# How you might see some of the dead again

The atheist position is that we won't see people who have died again. I don't disagree. But that's not the same as saying it's impossible, with or without the supernatural. There's an actual chance that's greater than epsilon. But it's still very small, I think it would be unwise to hinge any hopes on it, or to mislead (as I think Kurzweil can do) that it's almost assured given better technology.

The supernatural possibility is of course possible. Perfect souls might exist separate from the body and somehow encode all memories and thoughts separately from the brain, perfectly, so that even if the body is brain damaged and lives on another 30 years, the soul, while retaining that, is fully self-aware and has access to all memories and thoughts across all of its existence, and so is unharmed. And it may be that souls are reborn, or that they go have a party together after leaving their bodies, and remember each other (perhaps even across multiple lives if there is some rebirth before the party). Maybe that party's just the pre-party and once enough people arrive they all come back to Earth (some of them having just left!) in new bodies to have the real party. There are many ways, if you grant supernatural possibilities, for you to see the dead again, either while you're still alive or when you too die. Atheists could be wrong. But probably not. So that leaves us with the natural possibility. How might a form of resurrection work?

Is your grandparent the same person they were when they were in their late 20s? If your grandparent suddenly develops Alzheimer's and seems to lose many long term memories, but isn't all the way gone yet, are they still the same person? If they develop into full-blown unable-to-care-for-themselves Alzheimer's, brain cell death and all, are they the same person? If a cure is developed for Alzheimer's before they get to that point, and they receive the cure, so they stop losing long term memories and can form and retain new ones, are they the same person?

Either "yes" or "no" are valid answers to all those questions. It's fuzzy. But hopefully you can see the same, that each question could be "yes" or could be "no", and it probably comes down mostly to disagreements on what the words "person" and "same" mean.

If you look at people's revealed beliefs by their attitudes around those with Alzheimer's, I think you'll see a sort of curve where the person with Alzheimer's becomes less and less "the real them" to everyone around them. But every now and again there might be a spark, maybe it's even possible to routinely trigger it, or maybe the person holds on to one characteristic of their old selves, and so people are led to believe there's still some kernel of "the real them" inside, damaged and fragmented as it may be.

People are finitely complex. At the material level, there are only so many ways for the atoms in the brain to be rearranged. It's a huge number, but it's finite. So in principle, with infinite computing power, by brute force you could test every possible arrangement of brain matter until you found one you'd call "the same" as the person who is dead. And once you found one, you'd find many.

If you found one, and had a way to grow a new body for it or otherwise emulate the biological input/output process, is this not the same person? If you take a snapshot of all the atoms in your brain 10 seconds ago, and reproduced another instance of that, is this not the same person you are, just a bit younger? They know all the things you know, except anything you learned in that 10 seconds, but they still think like you so they would process the new information the same way. They are you.

Obviously we don't have infinite computing power, so this approach can't work. What approaches can, then?

We go back to the statement that people are finitely complex. But can we bound that any tighter? How many bits of complexity are "you", to whatever arbitrary closeness to whatever point of time you want to use as "you"? And what can we learn about you that can fill in some of these bits if we don't have all of them? Male and female minds are different, so if I know which one you are, I've learned one bit of information, and eliminated half of all the possibilities.

DNA is probably huge for how many bits it gives you about a person. If you really want to entertain the idea of resurrection of someone, you better keep some sort of DNA sample that could be used in the distant future. If you can make a perfect clone, a lot of the guesswork on the initial state of the brain is taken care of. You've greatly reduced the number of possible configurations, because the brain of the person you're looking for had to have evolved somehow from the starting one.

But even if you don't have a nice vial of spit carefully preserved, if the dead person is a relative then you can probably get a lot of the same details by the fact that there's you and you partially share DNA. For non-relatives, you might not be related, but you can find relations. The more relatives, the better. If you've got buried corpses, especially of the dead person, even better. Still not enough? Then maybe there's trace or touch DNA you don't even know about -- kept some random jewelry? Maybe there's something. Or if their old house hasn't been totally blown up, perhaps a deep search would find something, especially when you have information on what you can exclude and what you're looking for.

So by some means you've got a model for what their brain should start out with. Perhaps if you find a fingerprint, you can restrict your model to account for some of the stochastic changes that occur in the womb, since identical twins still lack the same fingerprint. Your next challenge is finding out how this model evolves, how it changes, into something you'd call "the same person" you once knew.

Kurzweil talks of memory extraction of living people for this. It's definitely a valid technique, but as people's memories aren't always that great, you have to treat it with some caution. But as a general rule, when you're trying to figure out how to evolve, shape, mold, or rearrange the person's starting brain model you have from their DNA, you should end up with something that conforms to the models of people who knew the person. If everyone remembered the person as being shy and reserved, then you shouldn't waste effort looking for brain models that aren't that way.

Besides memories, there are things like handwriting samples, video, voice recordings, purchasing decisions, interests, addictions, medical data, employment history, and many other rather objective details. Some things are time-stamped. They can all be used to constrain the type of model you search for. I don't know what the searching algorithm would look like, but again to at least try and match the idea of "same person" it has to match certain details at certain points of its evolution and have at least one plausible path of changes to get from one point to another.

If you think of a Go or Chess game that has been underway for several moves, freeze the current state of the board. Can you find that state from all the possible states? How much information do you need? If I tell you that white is down a pawn, or that there's two more black stones than white stones, you know that at some point in the game's history a capture occurred. There are probably many ways to get to the same state, but all the ways have to involve a capture somewhere.

And so it might be that a superintelligence, or at least vastly faster and smarter computers and humans working together, can narrow down and produce a mind, and embed it in a body or otherwise identical emulation if you're ok with that, and you have resurrected someone. They talk like you remember, they remember your shared experiences together (or at least most of them), as you watch them explore the new world and do and learn new things you see they're acting in a way you'd have expected.

But still the question: is it the same person?

And the answer is still the other question: what do you mean by "same" and "person"?

At least for some people, some of these resurrections will be same-enough, person-enough, to count. Some other people may never accept them. And some may mostly accept them but still have in the back of their mind "the real person is gone forever..."

Your own opinion on the matter is irrelevant. Without anything to stop them like laws against modeling minds in detail, people of the future will try this. If they're successful at all, some people will get great comfort from it, and some of the resurrections will be very happy to be going again.

If the Singularity is Near, then you might have the chance to live to see this, might have the chance to see dead people live again, and might have the chance to die and be brought back even later yourself. If superintelligence is at all possible, and such a universe with that possibility realized is at all friendly to human values, this seems so possible that I can understand why people like Kurzweil might give off the idea of inevitability, and for critics to use that as just another tired "hey look at these dumb atheist tech nerds still having faith in dumb religious things like resurrection" remark. It's not inevitable, and the belief of its possibility however strong isn't anything like a faith.

I don't even personally think it's likely for myself to experience any aspect of this, I miss a number of people and feel sad I'll never see them again. But I do think this sort of resurrection is likely in the realm of what a superintelligence could do if it wanted. I'm reluctant to believe very strongly in limitations on what superintelligence could do -- could any human of ancient Egypt before the Great Pyramids foresee how the world looks today or would they think such things impossible and beyond the limits of humanity? Could they even foresee just the Great Pyramids?

Of course this means if you make a good in-principle argument without invoking the supernatural or infinite computing power about possibility, I'm inclined to agree it's likely doable by a superintelligence and maybe even likely a friendly superintelligence does such a thing.

#### Posted on 2017-06-02 by Jach

LaTeX allowed in comments, use $\\...\\$\$ to wrap inline and $$...$$ to wrap blocks.