Continuity in Uploading

I don't acknowledge an upload as "me" in any meaningful sense of the term; if I copied my brain to a computer and then my body was destroyed, I still think of that as death and would try to avoid it.

A thought struck me a few minutes ago that seems like it might get around that, though. Suppose that rather than copying my brain, I adjoined it to some external computer in a kind of reverse-Ebborian act; electrically connecting my synapses to a big block of computrons that I can consciously perform I/O to. Over the course of life and improved tech, that block expands until, as a percentage, most of my thought processes are going on in the machine-part of me. Eventually my meat brain dies -- but the silicon part of me lives on. I think I would probably still consider that "me" in a meaningful sense. Intuitively I feel like I should treat it as the equivalent of minor brain damage.

Obviously, one could shorten the period of dual-life arbitrarily and I can't point to a specific line where expanded-then-contracted-consciousness turns into copying-then-death. The line that immediately comes to mind is "whenever I start to feel like the technological expansion of my mind is no longer an external module, but the main component," but that feels like unjustified punting.

I'm curious what other people think, particularly those that share my position on destructive uploads.

---

Edited to add:

Solipsist asked me for the reasoning behind my position on destructive uploads, which led to this additional train of thought:

Compare a destructive upload to non-destructive. Copy my mind to a machine non-destructively, and I still identify with meat-me. You could let machine-me run for a day, or a week, or a year, and only then kill off meat-me. I don't like that option and would be confused by someone who did. Destructive uploads feel like the limit of that case, where the time interval approaches zero and I am killed and copied in the same moment. As with the case outlined above, I don't see a crossed line where it stops being death and starts being transition.

An expand-contract with interval zero is effectively a destructive upload. So is a copy-kill with interval zero. So the two appear to be mirror images, with a discontinuity at the limit. Approach destructive uploads from the copy-then-kill side, and it feels clearly like death. Approach them from the expand-then-contract side, and it feels like continuous identity. Yet at the limit between them they turn into the same operation.

Comments

sorted by
magical algorithm
Highlighting new comments since Today at 2:48 AM
Select new highlight date
Rendering 50/88 comments  show more

I agree that uploading is copying-then-death. I think you're basically correct with your thought experiment, but your worries about vagueness are unfounded. The appropriate question is what counts as death? Consider the following two scenarios: 1. A copy of you is stored on a supercomputer and you're then obliterated in a furnace. 2. A procedure is being performed on your brain, you're awake the entire time, and you remain coherent throughout. In scenario 1 we have a paradigmatic example of death: obliteration in a furnace. In scenario 2 we have a paradigmatic example of surviving an operation without harm. I would say that, if the procedure in 2 involves replacing all or part of your brain, whether it is performed swiftly or slowly is unimportant. Moreover, even if you lost consciousness, it would not be death; people can lose consciousness without any harm coming to them.

Note that you can adjust the first scenario - say, by insisting that the copy is made at the instant of death or that the copying process is destructive as it transpires, or whatever - but the scenario still could go as described. That is, we are supposed to believe that the copy is a continuation of the person despite the possibility of inserting paradigmatic examples of death into the process. This is a clear case of simply stipulating that 'death' (and 'survival') should mean something entirely different. You can't hold that you're still speaking about survival, when you insist on surviving any number of paradigmatic cases of death (such as being obliterated in a furnace). There are no forms of death - as we ordinarily conceive of death - that cannot be inserted into the uploading scenario. So we have here as clear a case of something that cannot count as survival as is possible to have. Anybody who argues otherwise is not arguing for survival, but stipulating a new meaning for the words 'survival', 'death', etc. That's fine, but they're still dead, they're just not 'dead'.

I think this realisation makes understanding something like the brain transplant you describe a little easier. For we can say that we are living so long as we don't undergo anything that would count as dying (which is just to say that we don't die). There's nothing mysterious about this. We don't need to go looking for the one part of the body that maintains our identity under transformation, or start reifying information into a pseudo-soul, or whatever. We just need to ensure whatever we do doesn't count as death (as ordinarily conceived). Now, in the case of undergoing an operation, there are clear guidelines. We need to maintain viability. I cannot do certain things to you and keep you alive, unless I perform certain interventions. So I think the answer is quite simple: I can do anything to you - make any change - as long as I can keep you alive throughout the process. I can replace your whole brain, as long as you remain viable throughout the process, and you'll still be alive at the end of it. You will, of course, be 'brain dead' unless I maintain certain features of your nervous system too. But this isn't mysterious either; it's just that I need to maintain certain structural features of your nervous system to avoid permanent loss of faculties (such as motor control, memory, etc). Replacement with an artificial nervous system is likewise unproblematic, as long as it maintains these important faculties.

A lot of the confusion here comes from unnecessary reification. For example, that the nervous system must be kept structurally intact to maintain certain faculties, does not mean that it somehow 'contains' those faculties. You can replace it at will, so long as you can keep the patient alive. The person is not 'in' the structure (or the material), but the structure is a prerequisite for maintaining certain faculties. The common mistake here is thinking that we must be the structure (or pattern) if we're not the material, but neither claim makes sense. Alternatively, say you have a major part of your brain replaced, and the match is not exact. Somebody might, for example, point out that your personality has changed. Horrified, you might wonder, "Am I still me?" But this question is clearly absurd. There is no sense in which you could ask if you are still you. Nor can you coherently ask, "Did I die on the operating table?" Now, you might ask whether you merely came into existence on the operating table, after the original died, etc. But this, too, is nonsense. It assumes a reified concept of "self" or "identity." There is nothing you can "lose" that would count as a prior version of you 'dying' and your being born anew (whether slightly different or not). Of course, there are such things as irreversible mental degradation, dementia, etc. These are tragic and we rightfully speak of a loss of identity, but there'd be no such tragedy in a bout of dementia. A temporary loss of identity is not a loss of identity followed by gaining a new identity; it's a behavioural aberration. A temporary loss of identity with a change in temperament when one recovers is, likewise, unproblematic in this sense; we undergo changes in temperament regardless. Of course, extreme change can bring with it questions of loss of identity, but this is no more problematic for our scenario than an operation gone wrong. "He never fully recovered from his operation," we might say. Sad, yes, but this type of thing happens even outside of thought experiments.

All true, but it just strengthens the case for what you call "stipulating a new meaning for the words 'survival', 'death', etc". Or perhaps, making up new words to replace those. Contemplating cases like these makes me realize that I have stopped caring about 'death' in its old exact meaning. In some scenarios "this will kill you" becomes a mere technicality.

Mere stipulation secures very little though. Consider the following scenario: I start wearing a medallion around my neck and stipulate that, so long as these medallion survives intact, I am to be considered alive, regardless of what befalls me. This is essentially equivalent to what you'd be doing in stipulating survival in the uploading scenario. You'd secure 'survival', perhaps, but the would-be uploader has a lot more work to do. You need also to stipulate that when the upload says "On my 6th birthday..." he's referring to your 6th birthday, etc. I think this project will prove much more difficult. In general, these sort of uploading scenarios are relying on the notion of something being "transferred" from the person to the upload, and it's this that secures identity and hence reference. But if you're willing to concede that nothing is transferred - that identity isn't transferrable - then you've got a lot of work to do in order to make the uploading scenario consistent. You've got to introduce revised versions of concepts of identity, memory, self-reference, etc. Doing so consistently is likely a formidable task.

I should have said this about the artificial brain transplant scenario too. While I think the scenario makes sense, it doesn't secure all the traditional science fiction consequences. So having an artificial brain doesn't automatically imply you can be "resleeved" if your body is destroyed, etc. Such scenarios tend to involve transferrable identity, which I'm denying. You can't migrate to a server and live a purely software existence; you're not now "in" the software. You can see the problems of reference in this scenario. For example, say you had a robot on Mars with an artificial brain with the same specifications as your own. You want to visit Mars, so you figure you'll just transfer the software running on your artificial brain to the robot and wake up on Mars. But again, this assumes identity is transferrable in some sense, which it is not. But you might think that this doesn't matter. You don't care if it's you on Mars, you'll just send your software and bring it back, and then you'll have the memories of being on Mars. This is where problems of reference come in, because "When I was on Mars..." would be false. You'd have at best a set of false memories. This might not seem like a problem, you'll just compartmentalise the memories, etc. But say the robot fell in love on Mars. Can you truly compartmentalise that? Memories aren't images you have stored away that you can examine dispassionately, they're bound up with who you are, what you do, etc. You would surely gain deeply confused feelings about another person, engage in irrational behaviour, etc. This would be causing yourself a kind of harm; introducing a kind of mental illness.

Now, say you simply begin stipulating "by 'I' I mean...", etc, until you've consistently rejiggered the whole conceptual scheme to get the kind of outcome the uploader wants. Could you really do this without serious consequences for basic notions of welfare, value, etc? I find this hard to believe. The fact that the Mars scenario abuts issues of value and welfare suggests that introducing new meanings here would also involve stipulating new meanings for these concepts. This then leads to a potential contradiction: it might not be rationally possible to engage in this kind of revisionary task. That is, from your current position, performing such a radical revision would probably count as harmful, damaging to welfare, identity destroying, etc. What does this say about the status of the revisionary project? Perhaps the revisionist would say, "From my revisionary perspective, nothing I have done is harmful." But for everyone else, he is quite mad. Although I don't have a knockdown argument against it, I wonder if this sort of revisionary project is possible at all, given the strangeness of having two such unconnected bubbles of rationality.

Now, say you simply begin stipulating "by 'I' I mean...", etc, until you've consistently rejiggered the whole conceptual scheme to get the kind of outcome the uploader wants. Could you really do this without serious consequences for basic notions of welfare, value, etc?

No, and that is the point. There are serious drawbacks of the usual notions of welfare, at least in the high-tech future we are discussing, and they need serious correcting. Although, as I mentioned earlier, coining new words for the new concepts would probably facilitate communication better, especially when revisionaries and conservatives converse. So maybe "Yi" could be the pronoun for "miy" branching future, in which Yi go to Mars as well as staying home, to be merged later. There is no contradiction, either: my welfare is what I thought I cared about in a certain constellation of cares, but now Yi realize that was a mistake. Misconceptions of what we truly desire or like are, of course, par for the course for human beings; and so are corrections of those conceptions.

You are dodging the question by appealing to the dictionary. The dictionary will not prove for you that identity is tied to your body, which is the issue at hand (not "whether your body dies as the result of copying-then-death", which as you point out is trivial)

Upload shmupload.

Lets remove unnecessary complications and consider the more essential question. You are knocked out. While unconscious, a particle-for-particle copy of you is made with all identical energy levels, momenta, spins, colors, flavors, and any other quantum states associated with any of the particles in your body. The only differences are all the particles in the new copy are 3 m to the east of the particles in the original. The unconscious copies are placed someplace nice and revived approximately simultaneously.

Pretty obviously, neither copy feels like it is the other copy. Pretty obviously, each copy presumes it is the original, or presumes it has a 50:50 chance of being the original, but each copy thinks the same as the other, at least initially.

Further, if this process was done to you without your knowledge, and while still unconscious, the original you was destroyed, when the copy came to, it would have no inkling that it was not you and that you were dead. It would think it was you and that you are not dead.

Is there really any important difference between your existence now, and one in which your physical body was replaced by a particle-for-particle copy every so often? To WHOM or WHAT is that difference experienced?

My own conclusion is that the continuity of my life is something of an illusion. And that my distaste for dying is primarily something bred into me by evolution, it is not hard to imagine how it would provide a significant survival advantage over humans without that trait.

I don't know much what to do with this point of view, but one thing I don't do is pay money now to have my severed head frozen after I legally die. It is enough that I carry through on the more straightforward survival instinct type things that evolution has stuck me with.

Is there really any important difference between your existence now, and one in which your physical body was replaced by a particle-for-particle copy every so often? To WHOM or WHAT is that difference experienced?

Yes, it is of importance to the me right here, right now, in the present. Under one interpretation I wake up in the other room. In the other I do not - it is some other doppelgänger which shares my memories but whose experiences I do not get to have.

If I somehow find myself in the room with my clone, it's true that there's no way short of checking external evidence like security footage or somesuch to determine which is the real me. That is true. But that is a statement about my knowledge, not the world as it exists. The map is not the territory.

If I were to wake up in the other room with the clone nearby, it no longer matters which one of us is the original or not. He isn't me. He is a separate person that just happens to share all of the same memories and motivations that I have. I want to say that I wouldn't even give this copy of me the time of day, but that would be rhetorical. In some ventures he would be my greatest friend, in others my worst enemy. (Interestingly I could accuratly tell which right now by application of decision theory to the variants of the prisoner's delima.) But even when I choose to interfere in his affairs, it is not for directly self-serving reasons - I help him for the same reason I'd help a really close friend, I hurt him for the same reason I'd hinder a competitor.

The truth has real implications for the me that does exist, in the here and now. Do I spend not-insignificant sums of money on life insurance to cover cryonic preservation for me and my family, thereby foregoing other opportunities? Do I consider assisted suicide and cryonic preservation when I am diagnosed with a terminal or dibilitating disese of the brain? Do I stipulate revival instead of uploading in my cryonics contract, knowing that it might mean never being revived if the technology can not be developed before my brain deteriorates too much? Do I continue to spend time debating this philosophical point with other people on the Internet, in the hope that they too choose revival and there is safety in numbers?

Under one interpretation I wake up in the other room. In the other I do not - it is some other doppelgänger which shares my memories but whose experiences I do not get to have.

I don't understand how to distinguish "the clone is you" from "the clone is a copy of you". Those seem like identical statements, in that the worlds where yon continue living and the world where the clone replaces you are identical, atom for atom. Do you disagree? Or do you think there can be a distinction between identical worlds? If so, what is it?

He isn't me. He is a separate person that just happens to share all of the same memories and motivations that I have.

In the same sense, future-you isn't you either. But you are willing to expend resources for future-you. What is the distinction?

Is there really any important difference between your existence now, and one in which your physical body was replaced by a particle-for-particle copy every so often? To WHOM or WHAT is that difference experienced?

Yes, to the universe as witnessed by an outside observer, and to the law. The important difference is haecceity, which to an naive inside view is currently meaningless, but to any intuitive observer or reflective agent becomes relevant. Objective history exists, it's just that we humans-within-universe simply cannot access it.

There are countless unknowns about the universe that we know about. There are also almost certainly unknown unknowns. Haecceity currently lies essential to sense of self, but in a more broadly aware context perhaps its value could be concretized and rationalized.

How do you feel about the continuous uploading procedure described in "Staring into the Singularity"?

It's not the book, it's the story.

Moby Dick is not a single physical manuscript somewhere. If I buy Moby Dick I'm buying one of millions of copies of it that have been printed out over the years. It's still Moby Dick because Moby Dick is the words, characters, events etc. of the story and that is all preserved via copying.

A slight difference with this analogy is that Moby Dick isn't constantly changing as it ages, gaining new memories and whatnot. So imagine that Melville got half way through his epic and then ran out of space in the notebook that I want you to also imagine he was writing it in. So we have a notebook that contains the first half of Moby Dick (presumably, this is a pretty big notebook). Then he finishes it off in a second notebook.

Some time later he pulls a George Lucas and completely changes his mind about where his story was going ("Kill off Ahab? What was I thinking?") and writes a new version of the story where they go into a profitable, if ethically dubious, whaling business with rather more success than in the first version. This is then written up in a third notebook. Now we have three notebooks, the last two of which are both legitimate continuations of the first, carrying on from the exact same point at which the first notebook was ended.

There is no interesting sense in which one of these is some privileged original, as Eliezer puts it. If you can't get your head around that and want to say that, no, the published (in real life) version is the 'real' one imagine that the published version was actually the third notebook. There is no equivalent of publication for identity that could confer 'realness' onto a copy. In real life, neither or them are notebook 1 but they're both continuations of that story.

If Will Riker discovers that he was non-destructively copied by the Transporter and that there's another version of him running around, he will likely think, 'I don't acknowledge this guy as 'me' in any meaningful sense.' The other guy will think the same thing. Neither of them are the same person they were before they stepped into the Transporter. In fact, you are not the same person you were a few seconds ago, either.

Identify yourself as the book and your concept of identity has big problems with or without uploading. Start by reconciling that notion with things like quantum physics or even simple human ageing and you will find enough challenges to be getting on with without bringing future technology into it.

But you are not some collection of particles somewhere. You are the story, not the book. It's just that you are a story that is still in the process of being written. Uploading is no different that putting Moby Dick on a Kindle. If there's still a meat version of you running around then that is also a copy, also divergent from the original. The 'original' is (or was) you as you were when the copy was made.

Moby Dick is not a single physical manuscript somewhere.

"Moby Dick" can refer either to a specific object, or to a set. Your argument is that people are like a set, and Error's argument is that they are like an object (or a process, possibly; that's my own view). Conflating sets and objects assumes the conclusion.

What's the point of uploading if we have an AI with all the skills and knowledge of everyone not information-theoretically dead at the time of its creation?

I have no idea how to argue with the ideas about consciousness/identity/experience/whatever that make uploading seem like it could qualify as avoiding death. It occurs to me, though, that those same ideas sorta make uploading individuals pointless. If strong AI doesn't happen, why not just upload the most useful bits of people's brainstates and work out how to combine them into some collective that is not one person, but has the knowledge and skills? Why install, say, Yudkowsky.wbe and Immortalbob.wbe, when you could install them as patches to grandfather_of_all_knowledge.mbe, effectively getting (Yudkowsky|Immortalbob)?

Assume that any properties of a brain that can be quantified get quantified, and where the variables match up, the or process takes the maximum. So if brain A is better at math than brain B, brain A|B uses A's math ability. If brain M is male and brain F is female, though, brain M|F will learn from both perspectives (and hopefully be stronger than the originals for it).

So the benefits of a group, with none of the drawbacks, all crammed into one metabrain.

Because I want to be alive. I don't just want humanity to have the benefit of my skills and knowledge.

Because I want to be alive. I don't just want humanity to have the benefit of my skills and knowledge.

When I read this in the recent comments list, I at first thought it was a position against uploading. Then I read the other recent comments and realized it was probably a reply to me.

I get the impression that no one has a functional definition of what continuity of identity means, yet destructive copies (uploads, teleports, etc) appear to be overwhelmingly considered as preserving it at least as much as sleep. I find this confusing, but the only argument that seemed to support it that I've found is Eliezer's "Identity is not in individual atoms", which is a bit disingenuous, in that uploads are almost certainly not going to be precise quantum state replicators.

(I'd make a pole, here, but my last attempt went poorly and it doesn't appear to be standard markup, so I don't know where I'd test it.)

What probability would you assign to each of these as continuing personal identity?

  1. Sleep.
  2. puberty
  3. The typical human experience over 1-5 years.
  4. Gradual replacement of biological brain matter with artificial substitutes.
  5. Brain-state copying (uploads, teleportation)
  6. Brain-state melding (Brain Omega = Brain A | Brain B | Brain n )

1) 1.0 2) 1.0 3) 1.0 4) It depends on the artificial substitutes :) If they faithfully replicate brain function (whatever that means), 1.0 5) Again, if the process is faithful, 1.0 6) It really depends. For example, if you drop all my memories, 0.0. If you keep an electronic copy of my brain on the same network as several other brains, 1.0. in-between: in-between

(Yes, I know 1.0 probabilities are silly. I don't have enough sig-figs of accuracy for the true value :)

I don't think most people who believe uploading qualifies as avoiding death would also agree that adding a fraction of a person's brainstate to an overmind would also qualify as avoiding death.

I think the expansion and contraction model, as you've described it, would probably also result in my death. The being that includes the computer and myself would be a new being, of which I am now a component. When the meaty component dies, this is my death, even though there is now a being who perceives itself to be a continuation of me. This being is, in many ways, a continuation of me, just not in the way that I care about most.

I'm not completely sure of this, of course, but anywhere I'm not sure whether I'll die or not I prefer to lean heavily towards not dying.

I think we have to give up on uniqueness of identity in order to remain consistent in these kind of sci-fi scenarios.

edit: And I guess "identity" has to have a continuous value to - similar to the anthropic principle - being x% certain you are in a particular world is like being x% certain you are a particular person.

I don't acknowledge an upload as "me" in any meaningful sense of the term

What about if your mind is uploaded, then downloaded into your clone previously grown without any brain functions? Would you consider the new meat you as "you"?

Suppose that rather than copying my brain, I adjoined it to some external computer in a kind of reverse-Ebborian act; electrically connecting my synapses to a big block of computrons that I can consciously perform I/O to. Over the course of life and improved tech, that block expands until, as a percentage, most of my thought processes are going on in the machine-part of me. Eventually my meat brain dies -- but the silicon part of me lives on.

This is very similar to the premise of Greg Egan's short story, "Learning to be me".

I find this an immensely valuable insight: continuity, or "haecceity", is the critical element of self which naive uploading scenarios dismiss. Our current rational situation of self as concept-in-brain has no need for continuity, which is counterintuitive.

We know a good deal about the universe, but we do not yet know it in its entirety. If there were an observer outside of physics, we might suspect they care great deal about continuity, or their laws might. Depending on your priors, and willingness to accept that current observational techniques cannot access all-that-there-is, it might be worth embedding some value to haecceity near your value of self.

Contrast grow-and-prune uploading with slice-and-scan uploading: the latter will be anathema to the vast majority of humanity; they may "get over it", but it'll be a long battle. And slice-and-scan will probably be much slower to market. Start with Glass and EEGs: we'll get there in our lifetime using grow-and-prune, and our AIs will grow up with mentors they can respect.