If we look really closely, really copiously towards the future, unclouded by our own moral preoccupations and projected desires, what might we find there? And/or, how could even the most dispassionate such gaze fail to reflect back present-day biases? This closing-out-the-year conversation (transcribed by Phoebe Kaufman) focuses on Robin Hanson’s book The Age of Em: Work, Love, and Life when Robots Rule the Earth, which articulates the most kaleidoscopically coherent vision of a futurist society that I ever have encountered, a vision ranging across an impressive array of intellectual and artistic disciplines. Hanson, an associate professor of economics at George Mason University, and research associate at Oxford University’s Future of Humanity Institute, has a doctorate in social science from Cal Tech, master’s degrees in physics and philosophy from the University of Chicago, and years of experience as a research programmer at Lockheed and NASA — and writes prose that pivots just as nimbly from one field to the next. His diverse research interests have produced papers on, among other topics, and according to his own intriguing inventory: spatial-product competition, health-incentive contracts, group insurance, product bans, evolutionary psychology and the bioethics of health care, voter-information incentives, incentives to fake expertise, Bayesian classification, agreeing to disagree, self-deception in disagreement, probability elicitation, wiretaps, image reconstruction, the history of science prizes, reversible computation, the origin of life, the survival of humanity, very long-term economic growth, growth-given machine intelligence, and interstellar colonization. Our talk took up at least half of those concerns.
ANDY FITCH: First I want to introduce this book less as pointed forecast than as exemplary model for what you have called a serious futurism. You give the odds of a completely accurate forecast here as less than one in a thousand. So rather than stake its claims upon something like prophetic vision, The Age of Em seeks, it seems to me, to think through a futurism as detailed and methodologically rigorous as the field of history. I very much appreciate for example your attempts to describe the future close up (its plumbing, its clutter, its muddled ambiguities and ambivalences as shaped by the material conditions of a given moment), as opposed to offering some more abstracted account that seeks to project stark moral clarity.
ROBIN HANSON: There’s a lot of overlap and commonality with the kind of futurism I’m doing and historical analysis (especially economic history). I know many economic historians here at George Mason University. I enjoyed taking economic-history classes as a grad student at Cal Tech. I find that economic historians often try to summarize an entire place and time. They describe all aspects of the society that they can, even if they have to speculate a bit to address topics on which they have less data — which I appreciate, because I am trying to do that same sort of thing with the future.
Historians do sometimes set limits on themselves. For example, traditional historians might claim to just describe what they see, without being very theoretical. Of course they are in fact using a great many theories. We do have a number of concrete documents, artifacts, fossils, and we do use those to help interpolate, to guess the past — but we also use a lot of theory to try to fill in the details. My way of studying the future is pretty much the same thing, except there are fewer details to anchor on.
That means you need to know a lot of theory. You need to know a lot of our best stories about how various parts of civilization work, because that’s what you’ll mostly be using. The other thing I anchor on is concrete assumptions (i.e., I break the future into different possible scenarios, and I establish a scenario by a set of defining assumptions, and then anchor on that set to make predictions given those assumptions).
You seem to have great respect for disciplinary rigor, but also to remain skeptical of any starkly defined disciplinary boundaries. You describe yourself as “a scholar in general” rather than “an economist in particular.” You discuss wanting to leave a legacy less through launching some paradigmatic revolution in one chosen field than by building consensus across a variety of fields. I’d love to discuss that particular goal or gesture. And for added lived context, I’ll quote from your preface: “mostly I’m hungry for human communion. I’ve never felt as intellectually isolated or at risk as when writing this book, and I hope my desert days end now.” Do you wish for a recognized discipline in which writing this book might have felt less isolating, or does it make you just as happy to stay outside such professional trajectories?
Well if I want to spend a lot of time learning about things that I think are interesting and important, and I’m willing to spend my personal time pursuing that hobby, then an academic career does happen to be an excellent complement to that hobby. Academia gives you many excuses and opportunities to learn many useful skills. It gives you contact with people who know many things and may also have related hobbies. However, academia’s main products, I would say, are credentials and impressiveness. Its customers are students, research patrons, and media reporters, who basically choose to talk to and reward academia because it credentials people, and people want to affiliate with that credentialed impressiveness.
Academia, for example, is broken into these disciplines which reward people for doing well within a subfield, but it tends to discourage research between and across disciplines. Academics tend to do very impressive work using overly complicated methods, with people repeatedly studying the same few topics and leaving many other topics neglected.
So that’s a long prelude to saying that, in The Age of Em, I honestly want to tell you about what this future could be like, and to do so I’ve drawn on a wide range of fields. In each field I’m just drawing on the most basic results (I’m not drawing on very complicated tools or results). But I am drawing on an unusually wide range, and offering an unusually thorough integration of these results.
To begin to focus on the book itself: one central tension seems to arise between its microcosmic and macrocosmic elements. Like, say, Thucydides, you use the study of one particular historical pivot-point to bring in any number of potential outcomes, reflective perspectives, disciplinary methods. Like Thoreau, you offer a total sounding of one localized environment (Thoreau says: “I have traveled a good deal in Concord”). And in terms of your preface’s claim to a sense of lived isolation, your meticulous construction of futuristic cityscapes does somewhat remind me of exhaustive, self-perpetuating, self-referential projects by certain American folk artists who end up getting both admired and deemed insane. But also, like Montaigne, you start from your own ample private library, then expand forever outwards. You offer a comprehensive social sweep reminiscent of Plato’s Republic. But I most want to mention, if you can believe it, epic poets. Epic poets get praised for their displayed mastery of many different idiolects (basically, the specialized language of a given subfield). And The Age of Em offers something like an epic array of social-science theories (in contrast to sci-fi’s often fetishistic, more naive or parochial approach to the future). So when your preface wearily admits that this book “reads more like an encyclopedia than a narrative,” I instead think of epic poets, or of an author like Roland Barthes when he offers “the novelesque without the novel,” quotidian realism without dramatic plot.
Well I’ve long enjoyed science fiction, but I tend to enjoy the setting more than the plot or characters. Many science-fiction novels spend roughly the first third establishing the setting, and then the last two thirds are a long chase scene, and it can be better to just stop reading right there. So I’m happy to just be providing a setting. Now, you made comparisons to epic poets and people who make large integrated artistic constructions, and there is a sense in which the intellectual world has less use for those today. There is more of a division of labor, and people focusing on doing little parts of a collective effort — rather than individuals constructing large complicated projects that hang together with more total coherence.
Of course people are sometimes celebrated for defying that trend, and constructing a large integrated performance. And there’s a sense in which this is a special kind of impressiveness that you can’t show in any other way. However, there is the concern that, in your attempt to construct a large integrated work, without fine detail, without connecting to lots of other people doing more focused studies, from their point of view you run the risk of losing touch with reality. You are wandering off into strange places where you haven’t been validated, and haven’t checked yourself against other people’s critiques and contributions. I do take that potential criticism seriously. So instead I’m trying to mostly direct my contrariness to taking this one big assumption and working out its details.
There’s a whole field of science fiction and futurism where this one assumption of the emergence of emulations is very standard. But it’s unusual to take a very standard assumption in science fiction and futurism, and apply very standard social science to it. I’m at risk of becoming one of these iconoclast eccentric types who just goes off in way too many idiosyncratic personal-judgment directions. So I try to prevent that by focusing on mainly applying standard-consensus results (i.e., taking a standard-consensus scenario and applying standard-consensus social science to it), and that I hope has protected me somewhat at least from the risk of being useless because I’ve lost touch.
I approach The Age of Em largely from a literary background, with you facing no threat of appearing too exploratory or too eccentric. But at one moment you mention an English professor glibly explaining why English professors always hate economists, and I do want to outline here, for a more literary audience, some of The Age of Em’s clearly defined analytic parameters. I sense many of these parameters might alienate literary audiences (from you, from me for uncritically accepting them), but you clearly state them and build from there, and we can do the same. You assume (different, of course, from advocating) a low-regulation, low-coordination future, with most macro developments emerging as the aggregate of countless individuals’ localized, self-interested decision-making. You do this in order to allow for a clean, logical transparency in subsequent stages of your argument. You also presume the historical continuity of Malthusian living conditions, with maximalized population growth reducing resources (and here often wages) to a subsistence level for almost all beings. You set aside the last hundred years or so of Western-style bourgeois capitalism as offering a great anomaly, a dreamtime in which population growth has not kept up with productivity growth — thereby offering a surplus of wealth, leisure, social tolerance. You consider this dreamtime mode historically unsustainable, soon enough eclipsed (regardless of whether ems emerge) by greater fertility rates among people facing more meager circumstances, people with more conservative cultural practices. You sketch four great global-historical disruptions, with the emergence of humans as increasingly successful forager species happening first, the emergence of agrarian practices further securing human ecological dominance second, the emergence of an industrial era eclipsing agriculture’s centrality third, and now with the em economy/civilization to replace humanity’s industrial economy/civilization. Finally, you seek to describe (rather than to evaluate) the era that emerges after this fourth disruption, and you leave it to other futurist texts to postulate and track alternate scenarios, or subsequent developments (such as the AI-based civilization that you consider likely to follow the age of em). Any other basic parameters that we should lay out?
In a moment I’ll want to go back and question the status of some of those as assumptions (as opposed to consequences) of the analysis, but first I’ll mention other things that have more the status of assumptions. One is that I’m focused on the equilibrium of this new era, not the transition. The equilibrium should be easier to understand. Secondly, I’m focused on making simplifying assumptions as needed, to get concrete results, which is what theorists usually do. I describe that as looking for the keys under the lamppost (which again is what theorists usually do). One of those simplifying assumptions is, as you mentioned, supply and demand (i.e., a competitive low regulation). Another is that I assume the emulations are hard to modify. You mostly turn them on, turn them off, copy them, erase them, run them fast, run them slow. But you can’t do much else. There’s a limited range of tweaks, which only have a limited consequence.
Unlike some more fully engineered AI.
Then more generally, I assume that when we see practices today in our society and civilization, but we don’t fully understand where those practices come from, that these will continue — that they have some reason for being, and whatever analogous practices would seem to be their most straightforward continuation will probably continue. For example, if we swear today, then we’ll swear in the future. There are some theories about why we swear, but that’s less important than the idea that we’re swearing and we’ll probably keep swearing.
Now the idea that our values have changed over the last few centuries, but that this dreamtime won’t necessarily continue — that thought is a consequence of a certain perspective on where this value change came from. This is the certainly questionable (but plausible-enough) assumption that the most disruptive transition that humans ever had was the transition from foraging to farming, and that some forager (versus farming) sets of attitudes and styles roughly map onto our contemporary liberal-versus-conservative distinctions. The farming world used conformity and religious pressures to keep people in the conservative farmer styles, but as, over the last few centuries, we’ve gotten rich, those pressures have felt less compelling, and we have drifted back toward more forager attitudes. The threats that the farming world had to punish us if we deviated feel less compelling when we have financial resources with which to resist.
So if that’s true, if the main cause for our liberal trend over the past few centuries has been our increasing per-person wealth, and our comfort and autonomy based on that wealth, then if the emulation era brings back near-subsistence wages, it will plausibly restore more conservative attitudes and values. For the same reason that the farming world imposed its stronger conformity and religious pressures, the emulation world will find it useful to do that.
Maybe we should take a slight step back now and define emulations. You offer one direct definition: “An em results from taking a particular human brain, scanning it to record its particular cell features and connections, and then building a computer model that processes signals according to those same features and connections.”
I find it useful to offer an analogy between ems and porting software. Today, if you have an old computer running software, and then you get a new computer you want that same software to run on, one approach is to try to write software on the new computer that looks like the old software. This requires that you understand the software in some detail. But another approach is to port the software. You do this by writing an emulator on the new computer. To the software, this emulator makes the new computer look like the old computer. You don’t have to understand how the software was written or how it works. It will just run on the new computer.
So the idea with an emulation is to port the software that is in each of our human brains over to a more ordinary computer. In order to do that, we have to find a way to emulate, on an ordinary computer, the hardware in the human head. We need to have models for how each of the brain cells works as a computing device, models for all of the brain cell types — ready and available on our computer. Then we need to have scans of particular brains, to see where all the cells are, of what type, and connected to what. If we have good-enough scans and good-enough models, then we can port the software that’s in any one human brain and make a computer model of that particular brain, and get the same input/output behavior.
So if we hook it up to hands, eyes, ears, mouth, then we can talk to it and it will talk back. We can ask it to do a job. It should approach this task exactly like the original would in the same situation. So that means emulations are very human. They have human styles of thinking and arguing and getting mad and falling in love and striving for status. Emulations have all of these familiar styles, but they gain the advantages of being on a computer. You can run them fast or slow. You can make copies, etcetera.
And just to add a few basic logistical implications for the em future, a few broadest outlines of this conjectured scenario, you then say that:
This future happens mainly in a few dense cities on Earth, sometime in the next hundred years or so. This era may only last for a year or two, after which something even stranger may follow. But to its speedy inhabitants, this era seems to last for millennia. Which is why it all happens on Earth; at their speeds, travel to other planets is way too slow.
In terms of lived conditions within this economy, you tell us:
On the upside, most ems have office jobs, work and play in spectacular-quality virtual realties, and can live for as long as does the em civilization. On the downside, em wages are so low that most ems can barely afford to exist while working hard half or more of their waking hours. Wages don’t vary much; blue- and white-collar jobs pay the same.
I would assume that, hundreds of times by now, after you have sketched for audiences these external parameters, you have faced questions concerning the internal identity of ems, or why an emulation might not possess/resemble human consciousness, or why we should prioritize more radically engineered AI intelligence rather than copied or ported em intelligence. I’ll try to spare you from those questions. I’ll adopt your useful principle that everyday human experience already contains enough fragmentation, drift, drug and media distraction that we cannot separate some projected pure humanity from an artificial em identity. But, perhaps in relation to your software analogy, could you clarify the extent to which ems manifest an abstracted capacity akin (also, of course, superior) to their source humans’, and/or the extent to which they literally have that source human’s immediate historical/contextual disposition, as if waking from a daydream? Could your em answer this question I’m posing right now, about this book you have written? Or would your em simply have the capacity, given equivalent circumstances, to write an equivalent book? And to offer one practical point of reference: The Age of Em claims that the first ems most likely will emerge from humans at peak professional capacity (around age 50). But I assume that early ems have more chance of success if derived not from bossy CEO types, but from more precocious, supple-minded, technologically and experientially dexterous teenagers or 20-somethings (I realize this would raise other dangers and ethical questions).
If we scan you today and tomorrow we turn it on, it immediately has to be convinced that it is the emulation — because psychologically it was you a day before. Now it’s going to be in a new environment. It will get to try out changing its speed, and zooming around by jumping across the world on communication lines, and making copies of itself and talking to its copies. So it will very quickly start to experience new things that you don’t experience, and these will change it. But that change takes time, so it will be very recognizably like you until this experience is enough to change its personality and character and moods.
The very first ems are going to come from the most productive people, because if you’ve just spent many billions of dollars making this emulation technology, and you’re trying to now make money off of it by renting copies of ems, you’re going to be looking for the models that customers would immediately pay the most for — not necessarily CEO’s, because there’s actually not that large a demand for CEO’s, but occupations with a lot more potential customers, like software engineers, lawyers, and professions like that.
But very quickly (and perhaps I didn’t emphasize enough in the book how quickly), this will switch to people who are younger and more flexible. That is, as soon as the skills you need depart from just acting in the ordinary human world, to interacting in this new world, you will want more flexibility to learn and adapt to this new world. And people who are 50 years old and at the peak of their career probably don’t have as much flexibility.
Most ems after this point will remember a young childhood, perhaps of being human, and then most of the rest of their life being an emulation. As you indicate, this has great potential for ethical conflict, because the first scans will probably be destructive of the original human. So you have this prospect basically of the rich emulation economy going around to parents of very young but very promising children and saying “Can we please destructively scan your children so that they can have a chance at becoming successful emulations?”
But the very first ems will have an unusual combination of characteristics — high productivity, willingness to accept a destructive scan, known to the first companies, in a geographic or political regime where they’re allowed to be created, etcetera.
So numerous factors, both strategic and random, arise related to early em candidates. And here ems’ ongoing processes of competitive selection again seem good to discuss, since these probably provoke objections each time you mention them. When I first picture the selection of ems (who gets selected for copying and why), I think of breeding processes that have led to animal domestication, let’s say, where the best (from an immediate economic basis) specimen may be dumber, more passive, more drone-like, far from what we otherwise might consider an evolutionary pinnacle. So natural selection seems to get shut down, or directed towards some crass commercial calculation. And similarly, one could worry about the lack of the robust genetic diversity that sexual reproduction would have offered, further precluding possibilities for evolutionary innovation. Yet you provide a compelling alternative here, in the form of well-honed, well-funded, well-differentiated, exciting, and even risky childhoods for ems, from which only a few such ems will advance for copying. So you offer unprecedented generative diversity at early stages of em existence, and then unprecedented quality-controlled homogeneity at more mature stages (because only super-successful ems keep reproducing). We only have to rear a few successful ems, and can produce countless citizens spinning off from these. This all can happen quite fast, because the training can take place at a great processing speed (we’ll have to discuss speed). Also, if em training does take place at great processing speeds, this makes the environment seem, from the child em’s perspective, quite stable and quite nurturing. Of course doubts or regrets do creep in for me when you mention, for instance, moderate stress producing the most productive workers, or artistic industries deliberately cultivating ems with bipolar disorders (bipolar disorders just about to enter their manic phase — for a hyper-productivity). The fascistic-seeming implications of trillions of ems potentially emerging from a few dozen selected humans stands out, though I realize that billions of humans already have emerged from a select few ancestors, and that’s the way it goes. So again, I won’t raise pointed ethical questions, but will point to one smaller potential snag I see. In the future, whomever desires to copy him/herself will have a huge selective advantage. But what if this desire to copy oneself often stems from moody dissatisfaction, from narcissistic desperation, or some equivalent undesirable quality? What if the quality that makes someone a good em candidate in certain regards also makes him/her quite a problematic candidate in other regards, and then those negative aspects likewise get exponentially enhanced and disseminated? Or what other hang-ups or concerns do you have about this early and then ongoing selectivity process?
This seems a case where my trying to just be positive and not normative shows up in my style of analysis and discussion. I am comfortable with just turning the standard analytical crank here, to discover what kind of ems would be selected. I can just nod my head and say “Oh, that’s what they’ll select for.” Many speculative authors might experience an emotional barrier, because we have many opinions about the kinds of people we like and celebrate, and the kinds of people we don’t like. Some people might go so far as to say that if the wrong sort of groups win, then this is a hell, a terrible outcome — that what we valued in humanity has been lost. Whereas I’m going to say: “Well humanity is a pretty wide concept, with lots of different parts, and we’re just seeing some of those parts being emphasized here relative to others, but they’re all pretty human.”
One issue that you started out with is domestication. It has long been noted that the animals that best survive around humans are more willing to put up with human dominance, and less willing to defy us. Their brains actually get smaller, since they need less brain complexity if they are going to be subservient to humans (as with dogs, compared to wolves). And we have noticed the same thing happen to humans themselves. We have domesticated ourselves. We became more docile and our brains got smaller during the farming era. In some sense, that docility has even been accentuated in the industrial era, with its main distinguishing feature of the arrival of large organizations.
So even though we like to celebrate lone individuals all by themselves on the frontier, running their own farm, or their own trapping business, in charge of their own destinies, and though we somewhat denigrate and disapprove of organizational toadies who submit to a big organization’s directions and accept having a small place in it, accepting the constraints of organizational choices and playing organizational politics and not rocking the boat, this is in fact what our industrial era has been selecting for. The people who most succeed in our society go to the top schools. They often succeed exactly because they are very conformist, and they’ve always done what they’ve been told all through school. These people are very good at intuiting what teachers want and giving it straight back to them. This is who our society has been rewarding, in part because it’s what the organizations we have want.
Many people are uncomfortable with this. It is not the forager personality that we most reward, and I’m sorry to have to tell you that this trend plausibly continues in the em economy. Organizations just get larger, and people get more and more willing to find their place in an organization and to succeed in it. They are less inclined and less able to do things all by themselves. Honestly, they’re probably less creative. Already there’s only a limited number of professions in our world where creativity is substantially rewarded. We like to give lip service to creativity, but schools actually weed it out of people, and workplaces mostly don’t want it either.
Similarly, in most industries today, a small number of suppliers dominate. We’ve already been experiencing a convergence, reducing some kinds of variety, as certain common-standard practices compete across the world and win out. In the em world all of these same trends continue, and they now happen to labor markets too. Workers will be much less varied. So instead of having ten billion diverse human workers supplied, you end up with a few hundred human types that supply most of the global market. Yes, that is a great reduction of diversity, but plausibly it’s what this economy wants.
You’ve mentioned how your book cranks out consequences, and how an important part of this project depends upon thinking through such implications, without getting too emotionally disturbed along the way. And your use of the terms “person” and “people” today has intrigued me, particularly when you refer to ems. I don’t want to impose too stark a binary, but I wonder if we could pause to consider humanity’s fate (with “humanity” meaning current, non-emulated human beings). We’ve said, for instance, that a youthful em, processing information at a speed one thousand times human levels, and also aging one thousand times faster, experiences a remarkably calm external world. And if relative speed creates a sense of external stability, then humans face the opposite extreme. Thousands of years of subjective em experience will play out in two years of human experience. Or again, your book hints that, since ems agree to quite clearly defined life conditions before contractually accepting their fate, humans, by comparison, seem to face the converse — a bewildering passivity before existentially determinant developments taking place in Lilliputian cities much more impenetrable than Franz Kafka’s castle. And then of course, after these two unimaginable human years, during which hyper-intelligent ems have millennia to prepare for whatever destabilizing technologies might come next, humans will face some new, even more astronomically disruptive change, happening at an exponentially faster pace. Along the way, you predict something like the eclipse of human labor, the inevitable death of those unable to sustain themselves through investment capital or government wealth transfers. You acknowledge that a near infinitude of ems themselves will get displaced, which further suggests the catastrophic displacement of even less adept humans. You definitely posit vast ecological destruction, alongside a few nature or human preserves that sound like zoos or reservations. So yes, we might want to conceive of humanity’s future social position in relation to present-day retirees, but this sketched scenario seems more like retiring into Weimar Germany than like moving in with the Golden Girls. And then I have one desperate suggestion. If ems agree contractually to coming into being each time, could we build in certain pro-human contractual obligations? Could such contracts stipulate that all ems must plan to leave Earth after 1.5 human years, that ems must create a new sustainable living outcome some place in the cosmos, or…
Yeah, that’s just not going to work.
OK, because that was the hope. I would frame it as: given ems’ religious, authoritarian, legalistic, bureaucratic, conformist tendencies, could those tendencies allow us to shape this future more than we might suspect? I guess not.
It’s useful to make a comparison to the last major transition. You could imagine, just before the industrial revolution showed up, people being told that it was coming. They’re of course living mostly in farming villages, and you’re telling them that the future will be industry concentrated in cities. They frame this as the farmers versus the industrialists, and say “How can I save my farming children? My farming children will be pushed to the margins. My farming children will no longer be the center of action. They won’t run the world. They won’t have most of the wealth. They won’t have high status. My poor farming children will suffer compared to those industrialists.” If you think of your children as intrinsically farmers, that’s kind of where you’re stuck.
The farmers were pushed off to the margins by industry, and today there are farmers out there on the margins, but they don’t have the best land. They don’t have great resources or status. They don’t run the world.
So I’m telling you that there’s another era coming, where yet a new way of life will displace these present ones, and if you think of yourself and your children as intrinsically non-em, then you say “Dang, the humans are no longer running things, and we’re on the margins and at risk of instability.” But if I’m willing to think of my descendants as potentially including ems, well then I want my children to try and become ems, the same way that a farmer from two hundred years ago could say “Well, I want my descendants to try to become successful industrialists. I want to send my kids off to live in the city, to try to learn the city ways, to learn the new industry ways, and I want them to be a part of this future.” That’s a choice you would have, and that’s a choice we have now.
Over the last century or so, things haven’t actually been accelerating that much. Even though in the long run history has been accelerating, we haven’t seen the acceleration up close and personal, because the accelerations mostly come in a small number of really big jumps in speed. But now there is another jump coming, and maybe even another jump after that, and so the faster the change gets, the more you should expect that, if you don’t speed up with it, you will then experience a very chaotically changing world, and we can’t give much assurance of what happens to you throughout all that change. That’s just the nature of lots of change.
Amid this impending chaos (for some), I also would love to return to a few of The Age of Em’s concrete details, those helping to anchor its broader speculations. I just want to make sure we include some of these novelistic delights with which the book overflows. For instance, ems might meditate. They might like singing together. They’ll tend, as you’ve said, not only to swear, but to swear a lot. They’ll walk slowly, with wind posing a serious problem (their robotic bodies will average two millimeters in height). They’ll see the sun dimmer. They’ll possess nearly telepathic powers with recent copies. They might receive 24-hour mood-induced life coaching from surveilling members of their clan. I love those extractable details, and then more complicated scenarios you give us also. Complex social norms will arise around whether one em deems it worthwhile to talk to a fellow em, or simply to send a conversational bot in one’s place. So I, as a status-obsessed em, might send a stealthy bot in my place, to see if you have sent a stealthy bot in your place. Or as a socially anxious em, I might hit my own personal “undo” button after an embarrassing faux pas — reversing myself to a previous state. But I also might get unjustly seduced, coerced, tortured by somebody who learns to misplay my “undo” buttons. Or the section entitled “Merging Real and Virtual” alone could prompt thousands of questions about how ems, as they cross doorway thresholds, can rent new robotic bodies and things like that. But do you yourself have favorite novelistic touches in the book? And how does this em world exist in your mind today? Do you sense a coherent, visualizable, mappable, tangible totality? Do you experience a series of discrete postulations, localized conjectures? To what extent has this em world crystallized in your mind as a closed, self-sufficient system? To what extent does this world continue to evolve as you discuss it and maybe flesh out new logical conclusions, subverting previous prejudices or assumptions?
When I’ve shown drafts of this book to people over the last few years, many have mentioned that they prefer to get their futurism in the form of fiction, so wouldn’t I please write some fiction instead of this nonfiction stuff.
You get that response primarily from fiction readers? Or from academics too?
Often both. Certainly many people who enjoy science fiction say “Why don’t you write fiction based on this?” I of course know that it takes years to become a good fiction writer. I greatly respect the skill, and know I don’t have it. I have a long computer file full of little pieces, of fiction story-elements, including plot pieces and character pieces and things like that, but I haven’t tried to put it all together. Those pieces are part of what makes this world vivid to me.
Still analysts like myself are concerned that many people seem overly persuaded by the detached details of fiction. Writers will present a basic scenario in a story, where they make some assumptions about wages or robots or whatever. Then they fill the story with detail, and readers get so emotionally attached to it and enjoy the detail, and as a result find the larger thesis more plausible — just because they’ve seen it attached to this detail, even if the detail doesn’t offer any evidence for the thesis.
Or contradicts major premises within that thesis.
So I’m worried about the critique that I have unfairly persuaded people that this scenario is plausible — that I’ve given detail that doesn’t evidentially support the basic assumptions, and that just makes this scenario easier to imagine. To defend myself against that claim, I like to point to each detail and say “Look, I just said weakly that this is a thing that might happen, and I almost never take several pieces of detail and combine them together to get a conclusion.” If I combined random details A, B, and C together, and said “Ah, that implies D,” then I would really be stretching it. But I do think that, with most of these details, you’ll find that the rest of the book hardly depends on them — that you can change these details as you wish, deny some, assume the opposite, and still most of the book goes through just fine. I’m not essentially relying on their conjunction.
Over the years I have studied and talked about what’s called construal-level theory. This is a very basic but important area of psychology, which shows that all of us have two very different mental modes (and a continuum in between). We tend to do math in a concrete mode for example, but we tend to be more creative in an abstract mode. We also tend to be more gullible in our abstract mode. Often our abstract mode is where we are willing to believe things that sound nice, or because they’re socially acceptable, and our concrete mode is where we’re more careful about what’s actually true. When we think about the future, we’re often in a very abstract mode, because there aren’t many concrete details to anchor our thinking, and this means that we more often wander into wishful thinking. For example, in an abstract mode we tend to focus more on basic moral principles. We’re less tolerant in deviations from them, whereas in a concrete mode we tend to focus more on practical constraints, and less on basic principles, for choosing actions. In a concrete mode, we tend to believe that our theories have more random exceptions, and we’re less confident that our theories describe what actually happens. I think those are better habits in thinking about the future.
So let’s continue with concrete applications of key concepts from the book. I have a few questions related to speed. Faster em mental speeds correspond to increasingly smaller size, to smaller robotic bodies, and to slower subjective temporalities — so that the external world seems to operate more slowly for a fast subject. In each such case, the legacy of Einsteinian relativity theory and your own physics background intrigues me here. And just to clarify for readers how variable subjective speed plays out, you offer a diverse range of scenarios in which, say, managers run at high speeds in order to stay in touch with numerous underlings, in which work teams accelerate so that they can complete a project on deadline, or vary their speeds to stay in sync with each other and complete a project simultaneously. You offer equivalent possibilities for a romantic couple with divergent sexual appetites to adjust their speeds and coordinate cycles of desire. You note how easily affairs might take place, with secret rendezvous happening at an extremely fast pace (which subjectively, again, feels like quality time).
Certainly the idea of ems running at different speeds has shown up in science fiction for a long time, but usually science-fiction stories don’t follow through with these various social implications, because the author isn’t thinking very much about social science and related fields. So again I’m using very simple concepts and applying them relatively straightforwardly, but adding a lot more factors than people had done before. In some sense, people had been put off by these details. This seems so strange — you couldn’t possibly say interesting things about it. Who could possibly know about such a strange thing? I just sort of say “Well, let’s try.” Speed is actually pretty solid for predictions. We know that if you have a parallel computer program and you add more processers, you can run the program faster. So if you had a human brain on a computer, then you could run it faster. The question is how does that change things.
Just to include a few additional concrete consequences, I love your description of em city infrastructure — with neighborhoods, clustered around ems operating at a certain speed, and built at a corresponding scale or size, creating urban nodal points at which two different scales somehow must join in a single bridge or intersection. And do you envision an asymptotic approach towards an infinitely small city size as speeds increase? Do these accelerating cities housing trillions of ems approach infinitely small masses?
There’s a range over which when you double the speed you double the cost. I estimate that range to be roughly up to a million times faster than humans, and down to a million times slower. Beyond that range, however, the cost rises faster than a linear model would predict. That makes it prohibitively hard to have emulations much faster or much slower. So there is some sort of maximum speed. If you (as an em) and your competitors are in a race to produce a product first, you may go up to near the fastest speed but you still face limits.
But to return to your question about city infrastructure, the most interesting thing to talk about with speed is fragmentation. We are used to a world in which most people operate at nearly the same capacity. We have a lot of concepts of equality and affiliation with other humans on the planet, based on this idea that even though we vary somewhat in our abilities, we have roughly the same range. Yet once emulations can run on different speeds, that just changes. Some emulations are vastly more capable than others, and em society can’t keep these cultures synchronized. So today we often have a synchronized world culture if we want that. We all sort of know about certain events and track them and feel like we’re part of humanity because we do so. But emulations who run at very different speeds will have really fragmented cultures. For the ones that go really fast, any shared culture would seem intolerably slow. Or you could temporarily speed up and visit a fast culture and find they have a new kind of music you like. But by the time you can come back to the slow culture and tell people about it, and by the time people can try it out, the fast culture will have changed to something else, because now you’re back in slow speed again.
To follow up on copying as well, could we pivot towards the complex conjecture of a society in which, at any given moment, almost all individuals operate at peak age, at their peak time of day, feeling as if they have just come off a restful vacation? And you also suggest some broader trends toward localized cohesion stemming from the copying not only of solitary individuals, but of particularly successful social units — such as work teams or romantic couples. And once more you track not only broader economic developments, but also the concrete details, the lived scruples of, say, having to determine which of its friends an em will keep once it has made copies, versus which friends instead will get assigned to these forthcoming copies. Or if one couple gets divorced, that might make all copy couples quite tense, pushing them towards self-fulfilling prophecies. Or to combine these two concerns, if I emerge as a successful em, more successful than my partner, then more copies of me might get made, and, paradoxically, some of those copies will feel left behind. They won’t have a spouse. You offer intriguing speculations on impulsive em behavior, socially and perhaps sexually — sometimes as desperate attempt to cultivate a slightly non-normative, non-mechanistic sense of self. You suggest that non-reproductive sexuality might encourage enhanced experimentation. And we probably should include clans in this quick survey of em copying and its implications. Clans offer perhaps the largest, most powerful, most clearly delineated social institutions in the em world. On a more individuated level, The Age of Em also presents scenes of great trans-subjective intimacy between clan members. You’ll describe me feeling towards clan members how I feel about my own reflection in the mirror. Here I also recalled Freud’s conception of the narcissism of small differences, a social dynamic in which the most minute differentiation between one specialized group member and another (the type of difference that nobody else in the world would notice — say the methodological differences between two members of an academic department) actually creates the most painful social tensions, the most destructive emotional conflicts. I absolutely love your scenario of an em watching movies about the most bohemian version of itself. But I also envision that (just as in present-day family life) the strongest affinities might prompt the most incendiary tensions.
So this is one of the factors that mitigates against the “ticky-tack houses all in a row” sort of problem. From outside, you see a world of similarity and conformity and you say “How sad for them. They all feel the same.” But of course they’re going to focus on smaller differences and highlight those and feel just as different. Everybody can see difference if they want to. And the big interesting question that’s hard to project here is how ems will feel about copies of themselves. This is something I’ve really struggled with, because it’s just not much like any current sort of relationship, and the question is which current relationships will it seem most like.
There’s an interesting comparison between how we are spread across time and how we can spread across space. If you imagine a small civilization spread across a vast amount of time, in the sense that it eventually will have a total of a trillion people, but maybe only about one hundred thousand at a time, that seems a very noble, respectable civilization that is stable, sustainable, etcetera. Now imagine a similar-sized civilization spread across space instead, which only lasts for ten generations. This seems somehow selfish, self-controlled, crude, and destructive. We have less respect in a sense for civilizations spread across space than those spread across time. Similarly, when we see ourselves as individuals spread across time, we are more willing to endorse these selves. We can see these selves, spread across time, as the same person. There’s more of a risk when we are spread across space. We may feel instead competitive with our spatial counterparts.
I expect ems will try hard to avoid that. The clan will want its members to feel identified with each other and to work together. So clans will probably do various things to avoid this sense of competition among clan members. They might even prevent ems from directly observing each other. You might just have a voice in your ear that offers advice from your clan. You won’t actually see other clan members around you. To the extent that you know about other teams, you might want to make sure your team seems different from theirs. Teams might want to create a local team culture so that they can feel more bonded with each other, relative to these other teams. Of course the more that you notice these other teams, the more you’ll feel you need to go out of your way to make sure your team seems different. But if you could just not pay much attention to them, you may not mind as much. Just like you may not mind living in a house that looks exactly like 20 other houses as long as those other houses aren’t built near yours.
You mentioned the difficulty in conceiving of copies of ourselves, and of identity spread across space instead of time. I wonder if discussing spurs may help a bit. Your most basic example of spur-related activities involves sending a spur copy to the DMV, to complete a task it actually benefits the em not to experience or remember. You then hypothesize a plumber em who creates one hundred spurs of itself to complete the day’s activities — with most such spurs then retired at day’s end, with the plumber who had the most constructive day emerging as the em who will continue forward. Or an individual might complete a complex task by copying itself, assigning the spurs various tasks, and then maybe the source em runs at a faster speed to coordinate their composite efforts. Or the concept of spurs safes allows for any number of compelling scenarios (in which, say, a therapist spur could listen, respond, and then expire with no potential to divulge confidential information; or a boss could tell my spur something I’m better off not knowing, and my spur might exit the sound-proof spur safe and say “Yeah trust us: it’s for your own good that you don’t know this”). Or an almost-retired spur might spontaneously, secretly, decide to do something illegal on its source em’s behalf. Spurs of course also raise much broader questions that we haven’t touched upon regarding the stakes of death and of expendable personhood in this em society. So as we move again towards em-world scenarios that certain readers may wish to resist, could we also address the retirement of ems, specifically of spurs? Do audiences often liken these outcomes to self-serving euthanasia campaigns, to enforced genocide?
Spurs do tasks that you don’t want to learn from, that you don’t want to remember. Those two activities have always been bundled together. So even when humans do something somewhat boring or somewhat unusual, they might say “Well at least I’ll learn something from this. At least I’ll have expanded my range of what I know about the world.” But for ems, that would be more expensive. They will choose which few things to learn from, and for most other experiences will say “Well I would have to pay all these other costs, so let’s just do without.” Think of all the things we learn from today, and ask: well, if learning from things were a lot more expensive, where would I cut back? For which things would you say, “Let’s just do that and not learn from it”? At the end, the spur might tell you what happened, just so you know. You may even have a little record, but you won’t feel it yourself and really have seen what it’s like.
And yes, when you present a spur as something that lasts for a short time and then is erased, even though I argue that many ems would be fine with that, many people get indignant. But when you frame it as “Well the spur won’t end — it will just retire to a slower speed,” that really deflates a lot of the objection. Many people will say “Oh, I get it: you’ll just have a slower existence, but you’ll go on.” I think in some sense that goes too far in the other direction. Ems will in fact resent being reduced to slower speeds. Slower speeds will be lower status, and ems will consider a slower speed as a failure, as a reduction of sorts of who they are.
A large class of very slow ems who are retirees will be like ghosts in the em world. The em world really does have ghosts. In our literature, ghosts are these creatures that are all around us and mostly we don’t see them, but if we were to interact with some of them, we’d find that they don’t really know much of what’s going on, that they can’t really have much influence on things, that they are kind of obsessed with the past, and they’re not very interesting to talk to really (which is why we aren’t so interested in talking to ghosts, even if we could). The em world will have real ghosts like this — lots of slow and cheap creatures who aren’t keeping up with things. They don’t have much influence, and we can and do mostly ignore them. This will be the fear: that if you have to go slower, you’ll become a lower creature who isn’t in touch with things, who other people discount as being less important and ignorable.
Of course related questions about potentially controversial claims and recommendations do arise throughout the book. Your assertions that non-democratic political modes prove more efficient in certain contexts, that a return to authoritarian farmer-era values seems likely, that the em economy will most closely resemble a contemporary poor Eastern nation’s economy…these claims might come across as neutral predictions based on tracking certain objective trends. Maybe people can’t much criticize those passages when you mostly just tally up the numbers. But then your critique of a one-em-one-vote policy (a policy which would just incentivize the production of cheap, low-functioning ems) sounds logical, if a bit unsettling. Your support for pay-for-performance standards, for a taxation of leisure time…at this point you start setting off my self-protective professorial sensors. Your strict-constitutionalist legal scenarios, in which we simply bring a law’s writer back from retirement to get the “correct” interpretation, or in which interpretive flexibility becomes less necessary (because we can just copy one judge over and over and have that same judge deliver equivalent results in every case) begins to challenge my contemporary political preferences. And then by the time we reach considerations of sickness as a laborer’s excuse for laziness, or of blackmail and random executions as economically efficient models of justice…at that point I assume you have alienated certain general audiences. And your comfortability with that fact impresses me. I can live with these alienating instances, and assimilate them as part of the largely descriptive ramble on which you’ve taken me, rather than as policy-paper prescriptions. But given your past exposure to unfairly sensationalized political controversies (regarding ideas futures), given the ability of hostile reviewers easily to extract, isolate, magnify some of these controversial bits from The Age of Em, given our contemporary social-media call-out culture, I wondered how you imagined ideas big and small from this book might circulate. Do you care about such concerns?
I’m probably just not as calculating as you anticipate. As an economist and social scientist, I’ll say “Well, there are social technologies in addition to physical technologies, and we should think about how social technologies will improve.” I know very well as an economist that most people have a difficulty seeing institutional evaluations like this as neutral. To them, this analysis gets tied up with key values, and they support some institutions (and oppose others) because they identify these with their key values.
But again, I’m not trying to recommend this world. It’s not my job to make you like it or dislike it. My job is just to tell you what is likely to happen. I understand that this isn’t what a lot of people want from futurism. They really like futurism as a place to sort of talk about their values, and they’d rather just use the future as an excuse to reaffirm their values and to complain about people who disagree with their values, and they don’t care much about the actual future. That’s part of the tradeoff between me actually analyzing our likely future and me telling readers things they don’t want to hear.
I also wanted to add that I appreciate your critique of how contemporary knowledge-production too often trains one’s powers of reasoning in pursuit of adversarial combat, rather than proactive, practical, pragmatic modeling and application. I agree, for instance, that we care too much about projected em personalities, rather than the concrete social implications of em demographics. You’ve allowed me to suspend any number of “why” questions throughout the book. But, to close, I do still face a species-wide question more like “why bother?” — why bother to design em domination? You’ve stated in the past that, if humanity just continued with its current economic growth rates, it could achieve one thousand times our current annual production within 150 years. That sounds like more than enough to me. And what if humanity could innovate by extending our current dreamtime beyond expectations, and enacting more sustainable growth that way? You ask us to think through carefully our conception of the human, to avoid disowning our descendants if we don’t believe our ancestors should have disowned us. You give a utilitarian rationale for this future and its trillions of sentient beings, though I just see one era’s comprehensive set of life forms annihilating another’s, without any real improvements to come. So can we conclude our discussion of this well-traveled academic reverie, this sci-fi novelesque without the novel, with you making a persuasive case (if one exists) for welcoming this em future — for embracing it basically from a point of desire?
Throughout most of history, people who traveled to foreign lands tended to become richer and higher status, but still they were somewhat suspect, especially if you traveled too much and seemed to like those other places too much. We thought instinctively that our culture and place were better, and those other people were suspect, and we had to worry about the nefarious plans they had to undermine us. If you visited a foreign land and came back to describe it, you would realize that people would be interested in this foreign world, but they would probably find ways to see it as worse. They would probably find ways to read what you wrote as validating their sense of superiority, and the more that you seemed sympathetic to these other people the more they would question your own judgment or loyalty. You might hope at least to provide a wider sense of the range of possibilities, to make the case that maybe those other people aren’t as bad as they seem. Maybe there at least could be an alliance or trading with them. But you couldn’t really expect to make people in your world love this other world better than yours. That would probably be too much to ask.
The same sort of thing happens in traveling across time, and telling people about worlds past and future. Most everyone is going to think that their time is best, that the culture of their time is superior, and you probably can’t do too much to change that. You’re not going to make people love the future so much that they give up their own current culture and trade it in for the future culture. They grew up in their culture, and they learned that their ways were best. So here I am, a traveler to the future, bringing you back a tale of what the future world is like. I don’t think it’s reasonable for me to hope to make you think this future world is better than yours, or to make you love it.
But what I can hope to do is to make you appreciate some of the complexities, to make you think it’s not as bad as you thought maybe, to make you at least see there’s some hope for a modest degree of relation with that future, and trade with it — that maybe you can help it, or it can help you. And that’s the most I should plausibly hope for, in a travelogue of this sort. I travelled to a strange place, and in a sense I’m describing some of the most alien creatures you’ll ever hear about. No one else on Earth or even in the past might be quite as alien to you as these future creatures. My book lets you encounter aliens and ask yourself what you think of them. Most likely, most people will be repulsed by aliens, or at least dislike them, or feel superior to them. Still I hope on the margin that I can help you understand these descendants of yours, and maybe appreciate some of their concerns from their points of view, and make a modest degree of progress toward you understanding them and them understanding you.