that’s personal; maybe that’s sort of me going from one sort of hard side to the other. But there is something so rewarding about finding stuff out yourself. And I really love that part of reporting.
Warzel: I think that’s a great place to leave it. Teddy, Elizabeth—thank you for coming on and talking about this weird place that we find ourselves in, with this strange typing thing that we started doing a long time ago, that now robots do. But thank you for coming on.
Brown: Thanks so much, Charlie.
Spiers: Yeah. Thanks for having us.
Warzel: If you liked what you saw here, new episodes of Galaxy Brain drop every Friday. You can subscribe on The Atlantic’s YouTube channel, or on Apple or Spotify or wherever it is that you get your podcasts. And if you want to support this work, and the work of my fellow colleagues, you can subscribe to the publication at TheAtlantic.com/Listener. That’s TheAtlantic.com/Listener. Thanks so much, and I’ll see you on the internet.
This episode of Galaxy Brain was produced by Hadley Robinson and engineered by Erica Huang. Denny Thomas edited the video. Our theme is by Rob Smierciak. Claudine Ebeid is the executive producer of Atlantic audio, and Andrea Valdez is our managing editor.
Learn more about your ad choices. Visit podcastchoices.com/adchoices
The post AI Writing Has Entered Its Post-Shame Era appeared first on The Atlantic.
