X-Message-Number: 9489
From: (John P. Pietrzak)
Newsgroups: sci.cryonics
Subject: Probability (warning, long post!) (was Re: Darwin-Preliminary)
Date: Wed, 15 Apr 1998 05:18:40 GMT
Message-ID: <>
References: <>
<>
On 13 Apr 1998 21:34:36 GMT, (Ettinger) wrote:
>Mr. Pietrzak denies, in effect, that there is any way of estimating
>probabilities of future advances in various areas such as nanotech or others
>relating to revival of cryonics patients.
>
>Please see"Cryonics: Probability of Rescue" on our web site,
>http://www.cryonics.org
I've recently received, and have (briefly) studied a paper copy of
Cryonics: The Probability of Rescue" (thank you, Mr. Ettinger). I'm
not certain I have everything correct here, but let me try to sum up
and see if you agree with my summary below.
(Note for others reading this: it's really unfair to judge a paper
simply from another person's summary, particularly one who might not
know what he's talking about, so if you're interested in it please
refer to the source above to get all the details.)
(Note 2: I'll cross-post this to cryonet, as I suspect it may be of
some interest there as well.)
(Note 3: If you've already read Mr. Ettinger's paper, you can skip to
the end of this message to find my complaints about it. My paraphrase
of the paper is probably too long, but does give context for my
comments.)
-------------------------------------------------------------------
[Paraphrasing the paper, section by section:]
Introduction: This paper is intended to describe the "correct"
manner in which probability theory should be applied in judging the
probability of success in cryonics. Scientists may have a harder time
with the paper than laymen. (?)
I. Foundations of Probability Theory
A. Approaches taken to probability
1. von Mises -- a "frequency" theory of probability
A well known (?) form of probability, it deals with mapping a
problem onto an idealized infinite sequence of (identical)
experiments, and counts the probability as the frequency of occurance
of the event in this sequence. This approach is vague in application.
2. Doob -- another "frequency" theory of probability
This approach provides a somewhat less vague definition of the
bounds on the sequence of experiments, but is still not appropriate in
application.
3. Laplace -- "classical" probability (also frequency based)
This approach involves defining probability as the ratio m/n,
where n is the number of outcomes of an experiment (all being equally
likely and mutually exclusive), and m is the number of these which are
"favorable". Easier to apply.
4. Koopman -- A very different form of probability
This approach, arguing that frequency theory is inadequate,
attempts to bring intuition directly into the theory. It adds axioms
called "Laws of Thought" to the system.
B. Synthesis -- (of Koopman vs. the frequency-based probabilities)
We attempt here to extend the frequency theory of probability
to include "single events" and "subjective" probability. A
description of the process for applying von Mises is given, noting:
1) the limiting frequency can never be noted exactly
2) experiments will not be perfectly identical
3) isolated events are considered outside the theory, unless
an appropriate series of events can be simulated.
4) probability only refers to an experiment in the context of
a sequence.
Example: bets placed upon football game between teams W and M.
Bettor A: Chooses team M from A.P. Poll. A.P. Poll has
history of 80% success, so probability M will win is 80%.
Bettor B: Chooses M randomly (the guy has never even heard of
football before.) Probability M will win is 50%.
Bettor C: A coach, rating M four touchdowns better than W, and
having history of 95% correct choice of winner when given a
four-touchdown difference, chooses M. Probability M will win
is 95%.
The differences being the experience of the bettors.
[Another example is given, which I omit for space]
We underscore the idea that the relevance of information is more
important than it's precision.
Conclusion: "Thus, we treat frequency probability and personal
probability on a single objective basis, such that we can assign a
definite (although imprecise) numerical probability to any event of
whatever kind."
C. Some Probability Notions Clarified
[Most of these were unclear to me, so I'll not confuse the
issue by trying to summarize them here.]
II. Applications
Examples of problems with probability as currently used.
Example of medicine man bringing rain on day when low chance of rain
given by weather service, thus validated. Lacks a priori probability.
Example of ESP, again lacking a priori probability as well as poor
experimental controls. Quote: "A rough guess, properly used, is much
more useful than an exact datum inappropriately applied."
A. A suggested Estimate for the Exponential Life Parameter
An application of the above, originally applied to estimating
the mean life of vacuum tubes (in 1953). Bit hard to encode in ASCII,
but mainly is composed of comparing equation (2) giving theta-hat sub
F (without a priori knowledge) to equation (5) giving theta-hat sub B
(using a priori knowledge). In situations where samples are few, the
a priori factor dominates and produces "better" results.
B. Aspects of Probability Concerning Cryonics
We will now apply probability to the success of cryonics.
First, we will ignore the risk of global war or catastrophe, business
risks, and political risks. The focus will be on the scientific
chances.
C. Intuition and Probability
Examples:
a) people intuitively avoid cars when crossing the street
b) Goddard, Tsiolkovsky, etc. predicted moon rockets. Lacking
any problems with feasability, they had the intuition that
the political and monetary petty details would work out.
c) Leonardo da Vinci intuitively invented flying machines;
given, for example, birds, the scaling up to flying
machines was only details.
Many experts fail to intuitively accept cryonics due to the time
involved -- it will not succeed within their lifetimes.
D. Unlimited Wealth -- Not "Probable" but Certain
[I'll skip this part, I don't really care about it and it's
assumptions are even less defensible than those in the previous
section.]
E. Summing Up: "Probability" Derives from "Experience"
[Let me quote a portion of the summary verbatim:]
" 1. In the modern era, not a single goal of science, so far as I
know, has been shown impossible (although some have proven more
difficult than expected, and others become irrelevant). Odds-on for
success.
2. Many cells survive even uncontrolled freezing, and there have been
partial successes with freezing mammalian brains: odds-on that, even
in freeze-damaged brains, injury is limited and reparable.
3. The Precedent Principle, the Feinberg Principle, and the prospect
of nanotechnology assure us that the atom-by-atom manipulation of
tissue (frozen or not) will allow construction or reconstruction, in
finest detail, of any human configuration known, designable, or
capable of inference. [...] Odds-on that _you_ can be restored.
In still other words, we conclude the "gamble" of cryonics is odds-on
in your favor: the probability of success (from the standpoint of
technical feasability) is much closer to one (certainty) than to zero.
The number may be imprecise, but it is the best and most scientific
estimate available."
----------------------------------------------------------------
I guess I became disappointed in this paper by the time I got to the
Koopman section in the discussion of approaches to probability, becase
I knew it was going to delve into regions of which I don't approve.
In short, Mr. Ettinger's argument seems to follow this line:
1) Intuition is an unavoidable part of probability, and can be an
asset when included explicitly into the theory. (Part I of the
paper.)
2) A priori knowledge is valuable in determining the true probability
of an event, using the "Classical" Bayesian approach. This is
particulary true when few samples are used, as the a priori knowledge
can (theoretically) make up for the lack of information. (The first
part of part II goes over this, including the mathematical example.)
3) Mr. Ettinger provides some (dubious, in my mind) examples of
intuition which have, over time, been proven correct (in the Intuition
and Probability section). He's essentially trying to give a large
probability to the success of intuition; trying to show intuition as
the coach who can forecast success with a 95% probability.
4) Here's the magic part: we now use our intuitive knowledge of
cryonics, and apply it (a priori) to a probability function which has
(guess what) _no_ samples at all! Hmm, all we've got left, then, is
our a priori knowledge, so that becomes our probability. Take a look
again above at the material I quoted at the end of my summary of his
paper. Those three points, despite what you or I may think of them
subjectively, are not only the only inputs Mr. Ettinger provides for
the a priori knowledge, they are also (through the magic of Classical
Bayesian manipulation) the only possible outputs of his probabilistic
study.
Unless I've missed my mark completely, this twenty page thesis comes
just about as close to circular reasoning in it's conculsion as I've
ever seen. This is _always_ a problem when you bring (unconstrained!)
subjective reasoning into a theory, and the a priori element of the
classical Bayesian system seems pretty well general license to do
whatever you wish when the number of samples is small. The
combination of the two is deadly. :) Basically, I just can't agree
that he has in any manner reached a "scientific" estimate.
John
Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=9489