X-Message-Number: 15003
Date: Thu, 23 Nov 2000 12:05:13 EST
Subject: Serial uploading


A few of the uploaders, e.g. Mike Perry and Lee Corbin, understand my 
objections but have more-or-less plausible responses. However, I think they 
draw an arbitrary and ultimately indefensible line between permissible and 
impermissible uploading scenarios. 

For example, Lee says, as I read him, that isomorphism alone is not enough; 
there must also be ongoing activity in real time (even though scaling is 
allowed) and causal relationships in the succession of quantum states of the 
system (sentient being). However, I still think that the isomorphism 
postulate leads to reductio ad absurdum, even if at first we tentatively 
allow it.

We remind ourselves, first, that a minimum postulate of uploading is that any 
instantiation of a Turing computer could support sentience, and could emulate 
(as nearly as desired) any particular person and a segment of his history. 
(If we stipulate the computer as the ur-computer, the Turing tape, that is 
enough right there to turn off many people, but let's assume you can swallow 
that--even though, for reasons I have outlined previously, this situation is 
worse than might at first appear.)

Let us also sweep under the rug all the questions about time and its 
continuity or lack thereof, and the discreteness or fuzzy discreteness of 
quantum states etc.  

An irreducible postulate of the uploaders, as noted, is that "you" could be 
represented as a succession of sets of symbolic states in a computer--for 
example, marks on a paper tape, changing from time to time in a discontinuous 
manner. In other words, by hypothesis, a labeled subset of states of the 
computer corresponds to a succession of states of the person emulated. 

Another reminder: I don't raise any "symbols-not-semantics" objection, 
allowing the uploaders their contention that, at least in some cases, a set 
of symbols can have sufficient internal consistencies and implications so 
that only one interpretation is possible.

Another reminder: I am NOT talking about the problem of duplicates--that is a 
separate question. The problem addressed right now is just whether an 
uploaded system could be a person at all. (I also ignore here the question of 
unique physical requirements for feeling or subjective states.)

Now. If uploading is valid at all, it is valid not only from meat person to 
computer, but also (and more easily) from one computer to another. In the 
extreme, we could have a separate computer for every successive quantum state.

Let this sink in a little. If "your" successive quantum states were realized 
not in a single computer, but in a succession of computers, one per state, 
then--by the uploaders' thesis--"you" would exist and live just as surely as 
in one computer or in the meat.

But what is a "computer" in this context? Lee says it is the whole system, 
including its functioning and causal relationships over time (even though 
most of its functions do not reflect biology, but only mathematical steps 
leading eventually to approximations of representations of biological 
states). It seems to me he is implying that the "computer" must have a 
"purpose" in order to count as a computer, and therefore in order for the 
representation to count as a person.

I don't think so. In principle--as far as we know, according to majority 
opinion--there is a very small probability that "you" (or a Boeing 747) could 
spontaneously self-assemble from raw material in the environment, according 
to the laws of q.m. and thermodynamics. If that happened, surely the result 
would be a person, just as much as the more usual kind. 

So "accidental" persons SHOULD count. A meat person arising conventionally or 
accidentally or by some unusual artifice will equally have sentience, and the 
same for a computer person. This implies, it seems to me, that if a 
representation counts as a momentary (one chronon?) state of a person, then 
any accidental configuration, if it can be interpreted as a correspondence, 
should be considered by an uploader to be a person.

Need the interpretations or correspondences be the same? No. Surely in one 
computer, for example, numbers could be represented in binary, in another in 
decimal, etc.--yet the emulated person is the "same" when uploaded from one 
to the other. Therefore we arrive at the conclusion rejected by Lee and Mike 
and others--that if you "mine the data" and look hard enough at almost any 
substantial region of space and time, you will find configurations of matter 
(subsets) that could be interpreted as representing persons. 

This seems to be another type of reductio ad absurdum--either for the 
uploading thesis or for the universe.

Robert Ettinger
Cryonics Institute
Immortalist Society

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=15003