Re: poly: Anthropic => Aliens like us?

From: Nick Bostrom <>
Date: Sat Mar 13 1999 - 14:35:33 PST

Robin Hanson wrote:

> Apparently various fundamental physics parameters look
> suspiciously tuned to allow life like us. For
> example, certain nuclear resonances and cross sections
> look tuned to allow stars to produce lots of Carbon.
> An anthropic analysis would suggest that these parameters
> are different in different parts of a "multiverse", and
> it is no coincidence that we, Carbon based life, live
> in a place in the multiverse tuned for Carbon.
> It occurs to me that this could be taken as evidence
> *against* non-carbon based life where our physics
> parameters apply. After all, if most generic untuned
> places were friendly to life, then we shouldn't
> expect to be in a place tuned for our kind of life.

If most life weren't carbon based, then one could well wonder, Why
are we carbon based? You could seek to apply an anthropic argument to
show that this would be improbable.

However, we should bear two things in mind. First, if all the other
(non-carbon based) life were divided into many different kinds
-- silicon based, neutron-matter based, gas-cloud based etc. -- then
every outcome would in a sense be equally peculiar. The
neutron-matter life could say "Hey, if there were a lot of
non-neutron matter life, then we would be very atypical...". But
reflecting on the matter, these guys could say: "...but wait: if we
were carbon based, then we would be equally surprised about *that*,
so maybe we shouldn't jump to any conclusions. It's like a lottery,
where any number is unlikely to win, yet some number has to win.
Learning what number won doesn't in the generic case give you reason
to be suspicious that the lottery was fixed in advance."

So for the reasoning to work, your a priori probability distribution
has to favour something like "Either there is (only) life of type A,
or there is life of types A, B, C, D, ... ." rather than "Either
there is (only) life of type A, or there is (only) life of type B, or
..., or there is life of types A, B, C, D, ... ." For with the latter
distribution, finding that you are of a particular type (say B) would
not enable you to conclude that there (probably) is life of only
type B.

Second, there is the question of the reference class. What class of
entities should we regard ourselves as random samples from? Maybe it
only includes carbon based life (though I doubt that), in which case
we can't infer anything from the fact that we are carbon based.

> And if most places are unfriendly to life we should
> be very suprised to see our place happen to be tuned
> for two different kinds of life.

Unless the tuning-conditions for these two kinds of life are
identical. For example, if the expansion speed of the early universe
had been slightly greater, then nothing denser than very low
density helium clouds would have formed, and hence there would have
been neither carbon based life, nor neutron matter based life. In
most universes, the early expansion speed might be wrong for both
these kinds of life, but it is not clear that we should be surprised
that we find ourselves in a universe where the expansion speed is
favourable to both.

> A similar argument would apply to any other way in which
> the universe appears tuned for us. If this place is
> tuned to make planets, for example, then we shouldn't
> expect to find life here that didn't start on a planet.

Again, I'm not sure this would apply to life that started from an
environment that (by the laws of nature) will be present in same
universes as will planets. (And a lot of alternative life forms that
have been speculated about seems to require approximately identical
parameter values as carbon based life. .)

> This strikes me as the sort of argument that may have
> been already made. Barrow & Tipler make it?

I can't think of any passage where they do, off the top of my head.

Nick Bostrom
Department of Philosophy, Logic and Scientific Method
London School of Economics
Received on Sat Mar 13 14:39:22 1999

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