Hanson's Predictions: A 5-Page Age of Em Robustness Test
By Bryan Caplan
Robin argues that even if my main criticisms of The Age of Em are correct, most of his predictions remain true. I’m skeptical. On Twitter, Robin challenged me to randomly select five pages from his book to inform our dispute. Limiting myself to his chapters on Economics, Organization, and Sociology, a random number generator gave me the following pages: 189, 216, 241, 304, and 336. Robin’s in blockquotes, I’m not:
We have many reasons to expect an em economy to grow much faster than does our economy today. As mentioned in Chapter 13 , Competition section, the em economy should be more competitive in the sense of more aggressively and more easily replacing low-efficiency items and arrangements with higher-efficiency versions.
Seems reasonable. Indeed, if I’m correct to think ems will be “robot-like,” this just makes Robin even more right. As I’ve stated, however, I interpret “much faster” to be a few extra percentage-points of growth per year, not a monthly doubling time.
Reduced product variety and spatial segmentation of markets help innovations to spread more quickly across the economy.
With robot-like ems and fabulously wealthy humans, I’d expect the opposite: more variety, more spatial segmentation as the typical human enjoys global travel and multiple residences.
Stronger urban concentration should also help promote innovation (Carlino and Kerr 2014).
In an em world, it’s not clear what even counts as urban concentration in the relevant sense.
The fact that more productive em work teams can be copied as a whole should make it much easier for more productive em firms and establishments to rapidly displace less productive firms and establishments.
Most of the research that aids innovation is “applied” as opposed to “basic” research. Thus we expect most of this better and faster em innovation to consist of many small innovations that arise in the context of application and practice.
If, as Robin argues, ems are based on a small number of humans, why wouldn’t we expect this pattern to reverse due to the homogeneity of the innovators?
An em economy might further reduce the costs of crime and disease, and increase the gains from innovation. After all, well-designed computers can be secure from theft, assault, and disease.
On standard human definitions of “theft, assault, and disease,” this makes sense. But why should we think ems are less vulnerable to the cyber-analogs of these problems than humans are to the conventional versions? Why couldn’t computer viruses bedevil ems as germs bedevil us?
More important, traffic congestion costs could be much lower in an em city because most transport of ems could be done via communication lines, and most virtual meetings within a city do not require the movement of em minds at all. As congestion costs now limit city sizes, this increase in virtual meetings could plausibly tip the balance toward much larger em cities.
Telecommuting hasn’t done much on this front so far. So why think ems will lead to “much larger” em cities? Furthermore, doesn’t being a virtual being vitiate most of the social reasons to live near others?
Thus instead of today’s equal distribution of people across all feasible city sizes, most ems might instead be found in a few very large cities. Most ems might live in a handful of huge dense cities, or perhaps even just one gigantic city. If this happened, nations and cities would merge; there would be only a few huge nations that mattered.
With robot-like ems, this might lead to free-trade zones, but not national mergers. This is especially true if, as Robin claims, the Age of Em only lasts a couples of years.
Larger em cities could allow great increases in social complexity.
Very vague, but ok.
Similarly, mass-market ems can be comforted to know that most disturbing life events were part of the clan’s plan.
If I’m right, ems will primarily be “distubed” by doing a bad job for their human creators, not by “life events.”
When an em clan supplies workers to a very large mass labor market, there must be at least as many members of that clan as there are workers who serve that labor market, times the market share of that clan. Because this number will sometimes be larger than the total number of members in a small clan, clans who supply very large labor markets must also be large. This may create a correlation between clan size and mass-market work; larger clans may tend more to supply mass labor markets, while smaller clans more often serve niche labor markets.
If ems are robot-like, I don’t see why they would form into “clans” in the first place. They’ll be part of their human owners’ teams, not their own.
While ems that serve mass markets could more easily be copied as a part of large teams that retain most of their familiar friends and context, such copied teams could not as easily bring along copies of large shared community spaces such as city neighborhoods or parks. For very large spaces this is not a problem, as the different copies of the team can just share the same original space, and only rarely meet in the large space. For smaller spaces, however, ems must choose between moving to a new “parallel world” space, which can of course still look a lot like the old space, or sharing the old space and dealing with the prospect of meeting more often and mixing with similar members of other teams.
I think I’d have to read a lot of background information just to know what this means.
In sum, the lives of both mass-market and niche-market ems differ substantially from our lives, but in different ways.
Very vague, but ok.
Ems from large clans should be less aware of, and feel more secure in their identifying with a personal combination of personality, style, and early life history. They know there is in effect a whole “planet” out there of perhaps billions of ems just like them, “super-twins” to whom they can talk at any time, or ask for help. Thus when around ems from other clans, ems might feel more like expatriates, that is, emissaries from a foreign land.
If ems are robot-like, they’ll identify primarily with the owners’ interests, and simply won’t have a lot of “personality” in a human sense.
As discussed in Chapter 19, Managing Clans section, ems who made a plan and then split into many copies to execute that plan are likely to feel that they and their copies are the “same” person, while copies who openly disagree with each other are likely to see themselves more as different people. In an attempt to unify their members, clans may discourage open disagreement, focus more on the fact that they have a shared mental style and shared memories of early life, and focus less on recent opinions and memories. After all, those recent opinions and memories often differ.
Again, in my scenario I see little role for em disagreement or self-organization.
When doing physical jobs, ems could use interchangeable bodies. Virtual ems could also easily change their virtual bodies at any time. Thus em identities need tie less to details of a particular body. However, to create a vivid memorable identity that is easily remembered by other clans, each clan may still create and stick close to a consistent visual style.
People today stay loyal to product brands over surprisingly long periods (Bronnenberg et al. 2012). This weakly suggests that ems will also stay loyal to brands for long times. Brand owners are eager to offer deep discounts to young ems with promising futures. While the em world has few children, each child has on average a great many descendants. On the other hand, collective purchasing by clans makes for better-informed consumers, and such consumers tend to have a weaker attachment to brands (Bronnenberg et al. 2014).
Very vague, but ok.
Today, our physical illnesses give us excuses to escape social pressures and expectations. If we don’t want to do something, we can pretend to be sick. Ems, who never get physically sick, lack this excuse. Perhaps ems will expand their expectations of temporary mental illness to compensate.
Given Robin’s discussion of “software rot,” I’m puzzled by these claims. In any case, robot-like ems won’t be inclined to make excuses for anything.
Because ordinary humans originally owned everything from which the em economy arose, as a group they could retain substantial wealth in the new era. Humans could own real estate, stocks, bonds, patents, etc. Thus a reasonable hope is that ordinary humans become the retirees of this new world. We don’t today kill all the retirees in our world, and then take all their stuff, in part because such actions would threaten the stability of the legal, financial, and political world on which we all rely, and in part because we have many direct social ties to retirees. Yes we humans all expect to retire today, while ems don’t expect to become human, but em retirees are vulnerable in similar ways to humans. So ems may be reluctant to expropriate or exterminate ordinary humans if ems rely on the same or closely interconnected legal, financial, and political systems as humans, and if ems retain many direct social ties to ordinary humans.
As my earlier critique explained, Robin’s analysis strongly suggests this amicable human-em relationship will end in an objective year or two. With robot-like ems, however, it’s basically right.
Few ordinary humans can earn wages in competition with em workers, at least when serving em customers. The main options for humans to earn wages are in direct service to other humans. Thus individual ordinary humans without non-wage assets, thieving abilities, private charity, or government transfers are likely to starve, as have people throughout history who lacked useful assets, abilities, allies, or benefactors.
In our world, financial redistribution based on individual income has the potential problem of discouraging efforts to earn income, and thereby reduce the total size of the “pie” available to redistribute. In an em economy, however, where most all humans are retired, this problem goes away; there are fewer incentive problems resulting from financial redistribution between retired humans.
In Robin’s scenario, I don’t see why humans wouldn’t tax ems, leading to standard disincentive effects (or perhaps non-standard Malthusian effects).
Ordinary humans are mostly outsiders to the em economy. While they can talk with ems by email or phone, and meet with ems in virtual reality, all these interactions have to take place at ordinary human speed, which is far slower than typical em speeds. Ordinary humans can watch recordings of selected fast em events, but not participate in them.
Do CEOs “participate” in their firms? Of course; though they don’t have time to follow the details, they’re ultimately in control. I’ve argued the same holds for humans in the Age of Em.
Although the total wealth of humans remains substantial, and grows rapidly, it eventually becomes only a small fraction of the total wealth, because of human incompetence, impatience, inattention, and inefficiency. Being less able than ems, humans choose worse investments. Being more impatient, they spend a larger fraction of their investment income on consumption.
In my scenario, humans hold 100% of wealth regardless of their incompetence, impatience, inattention, and inefficiency. Humans own ems, ems own nothing.
Is Robin really wrong 90% or more of the time? I don’t know quite how to score this five-page dissection, but I’m open to readers’ suggestions. After performing this exercise, I’m more inclined to say Robin’s only 80% wrong. By any measure, however, his forecast accuracy seems low. My main complaint is that his premises about em motivation are implausible and crucial. But in several cases his reasoning from his own premises seems shaky as well.