CV works! 'Moral Perturbation Theory'

From: Marc Geddes (
Date: Thu Jun 17 2004 - 21:47:53 MDT

'Moral perturbation theory'

After thinking about the problem of extrapolated
volition I now don't think it would be so hard after
all. What made me really skeptical at first was the
fact that there would be a huge number of different
components contributing to the 'Collective Volition'.
But then I realized that not all these components
would be of equal importance.

In astronomy, within the solar system there are a
large number of bodies whizzing around. Aside from
the planets there are various asteroids and so forth.
Now when astronomers calculate the motions of a planet
they don't try to make an exact calculation. There
are just too many different gravitational influences,
and an exact calculation would be enormously complex.
So astronomers use something known as 'Perturbation
theory'. What makes the calculations tractable is the
fact that not all the gravitational influences are
equal. All they have to do is look at the biggest
gravitational influence for a first approximation.
Then they say that the actual motion of the planet in
question is this approximation plus a 'perturbation'
representing all the other gravitational influences.
The accuracy can be increased to any desired degree by
successively factoring in the influence of each other
body in the solar system in order of decreasing
gravitational influence.

To approximate the collective volition of humanity we
don't need to look at the brains of 6 billion people
at all. In fact I think that the brains of only a few
hundred should suffice as a very good approximation!
Further if we wanted to calculate the approximate
extrapolated volition of any particular individual,
we'd simply take the CV as a generic first order
approximation , then add the 'moral perturbations'
caused by the more specific mental characteristics of
the individual under consideration.

Here's my reasoning: Remember the extrapolated
volition is defined to be the volition that would
result if one thought longer, faster, knew more , was
more wise etc. In other words: the volition resulting
from an increased over-all cognitive capacity. So a
sampling of the people in the world with the highest
cognitive capacity would be a good generic first order
approximation of a typical humans 'extrapolated
volition'. We only need look at our bright and
brightest.. the Gandhi’s, Einstein’s etc.

Consider: as the extrapolations are projected forward
in time, small initial differences get magnified. The
principle is similar to two different objects
following slightly different trajectories under
gravitational influence. Initially there isn’t much
difference in the location, but as we extrapolate
forward the objects would get further and further
apart. By analogy the psyches of just a few hundred
of humanities ‘best and brightest’ people would ‘blow
up’ in influence relative to everyone else on the
planet as the extrapolation projected forward. In
fact this small group of people would end up exerting
the major 'moral influence' on the collective volition
in the long-term. So we can safely say that the
collective volition of humanity is largely represented
by only these few hundred people plus a smaller
'perturbation'. The smaller 'perturbation' would be
the moral influence of the rest of the people in the
world. Remember the astronomy analogy. To increase
the accuracy of the first order approximation, we
would simply expand the pool of people under
consideration, in order of decreasing cognitive
ability. The 'moral perturbation' on the
approximation would become smaller and smaller as more
people were considered.

Here's the really neat bit: To extrapolate the
volition of any particular typical human with
approximate accuracy we don't even need much detailed
information about that individual. We simply take the
generic first order approximation that is Collective
Volition, and say that the volition of the specific
individual we wish to approximate is simply the CV
plus a smaller 'perturbation' caused by the mental
characteristics specific to that particular
individual. To increase the accuracy of the
extrapolation for that particular individual we don't
need to look at all their mental characteristics, we
just start by factoring in the ones exerting the most
'moral influence' on their psyche. As more mental
characteristics are considered, the 'moral
perturbation' caused by the remaining characteristics
would be growing smaller and smaller aka the astronomy
analogy. A database of 'human mental characteristics'
would be consulted, moving from a small number of
generic characteristics exerting large 'moral
influence' on the psyche, to a larger number of more
specific characteristics exerting smaller 'moral
perturbations' on the earlier approximations.
Obviously the further these calculations were carried,
the more accurate the volitional extrapolation. But
the point is that using this 'moral perturbation
theory' only a very few mental characteristics should
be sufficient to produce reasonable approximations of
volition even on the individual level!

I also now think that the ‘sentient humane FAI’
approach would actually produce results more or less
equivalent to Eliezer's ‘non-sentient collective
volition process’ approach.

Here’s my reasoning for this: Suppose we created a
sentient FAI based on a generic altruistic human with
transpersonal philosophy. Blowing this seed up to the
super-human level would be more or less equivalent to
a 'first order approximation' of the Collective
Volition of humanity (because as I explained, one only
needs to sample the minds of a small number of
humanities best and brightest (i.e. a seed A.I based
on a generic competent well balanced human altruist),
then 'extrapolate the volition' of this small sampling
(i.e. make the seed A.I 'think faster, longer' etc as
in 'hard takeoff).

Imagine that the sentient humane FAI implemented a
global political system whereby 'rights' would were
assigned according to cognitive ability, as I
suggested. So the FAI (which would actually be
equivalent to the 'first order approximation' of
humanities Collective Volition remember) would have
the most say in what went on in the world (because
most of the cognitive power on Earth would reside with
it). Humans (having less cognitive ability than the
FAI) would initially have fewer rights than the FAI,
but they would still have some say. This is
equivalent to factoring in the 'moral perturbations'
caused by more specific non generic characteristics of
the human psyche to increase the accuracy of the
original 'first order approximation' of humanities CV
(represented by the altruistic FAI).

So if the first FAI was a sentient FAI based on a
generic altruistic human with transpersonal
philosophy, it could exactly duplicate the results of
the non-sentient CV FAI system simply by setting up
the global Meritocracy system I proposed (giving
rights to each sentient on Earth based on that
sentient's cognitive capacities and including itself
as one of the sentients).

"Live Free or Die, Death is not the Worst of Evils."
                                      - Gen. John Stark

"The Universe...or nothing!"

Please visit my web-sites.

Science-Fiction and Fantasy:
Science, A.I, Maths : - Yahoo! Personals
New people, new possibilities. FREE for a limited time.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:47 MDT