The Naked Scientists

The Naked Scientists Forum

Author Topic: description of my research  (Read 1500 times)

Offline paros

  • Jr. Member
  • **
  • Posts: 24
    • View Profile
description of my research
« on: 12/08/2013 19:18:58 »
In 2009 I completed a piece of software inspired by Ken Stauffer's Evolve4.0c.  These simulations are called "ecosystem simulations" because they are genetic algorithms performed in "real space" with organisms who must live out their lives in an environment.   We use simple grid worlds to perform this simulation, because evolution is gratuitously slow, and desktop computers are limited in power.   Superficially, this research is categorized as "Artificial Life" , or ALife.

There are two styles of  ALife to speak of. One is the 1990s-styled Alife, and the other is 21st century Alife. 

In the 1990s-styled Alife sim, the organisms are given high-level actions to perform in the environment such as "ATTACK" and "EAT", or most embarrassingly, "REPRODUCE".   These are actions normally associated with entities who live in video games.  Regardless of this roughshod approach, those simulations did actually exhibit Darwinian natural selection, and they would self-perpetuate for weeks on end.  1990s-styled ALife stands as proof-of-principle for natural selection actually working the way scientists say it does.

21st-century styled Alife will henceforth be called 21C-Alife for short.  In rudimentary form, the simulation will primarily be concerned with some sort of "machine" which builds a copy of itself.  In these simulations we do not take for granted behaviors as eating or even reproducing.  An "attack" command would not be allowed.   Instead we want eating to actually physically be performed by the simulation's kinematics. Reproducing does not happen instantly, instead the organisms within the simulation have the duty of actually constructing  a copy of themselves in realtime from parts.  Things like "eating" are then seen as very abstract actions.  "Eating" is the machine selecting and discarding portions of the environment that it takes into itself through a pore in its body ("mouth").  Abstractly, "eating" is a form of discrimination of the environmental material by the organism/machine. In most cases the concept of "body" is also abstract here.

Despite the name I have given, 21C,  initial work on this was done by John Von Neumann in the 1950s. He wanted a cellular automaton that could reproduce a copy of itself,  with the target of the research being the question "Can a machine build a copy of itself?".   Von Neumann showed the answer to that question is yes. 

For decades, this research floundered and went ignored.  Very recently it has been picked up again by Tim Hutton, Robert Freitas, Ralph Merkle, and myself.  There is a parallel research tract being done by biochemists to produce a nano-scale protocell using real DNA, PNA, or similar.  It is my opinion that both approaches should inform one another. 

A number of foundational questions lie at the center of this research.

  • What is the qualitative difference between the self-replication seen in crystal growth, versus the replication of offspring from a stored string/data/genotype?   In the most abstract, mathematical descriptions of self-replication, we can assert that the various branches of a snowflake are "copies of themselves".  But that seems to be lacking something; some je ne sais quoi, which we understand only intuitively. Can we quantify this intuitive notion?    Progress on this question may involve deeper understanding of the relationship between form and function.
  • What precisely is the difference between raw material to be eaten for the purposes of building up parts of a machine's body, versus the machine's body?  In other words, how random does the "soup of raw materials" need to be for the machine to be said to be "constructing" itself to begin with?    This question is asking for a quantitative answer to what the verb "to construct" actually means.
  • The existence of self-replicators has two questions. Existence question 1.  Can replicators only exist in our universe at the molecular level?
  • Existence question 2.  Molecular self-replicators exist. Look out your window to see a tree.  That was easy to confirm.  However, trees have no eyes, no central nervous system, and they do not copy themselves through the use of levers, arms, or grippers.  Pushing the proverbial pendulum to the far extreme, we can imagine factories wherein the human engineers have been replaced by powerful artificial intelligence Androids, complete with hand-eye coordination,  complex memories, and language.  So we have a spectrum with two trivial answers standing at the poles.  Molecular self-assembly, "Tree-like" replication on one end, and powerful androids working factories at the opposite end.  Existence question 2 asks, are there any plausible replicators between these two extremums?   

In regards to the existence question no.2,  it is the opinion of this author that Von Nuemann's constructor strongly suggests that there can exist such a replicator between the two trivial extremes. That machine would be an artifact operating at "macro" scales that is capable of building a high-fidelity copy of itself.  Its manifestation in reality may end up being smaller than can be seen with the human eye, (but still significantly larger than molecules) and some early versions may have to replicate underwater.

Starting from a discipline of computer simulation, an ALife simulation containing self-replicating constructors is desired. And this simulation raises questions about how large the "genotype/data/tape" would need to be relative to the size of the rest of the machine.  Orthodox approaches derived from Von Neumann usually have the tape's information in identical size to the rest of the machine's parts, but this requirement is not strict.  This is seen in the videos linked in citation.

In any case, this research is highly interdisciplinary, and so the citations are drawn from branches of science that are widely separated. 


Citations

http://en.wikipedia.org/wiki/Von_Neumann_universal_constructor [nofollow]




{book} Kinematic Self-Replicating Machines (Landes Bioscience, 2004) ISBN 1-57059-690-5

For comprehensive citations see, http://www.thenakedscientists.com/forum/index.php?topic=48610.0




 

Offline grizelda

  • Hero Member
  • *****
  • Posts: 740
    • View Profile
Re: description of my research
« Reply #1 on: 14/08/2013 18:09:01 »
Quote
  • What is the qualitative difference between the self-replication seen in crystal growth, versus the replication of offspring from a stored string/data/genotype?   In the most abstract, mathematical descriptions of self-replication, we can assert that the various branches of a snowflake are "copies of themselves".  But that seems to be lacking something; some je ne sais quoi, which we understand only intuitively. Can we quantify this intuitive notion?    Progress on this question may involve deeper understanding of the relationship between form and function.
  • What precisely is the difference between raw material to be eaten for the purposes of building up parts of a machine's body, versus the machine's body?  In other words, how random does the "soup of raw materials" need to be for the machine to be said to be "constructing" itself to begin with?    This question is asking for a quantitative answer to what the verb "to construct" actually means.
  • The existence of self-replicators has two questions. Existence question 1.  Can replicators only exist in our universe at the molecular level?
  • Existence question 2.  Molecular self-replicators exist. Look out your window to see a tree.  That was easy to confirm.  However, trees have no eyes, no central nervous system, and they do not copy themselves through the use of levers, arms, or grippers.  Pushing the proverbial pendulum to the far extreme, we can imagine factories wherein the human engineers have been replaced by powerful artificial intelligence Androids, complete with hand-eye coordination,  complex memories, and language.  So we have a spectrum with two trivial answers standing at the poles.  Molecular self-assembly, "Tree-like" replication on one end, and powerful androids working factories at the opposite end.  Existence question 2 asks, are there any plausible replicators between these two extremums?   


So you're looking for an analog of a natural system that replicates itself at its smallest level and replicates the replication of that replicant replicating other systems in the environment, paring off the losers by natural selection. Right. Carry on.
 

Offline RD

  • Neilep Level Member
  • ******
  • Posts: 8125
  • Thanked: 53 times
    • View Profile
Re: description of my research
« Reply #2 on: 14/08/2013 23:31:47 »
... There is a parallel research tract being done by biochemists to produce a nano-scale protocell using real DNA, PNA, or similar.

Careful now ... http://en.wikipedia.org/wiki/Grey_goo
« Last Edit: 14/08/2013 23:53:34 by RD »
 

The Naked Scientists Forum

Re: description of my research
« Reply #2 on: 14/08/2013 23:31:47 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums