The Age of Em is Robin Hanson’s first book. A normal economist might have started with a more popular topic rather than something niche, with the title even containing the short term “em” he coined for emulations and that most people would find meaningless. He actually sent me an advanced version not intended for publication but instead for critique prior to writing a version for a popular audience, but given all the distractions available on the internet I never got around to reading that. He doesn’t seem to have had any shortage of such critique as the book is full of caveats and responses to possible objections. The typical big-idea book tries to get away with impressive boldness and devotes relatively scant attention to possible counter-arguments. This is not a typical book. It’s sort of like a non-fictional version of scifi, as it’s focused on a technology that doesn’t currently exist and may never come into existence, but there are no characters or made up burdensome detail or even a real plot. Hanson has complained that there is so much more history than futurism, so this book represents the thing he thinks there isn’t enough of. I’d say the reason isn’t because the idea hadn’t previously occurred to anyone but him, but rather that a market for reader attention will be inclined that way even if an ideal sort of futurism might be more “useful”. I should note that Hanson is one of the thinkers I most respect, and I think he’s more likely to be correct than Yudkowsky in the AI foom debate over whether emulations will come first, but like the little girl doing a book report on a text concerning penguins, I found there was more than I was eager to know.

One of the things that distinguishes it from other non-fiction books is the citations. Usually a text will have numbered superscripts that refer to a line that particular chapters section in the references at the end. Hanson does not do that, but instead always inserts parentheses containing the surname of the author or authors along with a year and finally a possible number to distinguish between multiple works from the same author within a year, and all the references are listed together regardless of the chapter citing them and in alphabetical order by author. This is more reader-friendly for someone who actually wants to know who is being cited, and presumably the reason this practice is uncommon, aside from taking up more space, is that most readers don’t actually care. That an author says a claim is sourced is enough for them. Many of the citations in this book are to Hanson himself, including not just academic papers but also his blog posts. That was fine by me, but people who take blog posts much less seriously than peer reviewed papers in prestigious journals might think otherwise. A common occurrence in this book to state a stylized fact deriving from one of those citations to say that we should a society of emulations to differ from ours in some way, but then on the other hand a different stylized fact could suggest the opposite. Originally the term “Singularity” was used to refer to the period at which change advances so fast that people prior to it cannot make any useful predictions about anything afterward. Hanson thinks our current industrial era, which has been going on for just a few centuries, will be replaced by an era dominated by emulations which will itself only last a year or two in human time, before switching to yet another era of even faster growth. He doesn’t try to describe that more distant era, which could thus be considered after his Singularity, but thinks that understanding the era directly after ours will help us to understand what comes after that.

One of the systematic differences between economists and non-economists (as discussed in “The Myth of the Rational Voter” by Hanson’s GMU colleague Bryan Caplan) is that the former are far less likely to see harm from trade, particularly international trade. As Robert Wright might put it, they see an ever-expanding circle of positive-sum interactions. One explanation for why many people don’t assume the same is the evolutionary psychology theory that there was less mutual benefit and more conflict in our evolutionary history compared to our industrialized present. Hanson himself has written about how our industrial era retains some changes brought about by the farmer era, but removing the specter of Malthusian limits is also eroding some farmer era cultural adaptations so that attitudes from the hunter-gatherer era are making a comeback. Hanson expects the emulation era to quickly rapidly expand the number of ems until they earn subsistence wages, thus bringing back some farmer attitudes even while doubling down on some industrial innovations. However, I think he gives short-shrift to war. The “Organization” part of the book has a “Conflict” chapter containing a “War” section, but it is only a few pages. Historians have typically devoted a lot of attention to war, particularly in the past, and while one could accuse them of being biased on disproportionately focusing on that, I think there’s a reason people subjectively found it so interesting. As Trotsky is supposed to have said, war is interested in you. A Malthusian situation is far more zero-sum, and thus conflict would actually be more rational than in our industrial era where an Amish strategy of pacifism (particularly when a shrinking portion of the rest of the population are farmers) is viable. He makes various citations about factors that make war more or less likely, but most of his data are from our current industrial era of declining war. Recent findings in ancient DNA are showing that large scale “population replacement”, in which (at minimum) basically all males of a population are wiped out, was a repeatedly occurring phenomena. Hanson himself has recently blogged about his fears that the current lull in war may be something of a temporary aberration between periods of more conflict. Carrying on the analogy between eras, the em era could similarly alternate between periods of peace and war, accelerated by the subjective experience of time described in the book so that they are vanishingly brief through the eyes of the remaining biological humans. Both such humans and the slower speed cheap/physically large ems he discusses would be at a disadvantage in reacting to such wars. Hanson makes a lot of predictions based on his assumption that the em era will be more “efficient” in the economic sense and choose efficient institutions, and since war is negative-sum it doesn’t seem to belong. The contrary perspective from Edward Luttwak and Greg Cochran (and perhaps Joseph Henrich) is that war is one of the most important selective pressures that have caused the proliferation of more “efficient” institutions.

At the same time that I was surprised at the lack of space given to war, I was surprised by how much discussion there was of conflict within emulation “clans” and the ways clans would seek to minimize such conflict (especially publicly). Clans are descended from a common copy of a brain/mind, and per Hanson would be even more similar than identical twins are to each other. One of the problems with “group selection” in evolutionary biology has long been that genetic variation within a population is greater than the distance between populations, so that a free-rider can take advantage. Culture is another story, hence the cultural evolution discussed by Henrich. Em clans will be more unified relative to outsiders, which would seem to make inter-group conflict more important than intra-group conflict. Hanson thinks that having copies of em clans in all of the relatively small number of em cities around the world will make conflict less likely, but as noted I am less convinced. On the other hand, I myself have noted that the salience of differences in context dependent, so it’s possible that within-clan differences which seem tiny enough to dismiss in my eyes will be more subjectively important to the ems that actually experience them. The differences he points to which could vary greatly among clan members “include wealth, mood, vigor, recently acquired skills, knowledge of local conditions, and loyalties and ties to specific individual ems”. I’m not sure why “vigor” is included among them, as I hadn’t heard of it being particularly malleable, but he has had more recent thoughts about that factor among humans today.

I mentioned earlier that I agree with Hanson that rogue AI is unlikely to suddenly “FOOM” and take over the world before emulations come to the fore. An argument by analogy he gives in the book is that “In our world, most firms, cities, nations, and other organizations are much more powerful than are most individuals, and yet they remain largely under control in most important ways”. He had earlier made such an argument by analogy to a “City-ularity“. But in his imagined emulation world, there really will just be a few major em cities due to the greater costs imposed by distance due to the subjective time experienced by the smallest and fastest, and thus most centrally located and high-status ems. His sketched out/hoped for scenario of computing capabilities being the last requirement for ems (after brain scanning and cell modeling techniques) is supposed to result in thousands of different emulations all at once rather than a first-mover dominating. The few cities on the other hand, will have large first-mover advantages. This is of particular interest to me, since I tend to think that city-states are the ideal scale of governance (they are less able to mobilize large armies than nation-states, but that has been less important in our more peaceful industrial era) and that letting a thousand nations bloom will result in more innovation and “efficient” institutions. Currently cities considered very desirable like New York and San Francisco are adopting many housing/real-estate regulations that most economists would consider to be very inefficient. Hanson suggests combinatorial auctions as an efficient means of zoning here, while at his blog more recently he’s discussed some taxation schemes designed with a similar end. But the same incentives which result in cities adopting inefficient policies now could result in them not adopting his proposed institutions, and the first-mover advantage of those few cities could overcome that inefficiency just as the smaller advantage of superstar cities today permits them their inefficiencies. Perhaps Hanson is hoping that by writing in advance the emulations who come later will be read his suggestions when creating their cities, but few academics have that much influence. I think Hanson’s proposed predictions markets are a great idea we could implement right now to essentially dissolve a lot of verbal disagreements and improve a lot of organizations, but he hasn’t made much progress on putting them into practice.

There is much more that could potentially be written in response to the book, but I think most of that should be left to subject-matter experts. The relevant knowledge I have mostly stems from Hanson’s blogging, which limits the amount of value I can add. If this comes across as a self-serving excuse for not reading and reviewing that advanced copy when I could have theoretically been of some help, that’s a topic for Hanson’s second book, “The Elephant in the Brain”, co-authored with Kevin Simler, whose work I have highlighted here before. I’ll probably get around to reviewing it years after it has faded out of conversation within our corner of the blogosphere, or what remains of it.