X-Message-Number: 15211
Date: Sat, 30 Dec 2000 06:57:53 -0500
From: Thomas Donaldson <>
Subject: more comments re. Mike Perry's comments

Hi everyone!Here I am again, on what Mike Perry said. But before that I
willcongratulate those who saw some important sense in Platt'smessage, even if
(like me) they disagreed with it. In particular,Alcor should be much more open
about its problems in suspendingpeople. Such research may only be of interest
to a subset ofcryonicists, but it remains important because our conquest ofall
the many things that can go wrong in a suspension is basicallypart of how we
try to improve our suspensions ... so that YOU, toocan get a better suspension
than anyone before you.As for Mike Perry, he seems to completely ignore the
point thatfor us NOW, the possibility that someday, through advances
insoftware only, we can make a version of ourselves that willwork as fast as
we do. That version, if done in PRESENT timerates, will necessarily HAVE to be
parallel. Not only that, butit will probably involve many "neurons" rather than
only someEnter command or <CR> for more !CompuServe Mailmuch faster
processors .... each "neuron" would have to workon its particular job. I will
add that if you read my messagecarefully, I did NOT claim that such a system
had to be biologicalin the present sense. However it still could not merely
implementany of the contemporary "neural nets" in current computer
science.None of these allow the increase and decrease in number of neuronsor
the changes in connectivity that a human brain allows.Even worse, for such a
machine to work like us it must also havethe proper environment. If it is to be
much slower, than we mustsomehow imitate the universe of which it is aware ...
sometingwhich changes constantly. The job of making a system whichimitates us
and is SEQUENTIAL, rather than PARALLEL, will alsohave to imitate the entire
universe. Incidentally, such a systemwill cause problems for imitation by a
Turing computer: even ifwe make a "human" that works sequentially, the universe
itselfwill not provide suitable sequential input for it. Nor could itoutput
its results sequentially.Enter command or <CR> for more !CompuServe
MailTuring did some fine theory, and it still remains as fine theory.The
question is whether or not we HAVE to deal with parallelmachines (my answer is
YES), and machines with all the otherattributes of brains (yes, again)... if we
really wish to imitatea human being. I am not discussing the abstract question
ofwhether someday the universe will have slowed down enough thata single
sequential machine could imitate it, I am discussingour PRESENT situation. As
for that situation, it's not evenclear to me that a suitable machine with
neurons possessingthe features needed could really be imitated by a Turing
machine,even if we forget the timing issue (new processors and changesin
connection cause a problem, regardless). If we rightly INSISTon the timing
issue, then Turing machines clearly don't work,because they fail to imitate
that timing.And yes, it would be interesting here to consider not a
singleTuring machine, but a collection (importantly, one which can increaseand
decrease) of connected individual Turing machines. Given thatwe must constantly
deal with TIMING in the real world, such aEnter command or <CR> for more
!CompuServe Mailsystem may give interesting results.                Best

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=15211