X-Message-Number: 2266
From:  (Thomas Donaldson)
Subject: Re: cryonics: #2258-#2263
Date: Tue, 18 May 93 10:25:11 PDT

About the proposed resuscitation scenario:

1. The newly created version of the patient should NOT be created in the frozen
   state. Thawing itself will cause fatal damage, even assuming that we can
   project from the patient's state to what that patient's state would be if
   frozen in a perfect condition. This point has been made before.
2. It will take more than AI alone to recover the patient from a map of the
   location of his/her molecules. We still need to understand a good deal more
   about just what the state of a live person is, chemically. We don't yet
   understand how memory works, nor IN DETAIL where it is stored. Therefore
   no amount of AI can tell us how to put it back there. Since such storage
   and even the detailed anatomy of the patient's nervous system will very
   likely be unique to that patient, simple comparisons with others won't
   help.

   Finally, one major problem still remains after these: since we don't under-
   stand memory, we have no firm idea whether it even still exists in the
   frozen patient. I myself, and I believe most cryonicists, have joined not
   because we feel that survival of memory is certain, but because some 
   acquaintance with the LIKELY forms of storage has convinced us that memory
   will PROBABLY survive --- not that it will DEFINITELY survive :-( sorry!

3. The problem of understanding how we work, of course, is a problem in 
   biology. The same is true about survival of memory. Preserving location of
   every molecule gives far too much extraneous information, since we are
   not so much the location of molecules as an ever-changing system within 
   which our molecular parts are created and destroyed constantly. We have

   not summarized the state of a river by listing the position of its molecules.
   A human being, or a very intelligent computer, would need a good deal more
      information before he could describe that river simply on the basis of its
   molecules; that information could not be derived simply from the list. The
   reason is that from the list alone it's impossible to decide whether some
   concatenation of molecules is random or essential to the riverness of the
   river. And if an AI system DID understand rivers, it would not look for
   position of molecules but for the higher level features which characterize
   rivers. Those features (of which for human beings we still lack a complete
   understanding, or one sufficient enough to recreate an individual) which
   the AI system would want are these higher level ones.

   I'm not saying that AI systems  couldn't do this because they were AI. That
   is a red herring. I'm describing what a repair system of any kind should
   look for, and what it needs to find it. The "river", for a frozen patient,
   is damaged. To do any repair the AI system must first examine many cases
   of undamaged rivers to work out just what is a river, and what an undamaged
   one looks like. It would then be able to seek out exactly the information
   it needs --- which is not in the positions of molecules but in SOME of their
   relationships to one another (and not precisely, either: a subset of their
   possible relationships to one another).

   Incidentally, verifying that one of a subset of relationships exists is
   a much harder computational problem than verifying position of molecules.
   It looks hard enough to me that the computational problem alone would
   make such a method of revival fail.
			Best and long life,
			 	Thomas Donaldson

PS: About the last paragraph, an addendum. The problem I see is a combinatorial
explosion: we aren't looking for individual molecules but for a relationship.
If we have a list of molecular positions, then we must test all candidate 
subsets for this relationship --- and the number of subsets of a set is 2^N,
where N is the number of elements of the set. In the case of a human brain,
that number is very large ...

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=2266