X-Message-Number: 8402
From: Ralph Merkle <>
Subject: Estimating the probability of success of cryonics
Date: Sat, 19 Jul 1997 14:19:23 PDT

I tried to post this a few days ago, but it seems not to have made it.
-----------------------------

Periodically, someone posts an analysis of the form

Probability of success =

    probability of event 1 x
    probability of event 2 x
        .
        .
        .
    probability of event n

While this approach is entirely reasonable in the abstract, it
is only valid when the different events are independent.  Usually,
however, the events listed are correlated in various ways, resulting
in a serious error in the overall estimate.

For example, if two events are:

      government allows revival
and
      not put in zoo

we see that there is a likely correlation.  If, for example, after the
successful revival of people in cryonic suspension the Supreme Court
rules that they have the same rights they had prior to suspension,
then the government would be more likely to block attempts to put
you in a zoo.

Similarly, the items

      will awake without dire pain
and
      my cryonics organization is willing to revive me

are also correlated.  If there is a significant probability of waking
in dire pain, the cryonics organization will likely conclude that
further research is required and that those in suspension should not be
revived until the problem of dire pain is solved.

My experience is that lists with more than three or four items suffer
from severe correlations, which invalidates the simple multiplicative
estimate of the total probability of success.

In general, cryonics will fail only if:

1)  Information theoretic death occurs at some point in time
2)  Technologies that are feasible in principle are never applied
    in practice.

In addition, it is theoretically possible that cryonics could be a
technical success but would still be undesired if, on waking, you
found that being dead was preferable to being alive.  We could
therefore add a third item:

3)  I don't like the result.

Even this short list suffers from correlations.  For example,
unsatisfactory results (item 3) could be caused by failure to
apply appropriate methods (item 2).

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8402