I want to talk about AI. It's their confident wrongness that's endlessly kind of amusing and/or dangerous.
It's also devaluing a lot of the skills that made nerds valuable, like the ability to code and just sit there and do it and get into flow and stay there for 24 hours with your Cheetos and your Mountain Dew by your side.
I guess what doomers and boomers both have in common is a very high opinion of how powerful the technology actually is. I don't share that opinion.
I definitely expected to be on Mars by this point in my life and I'm a little sad that didn't happen.
Yeah, me too. I know.
We need to talk about the future. It feels like we're living in the most uncertain time I've ever experienced as an adult. Our politics is melting down. AI is disrupting, okay, not everything, but it feels like almost everything. Geopolitical alliances are shifting. And there's 85 flavors of Oreos and they seem to be multiplying. You'd be insane if all that change didn't make you a little nervous. But when I'm tempted to fall into a death spiral of anxiety, I also remember that uncertainty and new things and creative destruction, that's where all the good stuff about my life came from. Disruptive ideas, tearing down the old and replacing it with something better. Frightening, but also the root of all progress, even if it's a little scary or a lot scary when it's happening. And I'm lucky to live in America, one of the most disruptive ideas in human history.
So, this show is about how we make a better future for us and all the generations after us who are going to be living in our future. I don't think that's going to be easy, but I am very confident we can do it, which is why I'm launching a show called Reasonably Optimistic. No panaceas, no partisanship, no simplistic promises of solutions that aren't really solutions. It's just working through problems together with you and with a lot of forward-looking thinkers who will join me. For our first episode, I want to talk to someone who's been living in the future longer than most of us. Neal Stephenson has been writing epic science fiction novels for three decades that have not only identified a lot of what's coming, but inspired the people who built it. The metaverse, cryptocurrency, wearable technology. In addition to writing, Stephenson is a futurist who's been on the ground floor of startup companies like Magic Leap and the space exploration company Blue Origin, which in the interest of full disclosure is owned by Jeff Bezos, who also owns the Washington Post. And I could think of nobody better to talk to about where to find hope in the technological future and how we get there.
Neal Stephenson, welcome to Reasonably Optimistic.
Good to be here. Congratulations on your first episode.
Thank you. I'm very excited to talk to you as well as to talk to other people in the future. And I said in my introduction that you've been living in the future longer than the rest of us. Of course that's a little bit of poetic license. You've been living in a future, actually a lot of them, as well as an alternative universe that often bears some startling resemblances to our own as well as a lot of differences. So can you talk to me about what it's like to try to live in a future?
You know, at bottom, I'm a storyteller. My responsibility is basically to entertain people and keep them turning the pages. If you're writing science fiction, you sort of have to make the story hang together in a plausible way. It's got to make sense. You can get readers to go along with just about any crazy idea in the opening few pages of a book, but after that, you have to keep telling a consistent story. So, a lot of what I'm trying to do in writing these books is to be pretty meticulous about thinking through the way these hypothetical technologies would work. But beyond that, thinking through, you know, how would it make sense as a business, how might it make sense in the world of politics and society? And I think that if you can kind of succeed at those, it keeps people turning the pages, like I said, but it increases the odds that people are going to look on your work favorably as a, you know, predicting the future kind of thing.
Yeah, I think that's one of my favorite things about your books is that there's a lot of just thinking and then what. And I think my least favorite writers in general, but especially science fiction writers, are the ones who seem to start out with having some giant usually political point that they want to make. And then all of the characters and the plot have to be shoehorned into that destination. And the thing I love about your work is that it feels like you started with the idea and then thought about, okay, but then what would happen? What are the implications of that? You've done this so well. I want to ask you what is it about you other than your gifts as a storyteller, which are considerable, but what is it about you, about your background, about anything else that allowed you to think and then what so compellingly that some of the things have actually happened?
I'd say it's a few things that came together kind of fortuitously. One is sort of growing up in a college town where, you know, everyone around me was a sort of science and engineering PhD or studying to become one and I didn't realize that that was weird. Okay, that was just the normal way people were in my family and in my community. So I came out of that I think feeling comfortable with the engineering mindset and kind of the scientific mindset and with that basically kind of optimistic view that, you know, we're all part of a rational society and we're coming up with solutions to problems and building things that work using our rational faculties. So that's kind of the base layer that I came out of as a young person. And part of that is the sort of reading classic golden age of science fiction type science fiction stories, which all kind of reinforces that way of thinking. So I think a lot of what you were referring to earlier in your remarks, you know, is kind of downstream of that.
The idea that if you're really going to write a science fiction story that rings true to a modern sophisticated audience, you have to jettison this idea that the inventors of the technologies are going to have their way every time and that the way that the technologies get used are always going to be what they predicted, and you have to embrace the idea that once it gets out into the world, people are going to find their own ways of using it.
Yeah. So one of the technologies that's been changing us a lot over the last decade is social media, which has kind of melted our politics and I would argue, I think you would argue, many people's brains.
But that's been true of a lot of disruptive communications technologies, right? The printing press comes and suddenly we get wars of religion, we get the Reformation. One of the early things that books are used for is to write accounts of witches, which lead to witch burnings. And then you get the telegraph and suddenly society seems to implode. And again with the telephone and radio and even television, I think you can argue that 1968 is a lot about the fact that protests look great on television.
And all of those things are really hard, but our culture does adapt and I think ends up in a better place at the end of the line. Do you think we're adapting to social media yet or can we adapt, I guess is the better question.
Yeah. You know, another one in that list is loudspeakers. If you see what happened, like I was reading some historical stuff about the Spanish Civil War and just the ability to hang loudspeakers from trees and shout things out of them, you know, had a transformative effect on the politics of small communities there and in other, I think the same thing happened in Indonesia. But anyway, the point is the same. New communication technologies tend to get picked up by people who have things they want to communicate, you know, and at the top head of that queue are politicians. And there may be some initial lag, but before you know it, they're all over these technologies and they're using them to get power and to make the changes they want to make.
Definitely sympathetic to the idea that this has all happened before. It's caused disruptions before and people adjust. You know, you could make a contrarian argument that social media is more insidious because it's changing all the time in real time in response to how we use it and what we do. And so maybe it's got an adaptation loop that runs faster than we can adapt to it. So I think the jury's still out on that. I'd like to think that people are more skeptical about screens and social media than they were, you know, 10 or even 5 years ago. We're seeing, you know, some momentum developing around, say, the works of Jonathan Haidt, who's talked about kind of the way social media warps your brain. I think we're starting to see some pushback now from parents who are trying to delay the age at which their kids get exposed to social media. I think ironically that's more common in elite social circles. Like the parents who are actually working for tech companies are the ones most likely to not let their kids have cell phones until they're 16 years old or whatever because they have a better understanding of why these things are dangerous.
This actually brings me to The Diamond Age. I have a small confession to make, which is that I worked for your publisher very briefly when Diamond Age was in galleys and so I first read The Diamond Age. Yeah. I eventually, I was an editorial assistant at Bantam Doubleday Dell. I got fired after three months by an editor who said, 'I think you'll make a very good editor someday, but you are a terrible assistant.' Entirely justified.
But one of the things I remember even then, and much more strongly preparing for this interview, I went back and reread it. And it was actually, I have to say this is not usually the case, but when you're in a magical book and you're really deep in it, usually you only get that once, but it turns out if you wait 30 years, you can have that experience again. And it's terrific.
And that's one of the themes of The Diamond Age, right? Is the split between how the elites are handling technology for their kids and how the masses are getting it. Because you've talked about the ways that AI is melting people's brains. It's causing education to just not function as education.
Unless all you're trying to do is teach kids to ask a GPT something, which is not what education should be. It should be about struggle.
It should be about like getting things wrong a lot.
So can you talk about first how you did think of that 30 years ago and how you think about it now? You know, 30 years ago you were calling it pseudo-intelligence. Is that still how you think about AI and still how you think about the technology and cultural problems it presents?
Yeah. So I mean it started with the idea of, you know, a very simple idea of, you know, a girl finds a magic book that changes her life. The idea for it came from a mobile. My wife and I had a baby at the time. And so we had this mobile that would hang above the crib and it had these replaceable placards on it that would move around and in the first version they're very simple crude geometric shapes. But then as the baby's visual system improved and they learned how to focus on more detailed things, you would pop those off and replace them with more complicated cards. And so I just sort of decided to extend that idea to a whole book that would adaptively change, you know, in response to the changing needs of the child. And it kind of developed from there. You know, I at the same time had been sort of reading a lot about the Victorians. You know, steampunk was a thing at the time. And so I was thinking about their attitude towards educating young people, how it might differ from what you'd see in other cultures.
Can you talk about that a bit? You know, the elites sending their kids off to these actually quite miserable boarding schools where there wasn't enough food and the big kids used to bully the little kids at the behest of the administration.
Yeah. I mean well certainly, you know, to use your phrasing, the idea of struggle was a big part of it, but I mean that's not so much, that doesn't find its way into The Diamond Age so much, but there is this idea of the importance of teaching kids to be self-reliant and think for themselves and the inherent contradiction between that and what we normally think of as Victorian morality. We think of it as very stuffy, rigid, you know, always the same. But at the same time, it was a very dynamic society that treasured and rewarded people who had the vision to achieve great things. So how do you square that circle? You know, it's a kind of a perennial issue with no clear answer.
So in the book, the guy who starts the whole project is a very wealthy successful elderly guy who, you know, he's somebody who today in our generation would have been a tech founder, but now we're, you know, we're 60 years in the future and he's done all the things one does in educating, you know, privileged, you know, affluent kids and sent them to the right schools and made sure that they worked with other kids like that. And he's not happy with the results. He doesn't think that his kids came out of all of that with the same sort of dynamism and iconoclastic spirit that accounts for him being rich and successful. And so he doesn't want to repeat that mistake with his grandchildren, but he knows he can't just directly intervene in how they're educated. He knows that his kids are going to give his grandkids exactly the same kind of upbringing, the same schools and all that that produced unsatisfactory results the first time around. So, as a kind of subversive technique, he has an engineer make this book, this magic book, and he gives it to one of his granddaughters as a present because how could the parents object to grandpa giving a book to the granddaughter? And so that's all kind of his complicated plan. But along the way, there's a couple of extra copies of the book that get made and given to other little girls of about the same age. And, you know, the results vary, right? So, in a couple, you know, the book doesn't always magically have the results that he intended, but it does fall into the hands of one extremely underprivileged and abused little girl for whom it, you know, completely transforms her life.
Going back to what we said, right, that it's hard to predict what the results will be and that there's tradeoffs, right? It's never that this thing is either perfectly great or perfectly awful. So I want to talk about AI and where AI falls on that spectrum because I think right now the conversation around it is a somewhat unhelpful dialogue between the doomers and the boomers where it's either going to usher in fully automated luxury communism, or it's going to be the Terminator or the Matrix and we're all going to find ourselves, you know, huddled in the spandrels between various large robots trying to kill us. Where do you fall on the spectrum? How do you see it?
I guess what doomers and boomers both have in common is a very high opinion of how powerful the technology actually is. I don't share that opinion.
Yeah. I mean, at the end of the day, these pattern matching machines don't have an actual model of reality that is reliable that they can sort of test their outputs against. I mean, these things are changing and getting better all the time, but like the classic one from a few months ago was, you know, asking a large language model how many Rs are in the word strawberry.
And they'll very comfortably...
No, no, but it's a simple, like all you have to do is count them, right? I mean, just go through and count them. But that's not what they're designed to do. They're designed to do other things. It's their confident wrongness, you know, that's endlessly kind of amusing and/or dangerous. You know, I was talking to a lawyer friend the other day who said that he was using one of these systems to prepare briefs. I said, 'Well, isn't that really dangerous? I mean, they come up with all kinds of hallucinated references and they say stuff that's wrong all the time.' He goes, 'Oh, yeah, of course. I mean, they're totally wrong.' You know, so you just have to throw away all that stuff and ignore it. But in between all of the wrong stuff, they do a pretty good job of summarizing certain things and aggregating information. So, I thought, okay, well, that's a really telling, you know, point of view on how to use AI. You have to already know enough about the subject matter in question that you know like which 75% of the output should just be thrown away.
I think Charles Mann just had a very similar story that he told a couple weeks ago on his Twitter feed, you know, where he had used, he had just as an experiment asked ChatGPT or one of those to kind of effectively do some research for him and immediately saw that he was wrong. And then all of these AI advocates sort of jumped on and said, 'Well, sure, you just have to discard the wrong stuff.' You know, and so Charles was left saying, you know, okay, so AI is great for those of us who don't need it.
Well, I mean, I think that's true, though, right? I periodically try to get an LLM, which is not the only kind of AI, but the kind of AI that is most relevant to my job right now, to write a Megan McArdle column. And I am pleased to say that so far it has not come up with anything that I would consider on par with my own work. Although perhaps my readers would differ. But it is actually very useful as a kind of super Google search. The phrase 'please provide sources' very essential. Do not repeat anything it says without checking the link. But it's good at finding me things.
It is great for when I have writer's block and I'm having trouble getting started. I just tell it to write the column and then I hate the column and I'm so mad that I can then get started writing the column.
It's also really good for fact-checking and for kind of checking my logic. It's terrific for having weird conversations about ideas I'm playing with and that my husband is sick of hearing about. But I'm not sure it's good for kids. You know, you recently wrote that we may end up with at least one generation of people who are mental weaklings utterly dependent on technologies they don't understand. And so I think the question is how do we balance that? How do you think about getting kids, they do need to know how to use AI at some point and they need to be good at it, but they first need to develop some thoughts. I've always hated this thing that educators will say of 'we're not teaching kids facts, we're teaching them how to think.' It's like you can't think about anything unless you have some facts in your head to think about.
So how do we balance that? How do we build the strength that we need?
Yeah. So, and by the way, it's not just that those kids end up being dependent on the technology, but by extension, they're dependent on the people who own and operate those companies, right? So, just to state the obvious. So, I mean, I'd like to think that the people who run these big companies who've raised all these hundreds of millions of dollars could put a little effort into solving that problem. I mean, it just so happens that, you know, ChatGPT and other such programs just because of the way they were built are perfect at doing homework. And so kind of it's just an unintended consequence of that. The entire system of handing out assignments in school and grading students based on that has totally collapsed. It does not exist anymore and there's nothing we can really do about it short of throwing out that way of doing things. You know, we can say, well, all exams now should be written out by hand in blue books and that sounds great and would probably work. You know, the resources for teachers in most schools have been cut back to the point where it's just out of the question for them to sit there, you know, reading through all of these essays and grading them one by one.
Even if the kids can write, which they can't.
Even if they have penmanship that's readable. Yeah. So that's where we get into this whole generational, you know, problem that, you know, even to implement the write it down in a blue book strategy, you first have to have started teaching kids penmanship five years earlier. You know, I have a friend who's a college professor, who teaches online classes, is an economics professor, and I was talking to him about this problem a couple years ago when it first popped up, and he started seeing people cheating with it. And we were talking about how someone said to him, 'Well, you should just give oral exams.' He said, 'I have 2,000 kids in my class. If I gave all of them an oral exam, I could talk to each of them for 10 minutes a semester.' Right? And that's not how it has to be set up. But our institutions just assumed a certain amount of friction with writing, right, that no longer exists. And so it's going to be a substantial effort.
Right? And then they removed friction by, you know, everything went paperless. Everything, you know, became bits on the screen. That has completely imploded. I'm trying to remember the exact name of it, but Niall Ferguson published a piece a few weeks ago called something like 'The Cloister and the Starship'.
Yes. The Cloister and the Starship. That is correct.
Yeah. Okay. Got it. So the idea being that, you know, a way to properly organize education going forward would be that, you know, eight hours a day students are in the cloister, meaning, you know, you just have books, no electronics, no internet, and you learn by reading, writing, and having conversations with other human beings. And then the rest of the time they're in the starship, which just means, you know, completely unfettered access to the internet and AIs and all of that stuff. When you're finished with your education, you're good at doing both of those things, right? You know, and so that all sounds great. The obvious rejoinder, which I'm sure Niall is well aware of, is that that takes more teacher time and more resources than we are currently allocating as a society to education. We are going to have to really do some deep work on our institutions to figure out how, first of all, how we should be teaching this stuff and then second of all who should be teaching it, training those people, selecting the people who should be doing it. It's going to be a big project.
And I want to actually talk a bit about who this project benefits because when I look at your books the central conflict is often kind of nerd world versus normies, right? They are the autodidacts, the obsessives, the people who want to dive deep into some weird obsession. And as a kind of lesser nerd who has a lot of nerd-like obsessions but never learned to code, I've been looking at the world for the past 30 years and it seems like, do you remember the movie Revenge of the Nerds? It came out when I was a teenager.
I feel like the last few decades have been the Revenge of the Nerds. It's been a time when having that particular set of skills has become much more valuable than it used to be. I am a journalist because blogging became a thing at a time when I had free time to do it and that was a space of about 5 to 10 years when that particular set of skills was useful. When I look at nerd world now, I can tell two stories about where they're going. Not the ones with capital who are going to own OpenAI, but just the normal people who go to Silicon Valley, they become coders, they do tech things, they build robots. And one story is that AI is great for autodidacts. On the other hand, it's also devaluing a lot of the skills that made nerds valuable, like the ability to code and just sit there and do it and get into flow and stay there for 24 hours with your Cheetos and your Mountain Dew by your side, or I guess now it's more a Jamba Juice and some like chickpea snacks. But who benefits? Do the nerds get better because GPT enables them to go farther on their own or do they get worse because the skills that have made them so valuable and the rise of Silicon Valley are being devalued by AI?
Another word for that dichotomy you're talking about is the shape rotators versus word cells.
