Re: Is a general purpose mechanism possible?


feedbackdroids wrote:
JGCASEY wrote:
feedbackdroids wrote:
JGCASEY wrote:
...
My personal interest at the moment has been thinking about
how even a simple "intelligent" system can evolve from chance
and selection. Every variable exponentially increases the
unlikelihood of any viable configuration of parts. My view
is becoming that evolution depended more on innate self
organizing properties than on chance.

Without the "right" laws (constraints on the innate behavior
of molecules) there would be nothing to select and nothing
to do the selecting.



Would be interesting to simulate evolution of 2 organisms in
competition with each other. Each would be self-organizing in
its own right, but each represents the "external" environment
to the other. Of course, the first thing to put into each is
the innate urge to explore its world. IE, not simply sedentary
until something comes along to stimulate it.


I think that has been done between programs that play chess,
checkers and backgammon. They would have shown some of the
principles involved but will always hit a blank wall for the
reasons Richard Dawkins gave with regards to his biomorphs.



I'm not completely sure what this means re Dawkins, but I'll
go look up his book chapter 1.


At the end of the first chapter he writes about the problem of
simulating the complexity of the real world in a computer.


My point, that I also made several weeks ago regards something
you said, is that there is no such thing as the "same
environment", if you [or Gould] were to start evolution over,
because the environment evolved along with the organisms in it.
The classic example from early evolution is that the
cyanobacteria ate CO2 and pumped O2 into the atmosphere, and
this was originally a toxic substance to the anerobic bacteria,
so they all had to evolve to adapt.

Later, during the cambrian explosion there were 1000s of
different athropods evolving, each being food for others,
which established the sort of predator-prey antagonism we're
familiar with today. I may be wrong, but I think the urge for
species to survive in this new terribly hostile predator-prey
world greatly accelerated the rate of evolution. Compare that
only bacteria were around for 3.5 BY, but then 1000s of species
evolved in 5-10 MY or so during the cambrian explosion.
It seems clear that a similar antagonistic situation would
push simulated evolutions.


The problem is simulating it in all its molecular detail
along with the number of units that can fit into the real world.
Each organism would require vast amounts of memory just for
its DNA simulation and there were billions of them. And that
is just a small part of the organism and we still have the
environment. As Dawkin wrote we may as well get back to
the real thing.

http://www.msu.edu/~lenski/sciencearticle.html

Test Tube Evolution Catches Time in a Bottle

...

I am beginning to question the central role of random chance
and selection for in my own attempts to evolve something it
requires that I keep adjusting the odds until it worked. You
cannot adjust the laws of chemistry. It has to be possible
to start with before you have anything to select or anything
to do the selecting.




This is true. I was mainly thinking about how to push evolution.
Eg, we know what outcome we'd like to see, natural evolution did
not and had to do everything at random from first principles
[natural physical and chemical laws].

This is where I think self organization took a part. It was
not completely random. What was selected was the self organized
but random configuration of some early structure.

Life would have required self organization, due to molecular
affinity of the nucleic acids etc. to have anything to of use
to select.

By imposing laws that preorganize random networks they are
more likely to do something interesting such as duplicating
a simple net into say a central pattern generator.

The random element acts more to shake the parts which then
click into place according to their non random rules to be
selected by their environment.


This is one of the first tasks I gave myself to explore the
question of evolving things from raw parts.

Lets say we want to evolve, by chance and selection alone,
a simple full adder made out of NOR gates. How long would
it take? There are enough gates below to make such an adder.




One thing to keep in mind here, regards natural evolution,
is that certain aspects of the genome are highly-conservative.
Eg, certain homeobox/etc genes exist in the human genome which
are similar to those all the way back to even worms and other
arthropods. Check out the HOX genes and also PAX6 genes.
IOW, once a genetic "module" had evolved that worked, in some
manner it was effectively "sealed off" from further extensive
mutation, and rather what evolved was how this module was "used"
[combined with other factors] in later evolved animals. This
sort of modular conservatism would be highly useful in simluated
evolution. Eg, once you evolve something useful like an adder,
then build in a means to use it thenceforth unchanged. This is
apparently one way that natural evolution dealt with the problem
of combinatorial explosion you mention below.


And I think the evolution of computer languages give us reason
to believe that nature may have evolved the same solutions for
the same reasons, stability and reusability.

An "object" in OOP is "sealed off" from further mutations
but it can be extended to serve a different purpose.

Nature may use the rules of modular programming ...

(see Pinker "How the Mind Works, chpt1 toward end of
Natural Computation section) [ ... all of which seems
to be followed by natural selection as it designed
our modular, multiformat minds:

modularize
use subroutines (sub modules to compose a larger module)
each module should do one thing well
make sure each module hides something
localize input and output in subroutines
choose data representation that makes the program simple.]


--
JC

.

Relevant Pages

  • Re: Insights Ive gained from my Bio class so far
    ... about problems with evolution, the examples they give of problem organisms ... percentage of the life on earth. ... the selection processes that caused their evolution in the first place. ...
    (talk.origins)
  • Re: Insights Ive gained from my Bio class so far
    ... about problems with evolution, the examples they give of problem organisms ... percentage of the life on earth. ... the selection processes that caused their evolution in the first place. ...
    (talk.origins)
  • Re: Addressing Scientific Reductionism
    ... theory of evolution the greatest thing since sliced whitebread. ... Its emphasis is on how variation in organisms is generated, ... rather than how natural selection works. ...
    (sci.bio.evolution)
  • Re: Is a general purpose mechanism possible?
    ... is becoming that evolution depended more on innate self ... organizing properties than on chance. ... so they all had to evolve to adapt. ... and selection for in my own attempts to evolve something it ...
    (comp.ai.philosophy)
  • Re: "Random evolution"
    ... Selection", so he invented the phrase "survival of the fittest". ... all honesty, organisms aren't fitter, they're just luckier... ... Evolution is a random process, ... numerous random variations thus can *non-randomly* ...
    (talk.origins)