From: Hal Finney (hal@rain.org)
Date: Thu Nov 14 1996 - 15:19:00 MST
Robin Hanson, in his very interesting "Great Filter" essay
about our prospects for expanding into the universe <URL:
http://hss.caltech.edu/~hanson/greatfilter.html >, mentions in passing
an argument which I have heard only a little about:
> Note that this cause for concern has a different basis than the simple
> statistical arguments of Gott [Gott 93] and Leslie [Leslie 96] that all
> else equal we shouldn't expect many more future humans than there have
> been past humans.
I wonder if this argument makes sense to people. I apologize for discussing
it without having read the original but hopefully I have the gist of it
right. The idea is that, given a person chosen at random, chances are that
he will be average in most ways. In particular, he will be about average
in terms of when he lives in the grand scope of human history.
If we graph human population over time, and assume for simplicity that the
resulting total human population is finite, then the average person's
time of living can be found by multiplying each era by the population of
that era, adding these up, and dividing by the total human population.
In other words, the average person will live at a time when the total
number of people who have lived before him will be about the same as the
total number who have lived after him.
Now we take the step of taking ourselves as our randomly chosen people.
Applying the argument, we conclude that approximately as many people will
live after us as have lived before us. Because of the recent increase
in population, it follows that there can't be many years left for humanity
because most of the people who have ever lived are alive now. Our time
seems to be about up.
Maybe I'm not telling it right, but this argument seems quite weak to me.
First, I don't accept the notion that each person can view himself as a
randomly chosen individual. There has been no selection procedure such
as we normally think of in random choice. Rather, each person throughout
history has his own conciousness. And in fact I seem to be not at all
random even compared to people in my own time. I live in one of the richest
countries in the world; even among such people I have attended some of the
finest schools in the country; my interests and abilities are extremely
unusual.
Looking at it in Bayesian terms, I am going to live out my lifetime here,
this year, whether the human race self destructs in a few years or not.
Both outcomes are equally consistent with my life experience. So I don't
see how the knowledge that I am alive should cause me to change my
estimate of the likelihood of these two outcomes.
Even if we accept this probability argument, there is then the question
of what the sampling pool is. Is the argument consistent with a transition
to a post-human stage in the near future? Could we say that the argument
validly predicted the end of the human race, but that the ending is actually
a transition to another stage? Just how much of a transition is necessary
for the argument to retain its force? I don't see any way to answer these
questions. If there are no answers then I think the argument is invalid.
I actually have another idea which answers the argument on its own terms.
Think in terms of the Moravec/Tipler picture where humanity lives
forever but spends a lot of time simulating the past. Then Gott's
argument would suggest that the "average" person should be from the
era most simulated.
What is the most interesting period to simulate? Why, the Transition,
of course: those brief millenia when we went from being paleolithic
cavemen to posthuman gods. So naturally they'd look at that over and
over again. And our era is probably high on the list, the first age when
we began to realize what is happening. Even among our era, those people
most simulated would be those who became aware of the great changes in
store for them, such as Extropians. And the moments which would be most
simulated would be when they were actually thinking and communicating
these ideas... just as we are now.
Hal
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:35:50 MST