From: Dawsonzhu@aol.com
Date: Sun Nov 02 2003 - 21:51:42 EST
Dave Campbell wrote:
> As to evolution, both our theory and the implementation into models are
> probably faulty. Bats have really good radar, but they have had millions of
> years to refine it. If you ran the models for millions of years with millions
> of attempts, you would probably have less need to adjust the final results.
>
Although I do not claim vast expertise here either, the genetic algorithm
(GA)
is an optimization approach. What "parts" you put in are largely what parts
you get out. Time is not an issue, billions or trillions of years of CPU
time
will not find you better "parts", because only the arrangement of those
"parts"can be decided.
The argument might be better restated as a question:
(1) Is evolutionary theory (as currently explained) sufficient to produce
all the forms of proteins and ribozymes we observe today from some
reduced model set at the beginning of life?
Actually, even to have a GA to begin with means that some profoundly
sophisticated machinery is already in operation that is not explained.
Between the pea soup of nucleic acids (and amino acids) and gene
replication machinery, there is a very large gap in our understanding.
I know some people are working on it, but anyway, there are some big
question marks there.
(2) Is that reduced model set (also should add the GA machinery itself)
sufficiently simple that it could have evolved from hydrothermal vents
(or some other yet unexplored natural process)?
Let me put it yet another way, if we trace back the protein evolution to the
very beginning, can we show by simple mutation, horizontal transfer,
neutral mutation, pseudogenes, introns etc. that the level of sophistication
in protein structures we see in life today is all constructible with a
core system, or would we have to explain various proteins
that come essentially from nothing? If the latter case is true, then
evolution and that the mechanisms currently invoked would require some
further modification, or even that maybe ID has a point. If the former
case is true, then that is clearly the end of ID. I think we must admit
we don't know the answer here although the first case remains a strong
possibility.
So even put in yet another way, to get a GA to solve a problem, there
must be enough "solutions" available that by some means, the GA can
arrange those solutions in such a way that you will get the desired
result. For example, if you don't put wheels in the GAs options for
building a robot, the GA can't reinvent the wheel for you even if you
give it an enormous amount of CPU time to try. In short, a GA does
not substitute for the real effort of thinking in solving a problem, it can
only serve as an aid in finding for the best solution given what you put in.
Remember the golden rule of computerese: "Garbage in garbage out".
So if you're starting proteins (and nucleic acids) are not sufficiently
sophisticated, you'll only get "brown gunk", not "life" as we know it.
by Grace alone we proceed,
Wayne
This archive was generated by hypermail 2.1.4 : Sun Nov 02 2003 - 21:54:01 EST