|
But if artificial evolution is to become a powerhouse of creativity
on par with natural evolution, we must either grant it immense time
periods we don't have, or enhance it with further creative aspects of
natural evolution, if they are indeed there. At the very least, messing
with artificial evolution will illuminate the true character of
historical evolution of life on Earth in a way that neither current
observations nor past fossils can hope to do.
I do not find it alarming at all that evolution theory may be taken over
by postdarwinians without biology degrees. The great lesson which
artificial evolution has already imparted is that evolution is not a
biological process. It is a technological, mathematical, informational,
and biological process rolled into one. It could almost be said it is a
law of physics, a principle that reigns over all created multitudes,
whether they have genes or not.
The least-appreciated aspect of Darwin's natural selection is how
unavoidable it is. The conditions for natural selection are very
specific, but if these conditions are met, natural selection is
inevitable!
Natural selection can only occur in populations and swarms of things.
It's a phenomenon of mobs distributed in space and time. The process
must involve a population having (1) variation among individuals in some
trait, (2) where that trait makes some difference in fertility,
fecundity, or survival ability, and (3) where that trait is transmitted
in some fashion from parents to their offspring. If those conditions
exist, natural selection will happen as inevitably as seven follows six,
or heads and tails split. As evolution theorist John Endler says,
"Natural selection probably should not be called a biological law. It
proceeds not for biological reasons, but from the laws of
probability."
But natural selection is not evolution, nor can evolution be equated
with natural selection. In the same way, arithmetic is not mathematics
nor can mathematics be equated with arithmetic. One can claim that all
of mathematics is just addition compounded. Subtraction is addition in
reverse, multiplication addition in sequence, and all complex functions
built upon those mere extrapolation of addition. This is somewhat the
same argument of the neodarwinists: all evolution is the extrapolation
of natural selection compounded. While there is a grain of truth in this
perspective, it shuts off understanding and appreciation of more complex
things. While multiplication is precisely a form of serial additions,
wholly new powers emerge from this shortcut that would not be understood
if multiplication was only thought of as addition repeated. Dwelling on
addition will not get you to E=mc2.
I believe there is a mathematics of life. Natural selection may be its
additive function. But to fully explain the origin of life, the
remarkable trend toward complexity, and the invention of intelligence
requires more than addition. It needs a rich mathematics of complex
functions built upon each other; it needs deeper evolution. Natural
selection alone is not enough, not by miles. It must be alloyed with
more creative, generative processes to accomplish much. It must have
more to naturally select from.
What the postdarwinians have shown is that there is no such thing as
monolithic evolution run by one-dimensional natural selection. It would
be more fitting to say that evolution is plural and deep. Deep evolution
is an aggregate of many kinds of evolutions; it is a multifaced god, a
creator with many arms, working by many methods, of which natural
selection of variation is perhaps the most universal factor. An
uncharted variety of evolutions make up deep evolution, just as our
minds comprise a society of dimwitted agents and a variety of types of
thinking. Various evolutions proceed at different scales, at different
tempos, in different styles. Furthermore, this blend of evolutions
changes over time. Certain types of evolution were important in early
protolife; some are more emphasized now, four billion years later. One
variety (natural selection) will be ubiquitous throughout the plurality,
while others will be rare and specialized in their roles. Deep,
pluralistic evolution, like intelligence, is an emergent property of a
community of dynamics.
As we construct an artificial evolution to breed machines and software,
we will also need to allow for this heterogenous character of evolution.
In a functioning artificial evolution capable of open-ended, sustainable
creativity, I would expect to see the following dynamics (which I
believe reside to some degree in biological evolution but which may
appear artificially in a stronger form than we find in biology):
-
Symbiosis -- Easy informational swaps that permit convergence of
distinct lines -
Directed Mutations -- Nonrandom mutation and crossover mechanisms with
direct communication from the environment -
Saltationism -- Clustering of functions, hierarchical levels of
control, modularization of components, and adaptive processes that
modify a cluster all at once -
Self-organization -- Development biased toward certain forms (like four
wheels), which become pervasive standards
Artificial evolution will not be able to make everything. There will be
many things that we can imagine in full detail -- and that by the laws of
both physics and logic should work -- that synthetic evolution will not be
able to reach because of its constraints.
In an unconscious way the computer-toting postdarwinians are asking the
question: What are the limits of evolution? What can evolution not do?
The limits to organic evolution may not be ultimate, but its biases and
inabilities may hold answers to evolution's creative talents. Where are
the vacant black holes in the landscape of possible creatures? I can
only echo Alberch, the monster guy, who said, "I am more concerned about
the empty spaces, about the morphologies that, although conceivable, are
not realized." To paraphrase Lewontin, "An evolution that cannot make
all things, explains some things."
continue...
|