Sci-fi master Ted Chiang on AI, time travel and more

What science fiction usually wants from AI, from these conscious machines, is actually not exactly a butler but a slave. We want someone who is competent and obedient, and to whom we owe nothing. We don't want someone who makes any demands on us. We want a robot that you could treat like shit if you wanted. Is it right to treat our future robots as slaves? Celebrated author Ted Chang tackles that question in his science fiction stories, and in this episode of The Fiction Science Podcast. Greetings, Earthlings. I'm Alan Boyle, the mastermind behind Cosmic Log, and one of your hosts for The Fiction Science Podcast, coming to you from the place where science and technology intersect with fiction and popular culture. Join me and my co-host, Science Fiction writer Dominica Fetaplace, as we talk with Ted Chang about artificial intelligence, time travel, and how having one of his stories adapted for a Hollywood movie changed his life. If you could see your whole life from start to finish, would you change things? That line, spoken by actress Amy Adams in the movie Arrival, and capsulates the kinds of questions that sparked thought-provoking science fiction stories for Ted Chang. One of his novellas, Story of Your Life, provided the inspiration for the 2016 film. In his most recent short story collection, Exhalation, Chang spins imaginative tales about time travel, artificial intelligence, alien life, bioethics, and predestination. In Chang's hands, these themes go way beyond mere plot devices, delving deeply into the human condition, is one of Chang's trademarks. And that's one reason why he's receiving an award this month from the Arthur C. Clarke Foundation for his use of imagination in the Service of Society. He's also one of the newest members of the Museum of Pop Culture's Science Fiction and Fantasy Hall of Fame. Ted Chang came to the Seattle area 30 years ago to take a technical writing job at Microsoft, and technical writing is still part of his resume. But today, his fame stretches far beyond Seattle. The New York Times has called his stories mind-bending and deeply beautiful, while the New Yorker calls him one of the most influential science fiction writers of his generation. When Dominica and I connected over Zoom with Ted at his home in Bellevue, Washington, we started out talking about the Arthur C. Clarke Foundation's award and the coronavirus pandemic. But we ended up doing our own thought experiments about AI and time travel, and focusing on two of his short stories, one titled Life Cycle of Software Objects, and the other called The Merchant and the Alchemist's Gate. If you haven't read those stories yet, you might enjoy revisiting this podcast after you have. But you don't need to read a word to enjoy this conversation. Thanks so much for joining us, Ted, and congratulations on the Arthur C. Clarke Foundation's Unleash Imagination Award for using Imagination and Service to Society. And there's a special call out to the crises that have been facing humanity lately, between the pandemic and politics and rising concerns about the effects of technologies. Frankly, it's been a crazy year for crises. During the Foundation's ceremony, you will be taking part in a conversation on the themes of crisis and imagination, and I'd love it if you could share a little bit of your perspective about the craziness of the past few months and the role that the creative process can play in helping us make sense out of all this. I don't know that I have an enormous amount of wisdom to share on what has been happening in 2020 because it has been a crazy year, and it has been confounding for me as it has been for most everyone else, I think, as for the role of creativity, and I'll say that rather than talking specifically about the creative process, but I'll talk about the role of, say, art in general. One question that I've seen come up is, does art have any role to play in times of crisis, such as this? Is art important? Can art help us? I think art is important and art can help us in times like these, because art is one way to make sense of a world which, on its own, does not make much sense, art can impose a kind of order onto things, and I think art's power to help us. It is not unlimited, but I think that it is still essential. I guess we would probably be worse off without art than we are right now, so it doesn't offer a cure all, because I don't think there is going to be any easy cure all, but I think art will help us, art helps us get by in these stressful times. So you read a lot of science fiction, and you write a lot of science fiction, and it feels like a very science fictional year, so did being a science fiction fan help prepare you for the weirdness of this year or now? I don't think anyone was really prepared for this year, but a lot of people certainly have been comparing the events of this year with works of science fiction, the various movies about pandemics, or novels about pandemics. So I think certainly science fiction has given people a certain conception of what pandemics might look like, and the actual pandemic has turned out somewhat different than most of the ones that we've read about. Yeah, this is nothing like earth abides. But yeah, I guess science fiction is where we look to for our models of what a pandemic looks like, because I don't think most people have a clear sense of, say, what the influenza pandemic actually looked like, we may have some big knowledge of that. Yeah, like like 50 to 100 million people died, I learned that I'm learning so much about the year 1918 in the year 2020, Woodrow Wilson got it and didn't tell anybody, all these people got it. But yeah, so yeah, that that certainly provides I think a very close precedent, but yeah, it's not something that was really in the popular consciousness very much. I remember in the early days, people were asking, well, is this really bad? How bad was this? Or is this? And one of the experts was saying, well, this isn't the killing blow, but it came pretty damn close. And I was thinking at the time, I think this was in March or April or so, and I was thinking, well, gee, this doesn't seem so bad. It's not like 28 days or any of the science fiction films, but it did turn out to be surprisingly serious in its effects and it makes you realize how potentially fragile our whole economic system is or our public health system. There must be something that you will be chewing on in the months and years ahead about what the lessons of these past few months have been. I guess one thing that I hope people will take away from this is that we've not immediately descended into a kind of every man from self melee, you know, it has not been sort of like the end of civil society, which I think is a fairly common scenario in fiction and movie depictions of almost any disaster. People are immediately like, you know, resort to shooting each other for food or things like that. You know, maybe that's an indication of the fact that this pandemic did not have the sort of swift dramatic severity that a lot of movie pandemics do, but it's worth thinking about that traditional depictions of pandemics don't spend much time, I think, on people sort of coming together and trying to support each other. That is not typically a theme in stories about, you know, disaster or enormous crisis, because I guess the narrative is usually, you know, it's like the end of civilization and people have not turned on each other in that way. The greatest conflict that we see generated is, you know, I guess from people who don't believe in it versus everyone else, and again, you know, that might be a product of the fact that it is not as severe. If it looked like various movie pandemics, it'd probably be hard for anyone to deny that it was happening, and so it may be that people's resistance to believe in it. living in the pandemic has to do with the fact that they are fixated on sort of a very graphic, real disastrous scenario, and anything less than that doesn't qualify. Yeah, that's really interesting. It does seem like this disinformation and denialism is like a big problem in our culture, and it feels like it's becoming kind of bigger. You know, I'm a writer of fiction, and I sort of feel like in one sense, fiction is on the decline because who are people are reading books and stories, but then the other sense, like fiction that marks itself as true is on the rise, because people are believing these incredible things. What can be done about disinformation, do you think? I think that question is above my pay grade. I don't think there's going to be a simple solution to that. Yeah, I think there are a lot of people who are working on that problem. But I don't know if there's a clear path to fixing that problem. If I were a hardcore technologist, I'd probably say that AI is going to fix everything, that we'll have some agent that goes through and figures out what's actually true and what isn't. I know that's the sort of thing that you've written about is what artificial intelligence can do for us or is doing to us. Are there some ideas that crop up in terms of what AI can do, or at least what we can do so that AI doesn't harm? Well, that's a big question. I think Facebook would like to have us believe that AI will be able to identify disinformation and call it in an automated fashion. I'm extremely dubious about that. I'm really skeptical, partly because it's Facebook making that claim. But there is a question of how much is Facebook actually motivated to do that, how much does it fit with their business model to actually suppress disinformation? I don't know that it is. I don't think it actually works with their business model to suppress disinformation. But if they genuinely were interested in suppressing disinformation, would AI be the best way to do it? Yeah, I don't know. Is that a potential fertile field for science fiction to address that issue? You've addressed so many philosophical issues about this in the past. It seems to me like this would be a great place for a Ted Chang story to take it to the next level. Well, okay, so I guess this would be an appropriate time to talk about all the different things we mean when we say AI. When we talk about AI in science fiction, when we're talking about something very different than what we mean when we say AI in the context of current technology. In science fiction, AI means some sort of conscious software, something that is very much like a person, something that has most of the capabilities of a human being and often capabilities that have human beings, but it can be thought of as a person for most purposes. When people talk about AI in the real world, they're not talking about anything like that. They're talking about a certain type of software that is usually a superpowered version of applied statistics, statistical modeling, and when any ad which says, "Oh, our new product powered by AI," what they mean is it's powered by applied statistics. There's nothing like a person or anything remotely like a person, anything that has any of the characteristics that we associate with human intelligence. They just mean it's like it was driven by a lot of data which we ran some statistical analysis on and we're reusing that. It's a very interesting technology, but it's an entirely different technology than what most science fiction writers talk about. The fact that we happen to be using the same term to describe these radically different technologies, one which is real and one which is pretty much entirely hypothetical, that I think causes a lot of confusion. Yeah, thank you for that clarification. I thought that was just really well put. When you mentioned this applied statistics that got me thinking of GPT-3, that's the open AI text completion program, are you familiar with it? I've been reading about it. Yeah. If people ever ask you if this is going to replace human writers, because I get asked that all the time. And what do you tell them when you were asked? Oh, but I want you to tell them now. I mean, I guess, can you tell me what your rationale is before I go? Yeah, that's fine. Well, first of all, I always kind of object. There's always something like a little bit passive-aggressive about someone bringing it up. Do you think that this applied statistics program could do your job? And so I always even resent the question, I wonder if it's being asked in bad faith. So I want to assure you, that's not what I'm doing here. But I think that people, when they approach a story, they're kind of hungering for a human connection. Any type of story. I mean, this type of text completion maybe could have replaced sports reporting at this point if you just want to just report on the statistics in the order that they happened. We could do that now, but people don't really want that. They still want to hear humans give their opinions of the game. So people seem to want a human connection that I don't think an applied statistics program will be able to replace. Yes, I would agree with that. It is of course interesting to see what sort of things people initially think of as being sort of vulnerable to a replacement by AI software. It's usually not the job of the person asking these questions or, you know, people who ask these things, they're usually, yeah, I think they are maybe subtly casting aspersions on someone else's profession rather than their own. Some people will definitely be thinking about its implications for their own profession. But I guess I don't think most people asking that are, or at least I guess in the context that you're talking about. But so, okay, so with regards to GPT-3, I don't think GPT-3 or any of its immediate successors will pose a significant threat. If only because GPT-3 has an extremely short attention span, it can sort of maintain a train of thought over a few sentences, but you know, it loses that over a page. And so even to write a short story, you know, that would require a big leap and to write a novel, to write something like a 300 page novel, that I think that would require, like, maybe an entirely different approach than GPT uses. But more generally, what is the threat that AI generated fiction might pose to human authors? You know, AI is some program written with it, maybe a somewhat different paradigm than GPT. So I guess I don't think of it as being a major threat, and again, talking about, it depends on what sort of timescale we're talking about here. Because let's say someone, some software, is capable of generating a coherent novel. Not a, you know, not a great novel, but a novel that at least stays coherent for, you know, 200 pages or 300 pages. Would that pose a threat to authors? I'm not sure that it would, because I think that, well, like, okay, for one thing, if you, if you think about, like, the price of a book, and how much of that, the price of a book goes to paying the author, it's not that big of a fraction. So cutting out the author is not going to generate enormous cost savings for anyone. So you know, in that sense, it is a form of human labor, which doesn't have enormous market value, and so it's like semi-skilled labor, in a sense. I'm not going to call it semi-skilled labor, but it is something which, I don't know that having AI do it would generate, like, such a powerful cost savings that, you know, we'd have a, anyone would have a strong incentive to do that. If we could get software-generated novels that were coherent, but not necessarily particularly good, I think there would be a market for them, because if you can buy a novel for a dollar, something that is readable, there are plenty of people who would pay a dollar or, you know, or 50 cents or, you know, I don't know, you know, some small amount of money. for a readable novel, but right now I think there has been a sort of a boom in self-published books, self-published fiction. A lot of those are in ebook form though, a lot of those are very cheap, and you know, there is an audience for that, you know, there is a market for that, but I think that there's not put traditional publishing out of business. It's expanded the offerings available, and so there's kind of a wider range of fiction available for readers, but yeah, it is not the main problem that traditional publishing faces. So yeah, so I think that here I definitely agree with you, Dominica, that a lot of what people want when they buy a traditionally published novel are things that will be very hard to do with AI, for an AI to generate a novel that you think of is really good, that you feel like, oh wow, this novel was with both gripping and caused me to, you know, think about my life in a new way, that I think is going to be very, very hard. You know, then we're talking about AI in the sense that we normally talk about in science fiction terms. You might very well need an AI which is as capable as a person is in many ways equivalent to a person in order to write a novel like that. Maybe, maybe not, but I think you'd have to get a lot closer to that and the spectrum. So yeah, you know, it's not impossible, because I certainly don't think that AI as, as we see in science fiction, I don't think that's impossible. I think it is possible for software to be conscious and to, you know, be like a person, but that is a long way off. No one is really working on that. Certainly nothing that is going on currently in AI research resembles that in any way. Yeah, mostly it's, you know, sort of, yeah, how do we build more powerful applies statistical models? And again, you know, that is very interesting and useful and GPT 3 shows us how surprisingly useful it can be. But yeah, if you're looking for a connection with another person, yeah, that's another matter entirely. You addressed this whole question in one of your short stories, the life cycle of software objects. And I'd love to hear a little bit more about the backstory on that short story, because when I was reading it reminded me for, for better or worse, of raising someone who is, for example, on the spectrum, all the things that, that parents have to go through with challenging children. In that story and others, you really addressed this issue of parenting and how parenting applies and how you can twist the paradigm by using a story about virtual individuals. Can you talk a little bit about the process that went into knitting that story together and whether this applies to what we've just been discussing about how the field of AI and the science fiction sense might progress in the real world? Sure. So one of the, one of the things that's fed into my writing that story had to do with, I guess, a kind of dissatisfaction with the way AI is often depicted in the science fiction. And here we're talking about AI that is very much like a person, because a lot of stories they imagine something which is, you know, it's a product like a robot that comes in a box and you flip it on and suddenly you have a butler, a perfectly competent and loyal and obedient butler. You know, that, that, I think sort of jumps over all these steps and questions, because butlers don't just happen. Butlers have an entire lifetime, you know, behind them. Right now, the best way we know of, you know, to make a person, you start off with a baby and, you know, babies make terrible butlers. Children make terrible butlers. How many years of of experience do you think is a minimum before you can get a loyal and obedient and competent butler? 20 years? Maybe more. And, you know, and this is not a, you know, this is not a diss against babies, because you know, babies, they are incredibly smart. It's, it's just that, you know, if we could, if we could create an AI that was as smart as a human baby, you know, like, yeah, in some ways, like problem solves, you know, because that's, that would be an incredible achievement. But yeah, you know, that, that incredible achievement, a, an AI that is as smart as human baby, that does not give you the butler that you can just flip a switch and, and you wake up. You're still at least 20 years of, you know, training away from, from your, you know, from your butler, the butler that you're looking for. So yeah, I guess you know, one, one of the things that I was sort of interested in exploring in, in my story was, what does that 20 years look like? How do you go from something which is as smart as, as a human to something that is the useful worker that science fiction usually wants from AI? What is required to get there? And then if the things that are, that have to happen over the 20 years, are those things that will ultimately make it very difficult for this AI to, like maybe, like, maybe they'll be people so like, I don't want this AI to be a butler or like maybe, maybe I want more up in this AI because someone is going to have to invest a lot of time and effort. You know, like if you were to try and pay people, it's like, okay, you give someone, here's, here's a human baby. And I'm going to come back in 20 years. And you'll, you'll give me back someone who is a competent and obedient butler. Like, how much would you have to pay that person to do that for 20 years? You know, you probably have to pay a few people to split up the work. But that's going to be a lot of effort. And is that something that anyone would undertake just for money? Maybe the only way that you can undertake that, you know, that amount of effort is, if there is an emotional investment too. Because pretty much every person who is a good worker, they had parents and if they had parents who loved them were good, you know, good parents, you know, that might actually contribute to them being good workers. Because you might think like, hey, you know, people who are raised by really negligent parents, maybe they don't make good workers, maybe those are, maybe they don't make the best employees. You know, which is not to say that someone who is unloved by their parents will never amount to anything. I don't, I don't mean that. But I mean, I think that, you know, loving parents, you know, I think definitely increases the likelihood of a child turning into an adult who you want to hire as an employee or you want, you know, someone who basically is good at the things that we want human beings to be good at. Loving parents probably will increase the chances of that. Then when you have loving parents or some, you know, some equivalent, then there, there are issues like what science fiction usually wants from AI, from these conscious machines is actually not exactly a butler but a slave. We want someone who is competent and obedient and to whom we owe nothing. We want a robot that could treat like shit if you wanted. I mean, certainly one that, you know, like, you don't want to pay, you're not going to pay them a lot. You're not going to give them, you know, a great salary. Right. Just plug them in and use them. Yes. So, yeah, that's what that's what people sort of want from their robot butlers. But if someone invested emotionally in that for 20 years, like, well, then maybe they don't want that to be treated like shit. They want that robot or conscious software to have some kind of autonomy to have a good life. You know, we don't even have to go with the parenting analogy so closely. If you have raised an animal for many years and then you're going to give up that animal to somebody, you will want that animal to probably go to someone who will treat it well. You don't want it to go to someone who is likely to abuse it. If you've trained an animal, you've trained an animal to be, you know, a seeing eye dog or a bomb sniffing dog or anything like that, you have probably invested emotionally in that animal. And even from a purely practical perspective, it's like this seeing eye dog is not going to do the best job at being a seeing eye dog if you abuse it. This bomb sniffing dog is not going to do the best job at sniffing out bombs if you abuse it. So, those are some of the questions that I was interested in examining with my story. Everybody probably asks you how your life has changed since the movie Arrival came out and you're probably pretty tired about answering that question, but I think I had to try asking the question in any way. Is there anything you want to say about that? I don't know that it's changed my life in a dramatic fashion. All I can say is that it's been great to see my work reach, you know, much wider audience. Yeah, a lot of people saw the movie and a lot of people bought my collection of short stories after seeing the movie. I'm sure it also prompted people to buy my new collection as well. So, yeah, my work has been read by people who would not describe themselves as science fiction readers, but people who don't usually read a lot of science fiction. And that's been really gratifying. It's not something that I ever really expected. I also thought of myself as, you know, a science fiction writer writing for science fiction readers. So it's been it's been very cool to see that my work is connecting with people outside of the field that I came up in. Yeah, and they just wanted to add, I think, not only bringing new people to science fiction, you're bringing new people to short fiction. So as a writer, short fiction myself, I really appreciate that. Yes, that's, you know, that has been great as well because, yeah, it's, yeah, I think a lot of people, you know, they only read novels and they are, you know, they tend to overlook short fiction. And I think, yeah, they often don't think of themselves as short story readers. So yeah, it's, that has also been gratifying to see that, yeah, my work has maybe, yeah, also reached a sort of a crossover audience across those lines as well. Well, I really enjoyed your latest collection of stories, exhalation, and I'm holding my breath waiting for more. One of the themes in at least a couple of the stories and exhalation have to do with time travel and the deep personal issues that come from thinking about time travel. If you were to time travel back to an earlier point in your career or your childhood or whatever, are there things that you would like to tell your earlier self as some of the characters did in one of your stories and exhalation? No, I don't think so. You know, well, okay, so in the story you're referring to Merchant and Alchemist Gate, for one thing, you know, like the character he remembers hearing this information from his older self when he was young. And so, yeah, that's an entirely different, you know, scenario because, you know, like, I was not visited by an older version of myself when I was young. So yeah, so I certainly can't say like a kill. Oh yeah, yeah, that was the time. But, you know, I guess more generally, I guess I'm, you know, I guess when people play this, this sort of thought experiment, I guess, you know, it's not really clear to me, like, is there a time in your life when you think some advice from your future self would have actually radically changed your behavior? I guess I'm not sure how often that's the case. You know, I'm like, okay, so not speaking about me personally, but like just in general, like if someone had gone back, like, you know, like, it's just like, oh, like, if only I had known, like, I would have majored in this rather than that. I don't know, if, if, when you were entering college, if your older self came to you and said, don't major in this major in this other thing, would you immediately say, well, okay, I'll better do that. Like, I think you'd probably like, no, why? I don't, you know, like, you know, I'm so confused. Yeah, I don't know that it would have the desired effect. I suppose, as in the story, it's more about going back to that human connection. In fact, maybe going back, if you had the chance, you could tell your younger self, you know, you're going to go through some crazy things, but it's all going to be okay. That might be, that might be what I do. I don't know how much good it would do. I mean, maybe, maybe, but when you are in the, in the middle of suffering, hearing that eventually your suffering will end and, you know, it'll be okay, that doesn't make your suffering go away. Like, okay, someone who had an unhappy childhood, you know, if you went back and said, like, don't worry, you're going to be happy adult. Like, how, how is that useful? That's not going to solve their problem. That's where the tips come in. I suppose. I don't know. Yeah, but what tips could they possibly act on that would, you know, improve their situation then and there? I, I guess it's never been clear to me, you know, what form these tips would take? Do you want to say like, what would you tell your younger self, like, buy Apple stock? You know, like, that's not going to pay off for a long time. If you were miserable in high school or something, you know, I'm not, I don't know what, what piece of advice would help your younger self. You know, at least it's hard for me to think of that. Maybe, maybe, I don't know, Domenica, can you think of a point when, like, you were unhappy as a child and, like, some advice from your future self would have really, you know, cheered you up or made you feel better? You know, probably, I do think I kind of enact this exercise as a writing teacher, though, and as a math teacher. In a sense, I'm trying to teach my students, but also I find myself really talking to my younger self sometimes. Well, okay, so when you're talking to your younger self, I mean, I'm imagining that that helps you. Yeah. Yeah. It's not great actually because when you're teaching someone, you're really supposed to be looking at them and their situation and not, not projecting your own stuff, but it just, it just happens and you just sometimes slip. Yeah, as well. See, that's, you know, that's a difference. That's a whole different exercise, you know, like talking to your younger self might help you, but doesn't help, you know, it doesn't help the younger self. That's right. Yeah. I guess that just kind of supports your point. You know, if we're talking about a time travel situation, yeah, like, that's a very different type of conversation, I think. There are definitely some times when I would go back and tell my younger self, listen to what your dad is saying. I should have taken that advice when he was giving it. And do you think you, do you think, would that work? I think it would. I think it would. Yeah, I think it would because at the time when I was thinking, he's telling me something, I, you know, I think it could have pushed me over the edge. I mean, I would have been very impressed to have a visit from my older self. That would just blow me away. Maybe, maybe, but I also wonder like if, you know, maybe we were just all very resistant to advice from older people when we were young. I don't know that, you know, why would we be more likely to take advice from someone claiming to be our older self than we are from our parents who, you know, we have known our entire lives. Wow. We could keep going on this for another couple hours, I think, and over a beverage of your favorite choice. But I'm loving this conversation. I, I'm a little regretful to call it to an end. Maybe my future self will come back and tell me, oh, you should keep going. But I think we're going to have to close it off there. I really appreciate this and and really recommend to anybody listening out there to take a look at exhalation stories by Ted Chang and look forward to the next adventure. And listen to your future self. Well, I mean, actually, if you get a visit from your future self and his only piece of advice is to is to redo this interview, then I think that votes really well for the future. If that's the most important change that he wants to make, it's like, well, you know, like, okay, I think that's super, super reassuring. To learn more about Ted Chang and his writings, including exhalation as well as some of the thought-provoking essays that he's written recently, check out the Cosmic Log posting at FictionScienceClub.com. And while you're online, check out DominicaFetaPlace.com. Don't worry about spelling. You can follow the link from Cosmic Log. One of Dominica's short stories is included in volume five of the best science fiction of the year, a collection edited by Neil Clark. I'd like to thank Ted for the interview and thank James Emily for his rendition of the Cosmic Log theme. We'd love for you to subscribe to the FictionScience podcast and feel free to give us a stellar rating on your favorite podcast channel. Until next time, this is Alan Boyle, advising you to watch The Skies.