I recently had the opportunity to see George Mason Professor Robin Hanson talk about his book, The Age of Em. I also was able to work my way into having a long conversation with him after his presentation.
For those who don’t know, it’s perhaps the strangest book you’ve ever heard of. Hanson looks to project forward in time when the technology exists to easily upload human brains into computer simulations. These “emulated” brains will have certain characteristics from residing in computer hardware: they can make copies of themselves, save versions of themselves for later, or delete versions of themselves. They will even be able to run faster or slower than normal human brains depending on what hardware they are running on. Hanson spends the book working through the implications of this new society. And there are a lot of fascinating insights.
Hanson discusses the pure physics of this world, as suddenly speed of light delays in communication mean a lot; if an em is running at a million times human speed, then a bad ping of 50 ms is equivalent to over 12 hours for a message to get sent today. This leads to very close physical locations of ems, which concentrates them in large cities. Their economy also grows much faster than ours due to the rapid speed at which their brains are thinking, although they may be physically restrained by how quickly the physical manufacturing of their hardware can occur. The economy also quickly moves to subsistence wages, as even the most productive members of society can have their brains copied as many times as needed to fill all roles. Elon Musk is no longer a one of kind genius, and in fact anyone who cannot compete with an Elon Musk version in their job would likely be cast aside. For a more detailed summary and examples of bizarre ideas, I recommend Part III of Scott Alexander’s post on the book.
In that blog post, Scott goes on to discuss in Part IV the problem of value drift. Hanson does a good job pointing out that past human societies would not have approved of what we now consider acceptable. In some areas, the change in values in stunning. Merely 10 years ago, many had reservations about gay marriage. Merely 50 years ago, many Americans had serious reservations about interracial marriage. On the scale of humans’ existence as a species, the amount of time we have accepted that people have the right to worship their own religion is minuscule. The section of human history where subsistence existence was not the only option is likewise small. Professor Hanson told our group that by far the most common reaction to his painting of the future was rejection.
I even asked him specifically about it: Hanson had stated several times that it was not his job or intention to make us like or hate this future, only to know about it. I pointed out that many AI researchers were very concerned about the safety of artificial intelligence and what it might do if it hits an intelligence explosion. To me, there seems to be little difference between the AI intelligence explosion and the Em economy explosion. Both would be human creations, making decisions and changing their values rapidly, at a pace that leaves most “normal” traditional physical humans behind. If many of the smartest people studying AI think that we should do a lot of work to make sure AI values line up with our own, shouldn’t we do the same thing with Ems? Hanson’s answer was basically that if we want to control the value systems of our descendants thousands of mental years in the future, well good luck with that.
Scott in Part IV of his review demonstrates the problem with just allowing this value drift to happen. Hanson calls the era we live in the “dream time” since it’s evolutionarily unusual for any species to be wealthy enough to have any values beyond “survive and reproduce”. For most of human history, there wasn’t much ability to build cities or share knowledge because too many resources were focused on survival. Today, we have become so productive and intelligent that humans have elevated Earth’s carrying capacity high above the amount of people we have. We don’t have to spend all our resources on survival and so we can come up with interesting philosophical ideas about morality and what the meaning of life is. We’ve also harnessed this evolutionary competitiveness to fuel our market economy where the determiner of what survives isn’t nature, but human desires. Unfortunately when you switch to the Age of Em, suddenly the most productive part of the economy is plunged back into a Malthusian trap with all resources going to keep the Ems alive. Fulfilling human wants may be what drives the economy, but if there are other pressures on Ems, they will be willing to sacrifice any values they have to keep themselves alive and competitive. If the economy gives up on fulfilling human demand, I wouldn’t call that a drift in values, I’d call that an absence of values.
If we live in the dream time, then we live in a unique situation where only we can comprehend and formulate higher morality and philosophical purpose. I think we should take advantage of that if we can.
Hanson’s observations given his assumption that the Age of Em will happen are excellent, considering he is predicting far into the future. It’s likely things won’t work out exactly this way, as perhaps a single company will have a patent on brain scanning for a decade before the market really liberalizes; this could seriously delay the rapid economic growth Hanson sees. He acknowledges this, and keeps his book more of a prediction of what will happen if we don’t oppose this change. I’m not sure how far Hanson believes that regulation/intellectual property will not be able to thwart the age of em, but it seems that he’s more confident it will not be stopped than that it will be. This may be an economist mistake where regulation is sort of assumed away as the realm of political science. It’s not unprecedented that weird inefficient institutions would last far into the future. Intellectual property in the digital age is really weird, all things considered. Software patents especially seem like a way to patent pure logic. But there are others: banking being done with paper checks, daylight savings time, the existence of pennies, and, of course, Arby’s. There are also plenty of examples of new technologies that have evolved much faster than regulation, like supplements, e-commerce, and ride-sharing. It remains to be seen what brain emulations will be.
There is also the possibility that emulated brains won’t be the next big shift in human society. Hanson argues that this shift will rival that of the agricultural revolution and the industrial revolution. This makes a lot of sense if brain emulation is indeed the next big change. Eliezer Yudkowsky (and Scott) think this is incorrect and artificial intelligence will beat it. This seems like a real possibility. Scott points out that we often come up with technological equivalents of human biology far before actually emulating biology. This is mostly because biology has accidentally figured things out via evolution and thus it is often needlessly complicated. For example, aircraft usually fly via fixed wing aerodynamics, not by flapping. It seems likely that we will reach human level problem solving via software rather than via brain scanning. Even if we don’t, it seems likely that software could quickly optimize a simulation based on a preliminary brain scan that was too rough to get a proper brain emulation into hardware. But software assisted reconstruction could start experimenting with neuron simulation and create a software assisted brain emulation that is better designed and more specialized than any human brain emulation.
It also seems possible that other things could happen first that change human history, like very expensive climate change, a crippling pandemic (anti-biotic resistance), genetic and epigenetic engineering and of course some technological revolution we haven’t even imagined (the unknown). Certainly if we assume continued economic growth, either brain emulation, artificial intelligence, or genetic engineering seem like likely candidates to transform humanity. Hanson thinks AI research is really overrated (he used to be an AI researcher) and isn’t progressing very fast. But he was an AI researcher about 25 years ago and we’ve seen some pretty impressive improvements in machine learning and natural language processing since then. We’ve also seen some improvement in brain emulation technology as well to be fair. Genetic engineering was hailed as the next revolution in the 1990s, but has floundered ever since. Last year though, the use of CRISPR in genome engineering has dramatically increased the feasibility of actually picking and choosing specific genes. Any of these could drastically change human society. Perhaps any genetic improvements would be overshadowed by brain emulation or AI. I guess it depends on the importance of the physical world vs the digital one.
Of course, not all changes could be from improved technology. There’s a significant risk of a global multi-drug resistant pandemic. Our overuse of antibiotics, the difficulty in making everyone stop overusing them, and our highly integrated world means we’ve created an excellent scenario for a superbug to appear and spread. Anything resembling the 1918 Spanish Flu Epidemic could be devastating to the world population and to economic growth. Climate change poses a similar risk to both life and the economy. If either of these were to happen, it could significantly deter the Age of Em from occurring or at least delay it, along with a lot of the progress of our civilization. And that’s not even mentioning additional freak natural disasters like coronal mass ejections.
Overall, predictions are very difficult and if I had to bet, I’d bet that the next big change in human civilization won’t be emulated brains. A good competitor is definitely artificial superintelligence, but when you add in genetic engineering, natural disasters, drug resistant bacterial epidemics, and so on, you have to take the field over brain emulations.
Nonetheless, this book really does make you think about the world in a different way with a perspective both more global and more forward looking. It even makes you question what it means to be human. The ins and outs of the 2016 election really fade away (despite my continued interest and blogging). Political squabbling doesn’t compare to the historical trends of human civilization and the dawn of transhumanism.