From: Beltran Ibarra Davila-Armero (bida@cs.washington.edu)
Date: Wed Dec 01 2004 - 02:23:55 PST
The evolutionary origin of complex features
By Richard E. Lenski, Charles Ofria, Robert T. Pennock and Christophe Adami
This paper aims to prove Darwin’s theory of evolution through the
simulation of an evolving population of digital organisms that meet the
three necessary conditions for evolution: replication, mutation and
competition.
One of the main ideas is that complex features can derive from simple
ones. After the results explained all along the article, we can clearly
see that although they always start with a population which is unable to
“do” anything (no logical functions), they always end up getting the
most complex feature (the EQU function). Moreover, this achievement was
done albeit the fact that several mutations were needed to perform a
simple intermediate function.
A second idea expressed in the article is that the reward system was
necessary to complete the most complex feature. In fact, when they
rewarded only the EQU logical function, none of the populations evolved
to EQU, whereas a bit more than 34% achieve this in the reward-all
system. This reward system shows that somehow organisms need to be
“guided” through intermediate steps (the rewarding of simple functions)
in order to achieve more complex status.
Another interesting point is that all this experimentation proves that
the evolution theory based on the three precepts (mutation, replication
and competition) works. The evolution of each population has been under
those rules and has evolved by eliminating the weakest elements, just as
Darwin advocates.
As far as flaws are concerned, one point that is subject to comment is
the way the authors assert that digital experimentation is good enough
to allow results from these to be applied into the biological world. The
nature is somehow far more complex than simple organisms that operate
logical functions and there are more environmental factors that
influence the evolution of a population. It is maybe a little too
simplistic.
Another point is that Darwin’s theory of evolution is concerns all
organisms: sexual and asexual. This article only takes into account
asexual ones and leaves out the sexual ones. It could be a weak point in
their procedure because maybe the sexual organisms fail to come to the
same levels of complexity because of the higher mixing of phenotypes.
Some points were not very clear about what pivotal mutations are.
Although they give an explanation of these, they are not very precise
and they don’t explain how they calculate or come up with these specific
mutations. This is even more blurry because, as it is said, there are
not “must have” functions to achieve the most complex one.
Of course the first question that we can develop is how well this
demonstrates the theory of evolution and what are exactly the origins of
life. And on the other side are we able to predict how populations will
evolve. Maybe if the good reward functions were taken and a good pattern
of an accurate set of ancestors was implemented….
This is clearly a search problem with its state space (all the possible
the genotypes), and its heuristics (the rewarding functions). Somehow
there is some randomisation since deletion is acceptable (by deleting
some functions we lose some logical functions and fitness). Maybe we
could compare it to some local search (hill climbing) with side moves
which will allow moving down the slope.
This archive was generated by hypermail 2.1.6 : Wed Dec 01 2004 - 02:24:01 PST