On 17 Apr 2012, at 11:23, Robin Faichney wrote:

Hi Bruno,

This is very interesting for me, my approach to information is via the
mind-body and "hard" problems, and I'm sympathetic to
computationalism. On the other hand, I have difficulties understanding
much of what you say here. Let me focus on one point for now though.

Tuesday, April 17, 2012, 8:48:48 AM, Bruno wrote:

Let me sketch the reasoning shortly. If I can survive with a
digitalizable brain, then I am duplicable. For example I could, in
principle, be "read and cut" in Helsinki (say) and pasted in two
different places, like Moscow and Washington (to fix the thing).

The subject to such a duplication experiment, knowing the protocol
in advance, is unable to predict in advance where he will *feel to
be* after the duplication. We can iterate such process and prove
that at such iteration the candidate, seeing if he feels to be in W
or in M, receive a bit of information, and that his best way to
predict his experience, will be, in this case, to predict a random
experience (even algorithmic random experience): like WWMWWWMMMWM
...., for example. That is the first person indeterminacy.

It seems to me that, if I believe I am duplicable, and understand the
protocol, I must predict that I will experience being in both Moscow
and Washington. The process bifurcates one person, who becomes two
people with absolutely identical physique and memories immediately
afterwards, which will then begin to diverge. Both, looking back to
pre-bifurcation times, will say "that was me", and both will be
correct. There is no "essence" to be randomly (or non-randomly)
assigned to one location and not the other. The individual is now two
people and therefore can be and is in both cities.

You are right. That is why in more lengthy explanations I introduce the key distinction between the first person points of view, and the third person points of view. For the simple case of self-duplication the first person account can be defined by the content of a personal diary brought by the candidate: which means that it will be read and cut itself and reconstituted in both cities. Here the third person account is provided by the memory, or the content of a diary of an external observers, and is not going to be annihilated and reconstituted.

The indeterminacy bears on the first person experience(s). You are read and cut in Helsinki, and reconstituted in W and M. You are asked to predict where you will *feel* to be after the duplication.

We suppose that you assume comp, and so you believe that you will survive such duplication. We assume also that the correct level of substitution has been chosen, and all the default assumptions (no bugs, no asteroids demolishing a reconstitution machine, etc).

So you can predict, with this relative to comp certainty, that you will survive, and that indeed, you will survive in both city. But comp prevents the existence of some telepathy between the two copies. So you know in advance that both of "you" (the M-you and the W-you) will both *feel* to be in only one city.

One will write in his diary: "Oh, I see that I am in Washington and I can only intellectually believe that I have a doppelganger in Moscow", and the other will write "Oh, I see that I am in Moscow, and I can only intellectually believe that I have a doppelganger in Washington".

So from both their first personal point of view, they get one bit of information.

The indeterminacy comes from the fact that they cannot have predicted in advance which one they will become, before the duplication.

If, in Helsinki the candidate predicts that he will feel to be in W, the guy in M will rightly consider that he was wrong in his prediction made before in Helsinki, and vice versa. As you say yourself, bot copies shares their memory-life up to the experience in Helsinki, and to evaluate their uncertainty, they have to take into account all the future copies discourses.

If this is not clear, let me copy you a slightly more elaborate account. It is called "step 3", because it is the step 3 of the Universal Dovetailer Argument, which is the main argument showing the reversal between physics and information/computer science following the comp assumption. It takes into account preview critics similar to what you said. In fact you attribute your consciousness and identity to both copies (it is a 3-person-view on the 1-person-view, but the indeterminacy is on the 1-view themselves). You can say "tomorrow, you can join me in W and in M". But from your personal perspective you get after the duplication, you will feel to be one and entire in a well defined city (W *or* M).

If you get this, don't read what follows. If it is still unclear, read what follows, and if some questions remain, I will answer them here.

== UDA step 3 == (from the FOAR mailing list) == ( 

I assume comp throughout, and, like in the two preceding steps, I assume the supplementary "neurophilosophical hypothesis", according to which the biological brain, the one in our skull, is the generalized brain (= the portion of reality that we have to emulate digitally, at some right description level, for your consciousness to be able to relatively manifest itself.
The description, or substitution, level exists by the comp hypothesis.

So you have accepted, or are supposed to have accepted (tell me otherwise) that you can survive with a digital brain (this is just a rephrasing of the comp hypothesis, using the neuro hypothesis).

You have accepted, in UDA step one, a job on Mars, for which you need to use a digital classical teleportation device, twice a day. So you are read and cut on Earth and then, from the read (scanned) information which is sent on Mars, you (your body) is reconstituted there, at the right substitution level. Then you do your work on Mars, and at the end of the afternoon, you come back on Earth, using the same method.

No need to imagine that this can be done in practice. The reasoning needs only to be possible in principle, or in theory. Actually, a large part of UDA can be done with robots or programs instead of humans, given that we have given purely third person definition of the first person (personal memory) and third person (bodies encoding those personal memories) notions. This means also we can derive a lot of information by using the weaker "strong AI" thesis, or even the still weaker Behavioral Mechanist thesis in the place of comp, but it is more tedious to do so, and I continue to use comp.

Steps 1 (and 2), can also be considered as a clarification of the meaning of comp, and does not introduce anything really new.

Step 3 is a bit more involved.
Like in step two, but more importantly, the key relies in the understanding of the first and third person distinction.

Here is the protocol of step 3.

You (your body) are (is) read and cut (annihilated) in Helsinki (say). But now the scanned digital information is sent (by radio-wave, say) in two places: Washington and Moscow (to fix the things). We assume that he is not reconstituted elsewhere.

In Helsinki, you are told the entire protocol in advance. So you know that you will be read and cut in Helsinki, and that "you" , that is the body, in the relevant state (which exists by comp) will be reconstituted in both cities W and M.

Now we ask the following question to the candidate, in Helsinki. What do you predict will happen to you, from the experiential, first person, point of view. We ask "where will you feel to be?"

Let us imagine how the candidate can answer that question.

He might be a bit naive, and answer directly something like

"I will find myself in W and in M".

Well, that is not entirely incorrect. We can indeed join him, or a version of him, in both cities. But then, we remind him/her that the question, asked in Helsinki, before the split, was bearing about what he will feel personally. His answer was bearing on him, as seen by some external observer. So he agrees that his first answer was a bit ambiguous, and he makes it more precise by saying:

"I will feel myself in W, and I will feel myself  in M".

That is, again, not entirely incorrect, but still does not answer the question. Here the candidate imagines the existence of the two reconstituted bodies, made in each city, and he attributes them consciousness and feeling, and the right, provided by comp, to assert that they are "him" (they both do have the genuine memory of the person who was in Helsinki).

Now, if only by definition, the personal first person account consists in what will be written in the personal diary/memory. He will never write "I feel myself now being in M and I feel myself now in W", because the two bodies have become independent, and there are no reason, (in our theretical comp frame) to expect some telepathy between the two reconstituted persons. So, he can know in advance in Helsinki (relatively to the comp assumption), that in each copy of the diary, which he will take with him in the read and cut box in Helsinki, and which will be duplicated and reconstituted in W and M together with him, both the one in M and the one in W will have to write something like:

I know myself to be in W (resp. M) and I believe intellectually that there is another guy in M (resp. W), looking like me, and having the same right as me to claim that he is the same person as the one in Helsinki, despite I am not him, in the sense that right now, our personal feeling are different.

Put more simply, he will write "I am in W and I have a doppelganger in M", or "I am in M and I have a doppelganger in W".

Note that he cannot have any personal certainty that the doppelganger has already been constituted. From there, he can only develop third person belief about him.

The guy know all this in advance. He knows that if comp is true, he will survive the duplication, and that, in all possible future personal situation, he will feel to be in only one city, with an inferred doppelganger in the other city.

So, if he is asked in Helsinki where he will feel to be, he can only answer that he will feel to be in W or in M, but without being able to be sure if he will feel to be in W or that he will feel to be in M.

Indeed, if he write in his diary that he is sure that he will feel himself in to be in W, the copy in M, will have to admit that he was wrong when in Helsinki (which belongs to his memory). And, by comp, he is right about this. And vice versa, if he writes in Helsinki in his diary that he is sure that he will be in M, then the copy in W will have to admit he was wrong, when in Helsinki, as the question in Helsinki bears on such personal future self-localization feeling.

So we (and him) can conclude that in Helsinki, the candidate cannot predict with certainty what will the result of his self-localization be experienced specifically, in the personal way, except for the indetermination: Washington OR Moscow.


I hope I was not too long. I did take into account some critics recently made.

Garry, if you are OK with this, feel free to write a shorter version. I will certainly propose shorter version myself in some summary later, but this version takes critics which have been already done into account.

Any problem? Some variants of that experience can be given. The impossibility of predicting exactly what we can feel in digital self- duplication experiences is what I call "first person indeterminacy".

Like in Everett QM, the splitting, or differentiating, person cannot feel the split (as he would not been able to immediately distinguish this duplication experience with the one where we cheat on him, and reconstituted him only in one city).

Feel free to make any comment, or ask any question, if you have any difficulty with this. In some next post, we will have to say some words about how to quantify such indeterminacy.

You might test your understanding by considering iteration, 64 times say, of that experience, where both the W-guy and the M-guy come back (by planes, say) to Helsinki, and redo that experience, and then again and again.

Abbreviating "I feel myself to be in X" by "X", the 2^64 resulting diaries will all contain a "memory/story" describable by a string of length 64, and simple combinatory relations can show that most of them will be random (even algorithmically incompressible, and more and more so, relatively to the number of strings, as the number of iteration grows).



Robin Faichney


fis mailing list

Reply via email to