X-Message-Number: 15048
Date: Tue, 28 Nov 2000 20:49:25 -0800
From: Lee Corbin <>
Subject: Quantum Mechanics vs. Uploading

I have never believed that arguments about QM were at all
relevant to uploading, any more than questions about the
Many-Worlds interpretation are.  Probably these were first
invoked for illustration, or as part of a reductio-ad-
absurdum.  The two main concerns about uploading are (1)
can a program imitate a human being satisfactorily? and
(2) if it did, would it necessarily be conscious, or have
feelings, or have an "inner life", etc?  None of this has
anything to do, in my opinion, with quantum mechanics.

Robert Ettinger writes,

>Now the uploaders have painted themselves into another corner.
>When a computer describes or calculates the "quantum state" of
>a physical system, my understanding is that this means it states
>or calculates the value of each coordinate in the phase space
>of the system, as precisely as the uncertainty principle or the
>Bekenstein Bound allows. For a particle in a one-dimensional flat
>well this means it specifies x and p (position and momentum) at
>time t. x1 includes a short space interval, p1 includes a small
>momentum dispersion, and t1 includes a short time interval. 
>The computer then goes on to calculate the "next" (t2) values
>of x and p, etc.

For all I know, if the goal for some crazy uploader was to actually
calculate a _quantum state_ IN DETAIL, then perhaps you have shown
this to be problematic.  For as you write, 

>But this is not quantum reality! It is just selecting, out of the
>infinite possibilities, the one "most probable" succeeding state.
>For the computer to reflect quantum reality (as presently understood

>by most physicists) or to reflect Many Worlds, it would have to
>calculate ALL the possible successive "states" and would therefore
>effectively grind to a halt immediately.

To go around calculating quantum states is very ambitious!  I'm 
real glad that for all the computer programs that I've been asked
to write over the years, all they ever wanted was for the computer
to do stuff.

:-)  Imagine that the year is 1927, and that a city is planning on
replacing its venerable old cop who has been directing traffic at
Broadway and Main for decades.  Someone at the city council meeting
says, "Look, it's simply impossible to have a mechanical instrument
replace Officer O'Malley!  Even if it were possible classically, the
new wave mechanics shows that in order to calculate the next quantum
state, e.g., "blow-whistle", the calculations must be done for at
least 10^100 possible worlds!"

So at the risk of being unfair and changing the grounds of the 
argument, let me proceed to the actual reality of what is wanted.
First, by no means can a realistic program that replaces me do
EXACTLY what I would have done in every case, because even I cannot
do EXACTLY what I would have done if.... Robert Ettinger had not 
written a certain email!  Or if the phone rings!  Everything that
happens  throws me into an entirely new reality if anyone is foolish
enough to be really picky about it.  The mathematician E. Borel used
to find it very interesting and significant to perform classical
computations that show that displacing a gram of matter in the star
Sirius by a mere centimeter will change the state of a liter of gas
on Earth within a second (after the gravitational influence arrives
eight years later, of course).

Second, there is an incredibly broad range of successful "Lee Corbin
imitation behavior" that a program might implement.  I won't go into
the very interesting and complex question, quite relevant to our own
cryonic revivification scenarios, of how we know that we've got back
the same person or not.  But I will remark that even in the cases of
severe brain surgery, we are able to tell, even though the patient's
behavior is very greatly affected.

Last, do programs really "calculate" their next state?  We have possibly
an abuse of language here.  A program goes from one state to another
every time it rearranges some data or loads some data from the outside.
In the ultimate example, a TM writes a 1 or a 0, or moves and goes into
a different "state" (i.e., technically into a "state" defined as an
abstract quadruple or quintuple).  Does a neuron calculate its next state
when it fires or not?  I don't think that we want to talk this way, and
we probably don't want to say that an artificial neuron necessarily
calculates anything either.  I hope I'm not going out on a limb here,
but what say we reserve "calculation" for an actual arithmetic
computation?  It's possible that unconscious associations of this
term have led to some misapprehensions.

>Note carefully that the Turing Machine ITSELF is CLASSICAL, even
>though it can calculate quantum mechanics.
>
>So--once more--what do we have? The Turing Machine does not and
>CANNOT emulate a person, because a real person does not always
>evolve into the next most probable configuration. Sometimes his
>next configuration is less probable.  Therefore the emulation is
>guaranteed to be different from what real life would be. (At
>least "the" real life, as opposed to "a" real life.) 

As said before, "the" real life is a fiction anyway (since my
phone may ring at any minute!)  If it's the strict determinism
that is true of a TM that's bothersome, then of course TM's can
also simulate pseudo-random outcomes.  But more importantly, the
tape of an actual TM that was simulating a person would be as
extraordinarily rich and unpredictable as our own world is.  Its
behavior would, in effect, be driven by vast numbers of random
inputs---like someone trying to get its attention---just as ours is.

Lee Corbin

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=15048