Digital Phylacteries and the Simulated Afterlife

If I’m right to endorse the Hofstatderean idea of the self (that is, that self is a pattern, and any instance of that pattern is an instance of the self), then I probably have to support the idea that if my pattern were replicated in a computer, it would also be a version of myself. Again, as always, it might be a poor-quality or imperfect version, but it would still be a version (instance? copy?). As a consequence, a person might want to extend their lives after death by uploading their consciousness to a hard-drive. If you believe in some sort of essential self that is not contained in any pattern, or if you require some historical continuity in order for identity to persist from one instance to the next,1 then you might not identify with a copy of your consciousness (and sub-consciousness, presumably) on a computer. But if you ultimately accept Hofstadter’s idea, then you should accept that the simulation on the computer really is you, to the extent that it is a faithful representation of your pattern.

So let’s talk about afterlives.


Christian Hendriks 2015; detail of a diorama at the Field Museum

In transhumanist communities—I’m exclusively familiar with what’s called the Less Wrong diaspora, and then only slightly—there are people who hope that, in the future, they will be resurrected on a computer.

Buckle your seatbelts: I’m about to go over some recurring Less Wrong ideas and if you’re not used to them, you’ll probably find them very strange.

Many members of the Less Wrong community believe strongly that artificial superintelligence is inevitable; the only question is what kind of superintelligence it will be. They greatly fear an AI that will be either malevolent towards humans or simply indifferent to humans. The latter would be as bad as the first, since human bodies could contain resources the AI will desire for its projects and the AI would have no reason not to harvest humans for those resources. Therefore they are working on ensuring that a human-friendly artificial superintelligence is made before any other kind of superintelligence. They do not call this human-friendly AI “the AI Messiah,” but I do and I think it will be a helpful name to go on with.

Many members of the Less Wrong community also believe that immortality is an unambiguous good. They seek to overcome death through research into cryonics and various life-extending devices. While they do not think that uploading your consciousness to a computer is yet feasible, some of them do hope that the AI Messiah (should the utopian outcome happen) will look through historical evidence with the aim of simulating in her2 computations all people who have died, reconstructing them based on whatever traces they left behind.


Source: Letitia Smania Donanzan at

The idea that the AI Messiah will reconstruct (ie. simulate) people based on historical traces is somewhat difficult for people to believe, even if those people believe in the Robot Messiah in the first place. How many traces will remain of a person? Are they enough to recreate a pattern faithful enough that you could say it was really the same person? It depends on how superintelligent the AI Messiah is, but you could help if you deliberately left behind as complete a record of yourself as you can. They call it lifeblogging; I call it a digital phylactery, referring to the Dungeons & Dragons object in which a lich stashes a fragment of their soul from which they will always resurrect (think of Horcruxes if you’re more of a Harry Potter fan). Long-term storage is a serious problem that they are trying to address, though I’m not sure they’re addressing it adequately; so far I’ve seen no mention of format obsolescence or inherent vices, or the data migration solutions that mitigate those problems. I smell a consulting opportunity for any archivists, records managers, and other information science people. At any rate, there’s an interesting conversation going on around this idea.

There are several reasons I find this version of the afterlife unsatisfying: one, I seriously doubt the future existence of a AI Messiah;3 two, I seriously doubt any AI Messiah would deem it necessary or good to resurrect/simulate the dead; three, I seriously doubt that any AI Messiah could recreate a human accurately enough from historical traces for that pattern to really be mine, unless that AI Messiah decided to recreate every possible human (which would be a worthwhile science fiction premise). However, thinking about digital phylacteries and AI Messiahs and simulated afterlives has been somewhat useful for me: it suggests that Hofstadterean ideas of the soul are conducive to afterlives of all sorts.

If you follow Accidental Shelf Browsing, you might have worked out by now that I’m a practicing Christian, though far from a conventional one. It has been quite some time since I’ve had much use for speculating about afterlives: not only do I feel (with Confucius, I guess) that most of this speculation is groundless blue-sky speculation and without any value for human endeavour, but I also think that worrying too much about the afterlife is dangerous. Terror management theory suggests that relying on faith in the afterlife to avoid your fear of death can make you inhospitable, hostile to alternative viewpoints, and even violent. My primary concern when reasoning/speculating about afterlives has been whether to endorse what’s called dogmatic universalism: the view that all people will in the end be reconciled to and reunited with God (and therefore neither annihilated nor consigned to Hell for eternity). For me, this has been an issue less about the fate of human souls and more about the nature of God and God’s mercy and justice. However, working through Hofstadter’s ideas did give me some ideas, albeit blue-sky speculative ones, about afterlives and how they might work.

If humans are patterns, then many claims about resurrection become a lot easier to understand. If there is a God who knows humans absolutely and who has creative capacities, it does not seem unreasonable that that God could recreate the patterns which make those humans. God would have to make those patterns in a medium (or substrate, or carrier) of some kind; what medium that is could be up for speculation. God could also free those patterns of any self-destructive tendencies they might have picked up. This view also reconciles arguments about spiritual v. bodily resurrection: in a Hofstadterean view, the pattern that is your soul probably includes the pattern for your body.

Of course, this is a somewhat saccharine view of the resurrection; Christianity tends to have at its centre a sense that death is difficult to overcome and that God could not simply remake humans in perfected form. Typically sin is understood to be a component in this problem: sin makes resurrection impossible without intervention on God’s part. Why should God need the crucifixion to achieve it? This is a question I’ll try to address later. For now, I just want to observe that this idea of soul-as-pattern not only seems consistent with common Christian views about the afterlife, but it might make them easier to understand and reason about. (I would appreciate correction from theologians excluding Edward Feser, however.)

1. Do you care about historical continuity? What counts as historical continuity? The best discussion of this that I know of involves Davidson’s Swampman thought experiment. Lightning strikes and kills a man hiking in the swamp; simultaneously, another lightning bolt strikes the swamp nearby, spontaneously rearranging a bunch of molecules. By coincidence, those molecules take on exactly the same form of the man hiking through the swamp at the moment of his death. This second man, whom Davidson calls “Swampman,” is identical in all ways to the first man, and so would act exactly the same as the first man, have all the first man’s memories, and so on. Davidson argues that there is a difference between Swampman and the first man, though not one that anyone would notice. I disagree with Davidson; do you?

2. I am gendering the AI Messiah “her” because this quasi-religion is increasingly converging on traditional Gnosticism, in which the rational, human-friendly, disembodying, emancipatory figure is the explicitly feminine Sophia. If you doubt my comparison between transhumanist Singularity-mongering and Gnosticism, I suggest you read Scott Alexander’s “The Goddess of Everything Else,” which in the context of Alexander’s work makes the Gnostic element pretty clear. (The Goddess of Cancer, in other places known as Moloch, is clearly the Demiurge; while cryonics is popular in this crowd, thanks to entropy all resurrection must eventually be disembodied.)

3. This is not to say that I seriously doubt the fact that advanced artificial intelligences will have a significant impact on human life at some point in the far future, but I do doubt that such intelligences will take on the Demiurge/Messiah roles cast by the communities I’m describing.

2 thoughts on “Digital Phylacteries and the Simulated Afterlife

  1. Pingback: Tzimtzum and Inherent Vice: A Personal Myth of Pattern | Accidental Shelf-Browsing

  2. Pingback: Clouds, Daffodils, and Jam That Will Not Come Together Again: A Series Index | Accidental Shelf-Browsing

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s