poly: What probabilities?

From: Nick Bostrom <bostrom@ndirect.co.uk>
Date: Thu Apr 23 1998 - 18:53:02 PDT

When we discuss future possibilities, we say that somethings are
probable and others are not. Sometimes we assign numerical
probabilities to hypothesis such as (H) "Nanotech will come before
superintelligence." How are we to understand such probability

I know some people who think that many hypotheses we discuss are at
the moment so widely out of reach for the scientific method that it
doesn't make sense assign probabilities to them. I don't think this
is the right attitude. But I do think that there are some things that
need to clarified before such probability assigments will be very
informative. In particular, I think we have to distinguish between
several different senses of probability.

When I say "The probability of H is 40%." I could mean to:

1. Say something about my personal subjective probability. This
would be equivalent to saying: "Given the evidence that I have, the
rational betting odds for H would be 40%." Note that on this
interpretation, you could say "I think the of H is 1%.", and we would
not have disagreed with each other, and we could both be perfectly
rational (at least if we didn't know about each others' statements).

2. Say something about the common subjective probability relative to
the present knowledge basis of humankind: i.e. that the rational
betting odds, given by the sum total of present human knowledge is
40%. (This is ambigous: what human knowledge should be included? Only
general scientific knowledge, or also facts that are only known to
one person, such as the hidden intentions of some government
official?) (This is the sense of probability that would hopefully be
approximated by the actual betting odds in an real-money incarnation
of Robin's idea futures. Given such an institution (if it worked as
intended), the probability in sense 1 would tend to approach the
probability in sense 2.

3. Say something about the objective probability of the event. I
could, for example, mean that: if we were to create one million
planets almost exactly similar to present-day earth, then on 40%
of them there would be nanotech before superintelligence. An
ambiguity here is "almost exactly similar". It could mean:
(a) "exactly the same", (b) "same as far as the state of present-day
earth is determined by my present evidence", or (c) "same as far as
the present state of the earch is determined by the totality of
evidence available to humans today".

(A futher ambiguity for each of these interpretations is how much
computing power to assume that the rational estimator has.)

Depending on which sense of probability we have in mind, we might
give very different numerical estimates. Consider the question: "What
is the probability that nanotechnology will cause the extinction of
earth-descendent intelligent life?"

In sense 1, I might answer: "25%".

In sense 2, I might say: "I don't know, but I would guess somewhere
between 2% and 70%."

In sense 3, I might say: "I don't know but I think it more probable
that it is somewhere between 99.9% and 100% than that it is between
50% and 60%.

Nick Bostrom
Department of Philosophy, Logic and Scientific Method
London School of Economics
Received on Fri Apr 24 01:01:00 1998

This archive was generated by hypermail 2.1.8 : Tue Mar 07 2006 - 14:45:30 PST