"Personal Identity and Uploading", by Mark Walkerpost by gwern · 2012-01-07T19:55:00.726Z · LW · GW · Legacy · 19 comments
1 Uploading: prospects and perils 2 The equivalency thesis 3 Personal identity: psychological and somatic accounts Against somaticism: the big stroke Against somaticism: retrospective replicas Against somaticism: practical ethics 4 No branching 5 Types and tokens 6 The type/token solution to personal identity 7 Should I upload? 8 Further reading None 19 comments
Objections to uploading may be parsed into substrate issues, dealing with the computer platform of upload and personal identity. This paper argues that the personal identity issues of uploading are no more or less challenging than those of bodily transfer often discussed in the philosophical literature. It is argued that what is important in personal identity involves both token and type identity. While uploading does not preserve token identity, it does save type identity; and even qua token, one may have good reason to think that the preservation of the type is worth the cost.
If, like me, you think that uploading is possible (at least in principle), and so you hold that the first interpretation of these events is correct, then you must hold true the following three theses:
1. Computers are capable of supporting the important properties constitutive of personal identity, e.g., thought and consciousness.
It is clear that uploading will not preserve all properties we associate with Homo sapiens, e.g., basic facts about the human digestive system are not likely to be preserved in uploading to a robotic body. But these facts are not typically thought to be important for personal identity. Candidates for important properties include thought, consciousness, emotions, creativity, aesthetic experience, sensory experience, empathy and so on. For the most part, the question of which properties are important is not as serious as it may first seem, since uploading promises to preserve the essential aspects of the brain and nervous system, which overlap with the usual lists of important properties for identity.
…suffice it to say that if Searle is correct, then #1 may be false. For Searle thinks that a computer can never consciously think merely in virtue of instantiating a computer program, and the uploading process seems to be one of merely instantiating a computer program (Agar 2010, 2011).
2. It is possible to capture the information necessary to emulate the important properties of individual humans.
…If we slice off layers of your neurons, and record the information of each layer, the lower layers will change (due to trauma or death).1 If we flash freeze your brain, we may destroy some essential information. Philosophical questions arise as to whether the information encoded in the brain is sufficient to account for all the relevant properties. For example, consider a dualist who believes that we have souls in addition to brains, and much of what is morally important (e.g., conscious thought) resides in the soul. If the dualist is right, then scanning your brain could never be sufficient, for it would be necessary to scan your soul to access at least some of what is important. If it is unlikely that we will be able to scan souls, there will be an insurmountable obstacle to uploading. Notice how theses #1 and #2 may differ on this point: a dualist could consistently hold that a computer might have a soul; it is just that if computers have souls, it is not because we obtained the soul-building information from humans. (Perhaps God implants souls in humans and computers.)
3. It is possible to survive the uploading process.
To see how #3 differs from #1 and #2, imagine that at some point in the future we have created computers of sufficient complexity that it is agreed that they have the same morally relevant properties as humans: these advanced computers think and are conscious, they are accorded rights, and the scanning problem has been solved so that we are able to scan the brain in such a way that we are not worried about loss of information. None of this answers the question of whether you have been preserved during uploading or whether uploading merely makes a very good copy of you…If an individual can be uploaded once, then it seems the same individual could be uploaded twice into separate computers, and indeed, billions of the same individual all embodied in separate robotic bodies could be created.
To make the discussion manageable, I will focus on thesis #3, and assume without argument that #1 and #2 have been resolved in favor of uploading. So our question is this: assuming that computers can be conscious, have memories, and (robotic) bodies, and assuming that it is possible to scan and capture all the information of a human brain, does uploading preserve personal identity? I will argue that uploading does preserve personal identity; at least identity of a certain sort.
The fact that we are assuming that computers are capable of embodying all the same type of properties necessary for personal identity means that we can make use of the equivalency thesis:
Equivalency thesis: If it is possible for an individual to survive migration from a carbon to a carbon body, then it is possible for individuals to survive migration from a carbon to a silicon body.
…There are a couple of reasons for invoking the equivalency thesis. The first is so that we are not misled by a new form of racism: substratism (Walker 2006). Substratism is the view that one’s substrate is inherently superior to that of other substrates along the lines that racists think their race is inherently superior to some other race. In the present case, it would suggest the idea that carbon-based humans are inherently more morally worthy than silicon based beings. Consider the fact that we would not accept this argument: it is not possible for persons to migrate from one body to another because then it would be possible for people of skin color X to move to bodies of skin color Y, and Y skin color is morally inferior. We want to avoid the same bad argument in considering moving from one substrate to another. Notice that this does not beg the issue at hand, since it is possible to say that having a certain substrate (or even skin color) is constitutive of my identity; it merely prohibits saying that this property in itself makes for moral superiority.
The second is that it makes directly relevant an enormous amount of philosophical effort that has gone into exploring the possibility of carbon-to-carbon transfers. The question of carbon-to-silicon transfers thus may piggyback on this effort.
The thesis disturbed me the first time I saw it; it seemed to me that it either begged the basic philosophical question at point or it did not do any work. So I read on to see how it was used. It seems to be the latter case: the thesis is barely used and not really germane to the examples that criticize somaticism and argue for a type-token kind of personal identity. This is good because it seems like used in any kind of strong sense, it’s easy to criticize the thesis.
(Implicitly, it seems to scope over all individuals - that we could rewrite it as, ‘for all individuals that survive any carbon->carbon transition, there is a carbon->silicon transition they survive’. But this seems false: a book is made out of carbon, survives minute to minute or copy to copy, and can be satisfactorily uploaded, but can a squishy human brain? Can a bowl of water? If I take a stick of carbon and light it on fire, how do I upload the burning stick? What does an uploaded diamond do? One might say the physics of the constituent atoms can be uploaded and this is a correct emulation with any necessary properties like emergence, but then we’re back to the question-begging.)
Historically, there are two main schools of thought about what is required for personal survival; the psychological and somatic approaches (Olson 2002).
…the psychological account says that what is essential for survival is continuity of psychological states such as memory, beliefs, desires and personality. John Locke, an early proponent of this view, famously described personal identity in terms of psychological continuity, within an analysis of personhood as consisting in existence as “a thinking intelligent being, that has reason and reflection, and can consider itself as itself, the same thinking thing, in different times and places…” (Locke 1975). The person on Mars who awakens will claim to remember being Derek Parfit, and to have memories and a personality that are psychologically indistinguishable from the person on earth whose body was destroyed. Locke, then, would say that Parfit survived teletransportation.
Somaticist accounts suggest the survival of a particular body is critical for personal identity over time. Since the body on Earth is destroyed during the scanning process, Parfit ceases to be. A different person will awake on Mars. This person will of course have psychologically indistinguishable memories and personality to those of the late Parfit, but this person will not be Parfit. The new person will be but an infant in terms of chronological age: only a few minutes old. We will think of “somaticism” as the view that continuity of one’s body is necessary for personal identity from one time to the next.2 There are two ways that one might be a somaticist: one can believe that bodily continuity is necessary but not sufficient, or that it is necessary and sufficient. One easy case to distinguish these two is as follows: a piano falls on your head, and causes you to go into a permanent vegetative state. Your relatives discuss whether to “pull the plug.” Those who think that bodily continuity is necessary but not sufficient may say that you no longer exist, but your body continues to exist. Those who think that bodily continuity is necessary and sufficient will say that you continue to exist, albeit your cognitive capacities are non-existent. Both views qualify as “somaticism” in our sense.
The Vorlons,5 a mysterious and intellectually advanced alien species, make this offer: you can have an original undiscovered play by Shakespeare written in his hand, or a copy of the play made by one of his lackeys. You salivate at the joy this will bring to the world (not to mention the fame and fortune it will bring you personally). Since you can have only one, the choice, it seems, is a no-brainer. You should opt for the one written by the bard’s hand. But now consider this variant: the Vorlons tell you that the text written in Shakespeare’s hand is missing the last two pages, while they assure you the copy written by the lackey is a perfectly faithful reproduction of all the words in the original. While it would be great to have both, you reason that the most important thing is the play itself be preserved, not Shakespeare’s handwriting. The copy here is in some sense better than the original because the original has been damaged.
…The Vorlons, with their ability to see into the future, say the news is grim. In less than twelve hours you will have a massive stroke that will cause you to lose many of your memories and some mobility, and impair your intelligence. Your stroke will not be as bad as some: the damage from the stroke will not leave you completely cognitively impaired, but you will no longer be able to work as an academic. You will have to find some relatively mindless job befitting your new level of intelligence, perhaps in academic administration…there is nothing the Vorlons can do to prevent the stroke. They provide a radical alternative: creating a perfect replica of you – down to the molecular level – with the exception that the problems with the arteries to your brain will be fixed in the body replica. They insist, however, that only one body can survive. You must choose tonight whether the replica or your current body survives.
…Obviously, this example is structurally similar to Parfit’s [suicidal teletransporter], but with one big exception: what is gained by having the replica survive is much more significant in this case than in the Teletransporter to Mars case. Parfit offers the incentive of avoiding three weeks of space travel. (We might not even sacrifice our original wedding ring for an exact replica if the benefit is merely avoiding three weeks in a spaceship). Here the incentive is the possibility of not having one’s life radically altered by the stroke. The attachment to one’s body does not seem worth the cost in this case.
…It is worth noting that not all somaticists are likely to be convinced by this example.6 But it should convince a few, and points out one of the heavy costs of somaticism.
…perhaps it is this fear of the unknown, rather than a commitment to somaticism, that explains the reluctance to use the Teletransporter. We can test this thought by considering a retrospective rather than a prospective version of a replication scenario.
Suppose that every night when people sleep their bodies (including their brains) are scanned by a swarm of nanobots and a molecule for molecule identical body is beamed from a hidden alien spaceship in orbit; the old body is vaporized in a manner that is undetectable by the human eye. Scientists discovered this fortuitously: physicists noticed a spike in neutrino levels every time psychologists in the adjoining lab conducted sleep experiments. Intrigued, scientists built a chamber to isolate subjects from neutrino influences and then had test subjects sleep in the chamber. Once the experiment was initiated, a hologram of a Vorlon appeared in the lab and spoke thusly:
We are an ancient race known as the “Vorlons.” We battled another species, the “Shadows,” just as your species was beginning to evolve on this planet. One of the toxic effects of our war was a type of radiation that kills all higher intelligences within three days. We have no way of eliminating the radiation, but we have left advanced technology to recreate your bodies from different molecules every day so that the radiation will not harm you. We left the galaxy eons ago. You are hearing this message now because you have advanced technologically to the point where you can detect our technology. If you interfere with our replicator technology, you will quickly die of radiation poisoning.
What should we make of this? It is clear that dismantling it is out of the question since all humans will die within three days. If you are a somaticist, you must conclude that you have been alive only for a very short while. In fact, you have existed only since last night. After all, the physical continuity of one’s body has lasted only this length of time. However, most of us, I think, would conclude the opposite. That is, that we have existed for years: that we do not cease to exist every night and a new person comes into being.
…the somaticist must now explain how so many people could be mistaken about their own identity retrospective case; after all, it seems very likely that, upon learning about the Vorlons’ technology, most would conduct their lives as if they hadn’t just come into existence that day. Who is going to say such things as: “I do not have to look after these children you call mine: how can I have children if I myself was born today. I can’t use this driver’s license, it is someone else’s – I was just born today. I’m not qualified to teach any classes: a postgraduate degree is required, which takes years to earn, and I was just born today?”
This example reminds me strongly of Nick Bostrom’s reversal test for the status quo bias; an example would be a drug that increases IQ 10 points may be feared and rejected, but would it be accepted if scientists discovered new pollution will reduce IQ 10 points and that drug would compensate? I like his reversal test, and I like this example as well.
…To emphasize, let us suppose that despite the fact that the preponderance of reasons seem to be against somaticism, imagine that the metaphysical reasons for and against somaticism are exactly balanced. Does this imply that we should be neutral on the issue? I think not. It may be that there is a further court of appeal to decide the issue, specifically, practical ethics. That is, the suggestion is that if our metaphysical arguments and intuitions cannot decide the metaphysical issue of personal identity, it is permissible to decide the issue on non-metaphysical grounds.
Imagine two persons, McCoy and Hatfield, who want to kill one another. They are co-inventors of the first replicating machine. McCoy thinks it should be used on humans, Hatfield believes that it never should be so employed. McCoy believes in the psychological continuity thesis of personal identity, whereas Hatfield believes in somaticism. How should we reason about personal identity in terms of what is good for society? We can imagine two possibilities: society adopts for, social and legal purposes (its “public norm” for short), somaticism or psychological continuity. Which is better for society?
Consider first using somaticism as the public norm. McCoy could kill Hatfield and then hop in the replicating machine. We would be forced to say, because we have adopted somaticism as our public norm, that McCoy is dead and the replica of McCoy (call this person “McCoyson”) is a different person. Since McCoyson was born after the crime, McCoyson cannot be responsible for the crime. (We have long abandoned the idea that one can inherit personal responsibility for the sins of one’s ancestors). This crime would be ruled a murder-suicide in a somaticist jurisdiction. Of course McCoy then has every reason to commit the crime, as he does not believe in somaticism.
…If psychological continuity is the public norm, then neither Hatfield nor McCoy will have reason to commit the crime based on replication. As before, Hatfield will not because he will consider this equivalent to suicide. McCoy will not because the public norm says that McCoy will survive the replication and be subject to criminal sanctions. Since a public norm of somaticism is more likely to lead to negative social consequences, this gives us some reason to reject somaticism.
Debates about identity preservation and uploading invariably get hung up on the “branching” problem, and this probably provides the strongest support for somaticism. The problem is that it seems there is only one of me. But uploading seems to allow the possibility that there could be hundreds, if not millions, of “me.”
Using the equivalence thesis we can see how this is exactly the same problem as the problem of branching that philosophers discuss in connection with carbon-to-carbon transfers. Parfit extends his Teletransporter case in exactly this way:
Several years pass, during which I am often Teletransported. I am now back in the cubicle, ready for another trip to Mars. But this time, when I press the green button, I do not lose consciousness. There is a whirring sound, then silence. I do not lose consciousness. I leave the cubicle, and say to the attendant: “It’s not working. What did I do wrong?” “It’s working,” he replies, handing me a printed card. This reads: “The New Scanner records your blueprint without destroying your brain and your body. We hope that you will welcome the opportunities which this technical advance offers.” (Parfit 1987, 199)
Of course there is no reason to stop at one replica. Using Parfit’s Teletransporter thousands of organic molecule-for-molecule identical persons could awaken in the same instant, all claiming to be Mark Walker.
…What does the psychological account have to say about multiple replicas? Here opinions differ. On the one hand, it seems that if there are multiple replicas, and they are all psychologically indistinguishable from the original, then each of them has as good a claim to be me, and so they are all me. The contrary “no-branching” view is that at most one replica is me, for there can be only one me (Shorter 1962).
The question then is whether there can be “branching”: more than one of me. I will argue that both sides of the debate are correct; there is a sense in which there can’t be more than one of me, and a sense in which there can be multiple versions of me.
The No-branching Argument:
- P1: Multiple replicas X, Y, Z…. of an individual O (the original) are numerically non-identical with each other, that is, X is not identical with Y or Z, Y is not identical with X or Z, and so on.
- P2: Preservation of personal identity requires preservation of numerical identity.
- C: Therefore, not all replicas X, Y, Z… preserve personal identity of O.7
…I want to suggest that the problem with the no-branching argument is that there is a critical ambiguity. To explain the ambiguity it will be helpful to review the type/token distinction.
(For those not familiar with the literature, ‘numerical’ here is being used in a sense of complete identity - there being complete logical equivalence. So for example, everyone reading this is numerically identical with themselves, and numerically not identical with the pope.)
Walker invokes the type-token distinction:
There are twenty tokens of the word the, but a single type of the word the. The argument to be canvassed is that if we think of personal identity as ambiguous between types and tokens, then the no-branching argument may be rejected.
for 2 false anti-replication personal-identity arguments:
No-branching Token Argument
- P1’: Multiple replicas X, Y, Z…. of an individual O (the original Hamlet penned in Shakespeare’s hand) are numerically not (token) identical with each other, that is, X is not (token) identical with Y or Z, Y is not (token) identical with X or Z, and so on.
- P2’: Preservation of play-identity requires preservation of (token) numerical identity.
- C’: Therefore, not all replicas X, Y, Z… preserve play-identity of O.
No-branching Type Argument
- P1’’: Multiple replicas X, Y, Z…. of an individual O (the original Hamlet penned in Shakespeare’s hand) are numerically not (type) identical with each other, that is, X is not (type) identical with Y or Z, Y is not (type) identical with X or Z, and so on.
- P2’’: Preservation of play-identity requires preservation of (type) numerical identity.
- C”: Therefore, not all replicas X, Y, Z… preserve play-identity of O.
The reader can guess what comes next: he’ll make the move of saying personal identity is the ‘type’ and any upload or copy is the ‘token’. We accept that while the original Hamlet is valuable in many respects, Hamlet survives the destruction of the original if an appropriately faithful copy is made.
The ontological status of abstract entities is a perplexing and contested issue (Wetzel 2009), but there is no reason to think that it is more perplexing in the case of persons rather than literature, and we are committed to types in the case of literature.11
The No-branching Argument in terms of Tokens
- P1’’’: Multiple replicas X, Y, Z…. of an individual O (the original) are numerically [token] non-identical with each other.
- P2’’’: Preservation of personal identity requires preservation of numerical [token] identity.
- C’”: Therefore, not all replicas X, Y, Z… preserve personal identity of O.
There are two problems with this argument. First, it is question begging. The entire issue is whether personal identity can be explained in terms of preservation of type identity, and so P2’’’ prejudges the issue.12
The other problem is that it is difficult to see how one can insist on non-branching without collapsing into somaticism. To see this, consider the case where the original Mark Walker’s body, O, is destroyed when three replicas X, Y, and Z are created. Either O is not identical with any of X, Y, Z, or O is identical with one of X, Y, Z. If the former, then non-branching is simply somaticism in disguise. If it is asserted that O is identical with exactly one of X, Y, Z, then any choice would be arbitrary in the sense that choosing one among the thousand to be The Mark Walker would not be choosing based on any intrinsic differences. We could, for example, have all the replicas draw a number out of a hat and designate the winner of the lottery The Mark Walker. But an appeal to a lottery shows that precisely no intrinsic properties are used to individuate: it is the process (the lottery) that does the individuating. We could do the same for Hamlet. We could assign a number to every extant copy of Hamlet and have a lottery to find out which is The Hamlet, and which are mere copies. But, of course, no one would be impressed by this.13
Points in favor of the type-token:
- explains reluctance to sacrifice a rare token for small gain in Mars case (destructive)
- explains why people would sacrifice rare token for large gains, in first Vorlon case
- explains our lack of concern, in second Vorlon case
- explains precedence of Earth original over Mars copy, in second Mars case (non-destructive)
- not too paradoxical in cases of multiple copies and no surviving original
Still, it may look as though this is tantamount to an argument against uploading: if there is any loss in uploading, even if it is only token identity, why would anyone want to sacrifice some identity?
(This hearkens back to a previous JET paper I covered, “Ray Kurzweil and Uploading: Just Say No!”, Nick Agar. Agar is not cited for this part of the paper.)
The answer is that there are considerable advantages (or at least purported advantages) to being uploaded, including immortality and enhancement. Except for the completely reckless, forgetful or lazy (ahem), everyone backs up his or her valuable computer files. But once we see that people too can be backed-up, it appears that virtual immortality is assured. For so long as there are operating computers, one can simply transfer the files that comprise oneself from computer to computer. If the hardware on one computer fails, you simply move to another computer…As for enhancement, one possibility is that our senses could be radically enhanced: robots presently make use of a sensory apparatus that detects light in parts of the spectrum not available to (unaided) human vision (e.g., infrared, x-rays, etc.), sounds that are beyond normal human auditory range, and so on. In terms of enhancing cognition consider that it is a relatively routine matter to add memory or computing power to today’s computers. If one is uploaded to a computer, then it seems that it would be a relatively routine matter to enhance one’s memory or cognition: just add more computer memory or processing power. The sky is literally the limit here
…It is beyond the scope of this paper to argue that these purported benefits of uploading really are benefits, but, if they are, the temptation to upload is clear. And just like in the stroke case, it is clear why it might be rational to forgo token identity survival for these advantages.
There doesn’t seem to be any discussion of this paper online. My own views on personal identity tend to the psychological pattern, which does not seem to be very different from a type-token theory of personal identity, if there is any meaningful difference at all, so this was a less challenging paper to read than the others, the equivalency thesis aside. The examples may be worth remembering.
Comments sorted by top scores.