Welcome to Scribd. Sign in or start your free trial to enjoy unlimited e-books, audiobooks & documents.Find out more
Download
Standard view
Full view
of .
Look up keyword
Like this
2Activity
0 of .
Results for:
No results containing your search query
P. 1
M&PI 15-MAI-91

M&PI 15-MAI-91

Ratings: (0)|Views: 70|Likes:
Published by Riccardo Mantelli

More info:

Published by: Riccardo Mantelli on Apr 01, 2008
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

05/08/2014

pdf

text

original

 
MECHANISM AND PERSONAL IDENTITY
Bruno MARCHALI.R.I.D.I.A. Université Libre de Bruxelles50 av. F. Roosevelt. CP194/6. B-1050 Brussels, Belgium
The soul is a number which moves itself. Xenocrate (see 44)
Abstract
: Some thought experiencesseem to refute the possibility of subjectiveexperience for machines. By using therecursion theorem of Kleene, I try toinvalidate these refutations. A newparadox occurs. I generalize an idea usedin the foundation of Quantum Mechanicsto suggest a step toward a solution.
*
Key words
: Machine, Recursion,Duplication, Modality, QuantumMechanics.
1. I
NTRODUCTION
I give an intuitive definition of a strongversion of Mechanist Philosophy (calledsimply Mechanism) and I present someparadoxical situations which look likerefutation of this philosophy. Then I willbe more precise about Machine andsketch a more rigorous "mechanist theoryof Identity" based on the Kleenerecursion theorem which throws somelight on these paradoxes. A new kind of paradox appears then. This paradox bearssome relationship to the problem of measurement in Quantum Mechanics, forwhich there already exists a mechanistsolution (see 12, 14, 18, 43). Then Isuggest that the present approachgeneralizes that solution with theconsequence that mechanist philosophywould fit both with "Strong A.I." andwith quantum mechanical facts.
 
*
The following text presents research results of the Belgian National incentive-program forfundamental research in artificial intelligenceinitiated by the Belgian State, Prime Minister'sOffice, Science Policy Programming. Thescientific responsibility is assumed by theauthor.
2. M
ECHANISM
The simplest intuitive strong mechanistaxiom is the following non metaphoricalclaim : "my brain is (at some level) acomputer". The Mechanist will agree thathe is not able to prove that a computercould vehiculate subjective experiencesbut he feels that he is not able to provethat for a human or even for himself. Sohis claim will rest on a semi-empiricalanalysis. The Mechanist does not believethat a neuron (or anything...) is able tocollect in one step an infinite amount of information from the neighbourhood sohe believes that there is a level at whichsome digital machines can replace anypart of his brain (including biochemicalregulatory pathways).Consequently the Mechanist will admitthat there is a level of organization of hisbrain such that some mechanicalsubstitution of all parts will preserve hisidentity (see 19, 20, 24, 28, 32).
3. T
HE PARADOX OF THEPOSTPONED DUPLICATION
Mechanism entails that we are duplicableentities. Let us call "translator" a machinewhich consists of a transmitter-annihilatorpart and a receiver-builder part. Themachine is supposed to be 100% reliable.An (intuitive) mechanist is someone whodoes not fear to use a translator as avehicle (although he does fear death).Suppose now that the transmitter-annihilator part is in Brussels and that thereceiver-builder part is in Washington.The Mechanist trusts the vehicle and thismeans that he believes in Brussels that theprobability of finding himself inWashington after using that vehicle, isone. What will happen if there are tworeceiver-builder, one in Washington and
 
one in Moscow ? It is not difficult, at anintuitive (although Platonist) level toshow (by iterating the experience) thatmechanism entails that the probability is1/2. Moreover, a strict subjectiveindeterminism (inside OR) appears herealthough from the outside point of viewthe situation is determinist (outside AND)(see 24). I call that form of indeterminism: mechanical indeterminism.The original has not privileged status. Soif a mechanist uses the translator with asingle receiver-builder in Washington anda transmitter
without annihilator 
inBrussels, the probability of finding itself in Washington, or staying in Brussels, isstill 1/2. Now the instantaneous statedescription read by the transmitter can beput on a magnetic tape -or a crystal-,which permits postponing the building of the copy. Let us look at the followingargument given by Mister X.Let us call t
1
the instant of capturing mydescription in the crystal. Supposet
0
< t
1
< t
2
, I know, said Mister X, thatin the interval (t
0
t
1
) I can postpone in theinterval (t
1
t
2
) the decision between usingthe crystal to build a copy of myself or todestroy that crystal. But this entails that Iknow during (t
0
t
1
), that I will be the onewho will take that decision in (t
1
t
2
), so Iknow that the probability to remain theoriginal (the one who will take thedecision) is 1. (and this reasoning can beused to discourage any use of a translatoras a vehicle). The copy will appear to belike me but I will know a posteriori that Iwill not be him. Note that all thereasoning here can be done in the interval(t
0
t
1
)and that makes the argument apriori. Suppose now that at t
2
he decidesto build the copy. Mechanism entails, atleast, that the copy will say : "Oh, myGod! ...I was wrong", and thereconstituted copy will also realize that hewill not be able to convince his older self (due to the fact that he knows that he hasbeen convinced by the a posteriori -> apriori argument). Who is right ? Anywaya mechanist (like the reconstituted copy)will have to admit that if there is nobackward causation the probability willdepend on his self knowledge and inparticular on his ability to respect his owndecision.
 Here is another question
: what is theprobability in (t
0
t
1
) of staying alive at t3with t
3
> t
2
if the reconstituted copy isdestroyed in the interval (t
2
t
3
) ? It seemsthat the probability of being destroyed is1/2 although the probability of stayingalive at t
3
is 1. There are a lot of translator-like paradoxes possible (see24). The following paradox can be seenas a kind of limit of such thoughtexperiences.
4. T
HE PARADOX OF THE FILMEDTWO
-
DIMENSIONAL COMPUTER
There exist two-dimensional computer(see 11). Suppose that the low levelmodules of that bidimensional machineact and react by luminous messages. Themachine is embedded in semi-opaquewhite smoke between two panes of glass.The sensibility of the modules is such thatthe presence of light suffices to triggerthem. So it is possible to activate amodule from the outside by lightprojection. Mechanism entails that it ispossible to compile for such a machine aninstantaneous state description of a brainof a dreaming person. Mechanism entailsthat the evolution of the correspondingprocess on that two-dimensional machinewill vehiculate the dreaming experience.Now we can film the evolution of thatmachine. The question is : does the filmvehiculate a subjective experience ?Because there is no more mechanicalcausation in the film it would seemfoolish to expect, even from a mechanistphilosopher, an affirmative answer to thatquestion. The trouble is that, relatively tothe initial context which here is just theinstantaneous state description D (that iswhy I talk about a dream), the behaviourof the film is always locally equivalent tothe behaviour of the two-dimensionalcomputer. It suffices to project the filmin real time
and real space
on it, it beingreset at the instantaneous description D.We can remove one, two, three anypieces (modules) of the two-dimensionalcomputer without changing anything. Soit seems that mechanism entails that thesimple projection of the film generates theexperience for that situation is reducible
 
to the emulation of the two dimensionalmachines when all parts of the machinehave been removed. The trouble is that,for the same reason, we can delete anyparts of the movie and, of course, theentire film itself. Does
nothing
emulatedreams ? Because any decision to choosea frontier between what can and whatcannot vehiculate the experience, duringthe two finite removing processes (of themachine's parts and of the movie's parts)seems to be arbitrary, the reasoning lookslike a refutation of Mechanism by areductio ad absurdum. Nevertheless, Iargue that such a refutation is not valid(more details are given in 24).
5. I
DENTITY
The identity theory sketched here is basedon the second recursion theorem of Kleene and its formal (and Platonistic)version known as the diagonalisationlemma (see 1). The embedding of thesubject in the object I want to performdepends on the closure of the set of intuitively computable functions (ICF) forthe diagonalisation operations.Paradoxes of self-reference aretransformed into infinite processes. I willalways implicitly use Church's thesis (see22). The code of a program will beconsidered here as a necessary bodywhich permits the program to manifestitself relatively to an universalenvironment. The theorem of Kleenepermits us to write programs which areable to output an intuitively computabletransformation of their own code.Informally, the idea is the following : it isnot difficult to write a program P which,given as input the code X of aprogram X, outputs the result of atransformation T applied to the code of aprogram which compute X(
 
X ), (firstdiagonalisation). So P(
 
X ) outputsT(
 
X(
 
X ) ). P has a body P . Theresult of the application of P on its owncode P . : (second diagonalisation) :P(
 
P ) outputs : T(
 
P(
 
P )). Forexample, if T is the identicaltransformation, P(
 
P ) will be a self-reproducing program. It is a programwhich builds a copy of itself like a manwho uses a translator. The method isconstructive. In LISP (for instance) itgives the following self-reproducing
expression
:
((LAMBDA (X) (LIST X (LIST(QUOTE QUOTE) X))) (QUOTE(LAMBDA (X) (LIST X (LIST(QUOTE QUOTE) X))))
.Some mathematician (see 13 page 227)argue that such expressions are not trulyself-referencing because a mathematicienis needed to interpret them. But here weknow that a universal LISP program cando the work. The LISP expression abovedoes correctly reproduce itself relativelyto a LISP interpreter as a amoeba doescorrectly replicate itself relatively tonatural law. I will insist on Kleene 'stheorem by giving a little less informalproof which will give me the opportunityto introduce useful notations. ICF arecharacterized by the fact that you candefine them finitely with finite languages.It is thus possible to enumerate the set of the ICF :
φ
0
,
φ
1
,
φ
2
,
φ
3
, You canidentify the indice i of 
φ
i
, with theprogram which computes
φ
i
. Thefundamental properties of such sequencesare :
u
i
x
φ
u(i,x) =
φ
i(x) (1)
u is a universal program where the code uis able to emulate
φ
i
, and :
s
i
x
y
 
φ
i(x,
y
) =
φφ
s(i,x)(
y
). (2)
So parametrization can be automated,with program s. Now Kleene's theoremis (not in his most general form) :
t
e
y
φ
e(
y
) =
φ
t(e,
y
)
e computes the transformation with code ton itself.
y
denotes a n-uple of parameters. The proof is the same asabove :
λ 
x
y
.
φ
t
(
φ
s
(x,x),
y
) (firstdiagonalisation) is certainly an ICF withvariable x,
y
. So there is a r such that :
φ
r(x,
y
) =
φ
t(
φ
s(x,x),
y
).

Activity (2)

You've already reviewed this. Edit your review.
1 thousand reads
1 hundred reads

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->