Loading ...
Sorry, an error occurred while loading the content.

Re: [GP] Self reproduction

Expand Messages
  • Howard A. Landman
    Telmo, ... The separation of phenotype from genotype was a very big step in biological evolution (see /The Major Transitions in Evolution/ by John Maynard
    Message 1 of 14 , Apr 19, 2010
    • 0 Attachment
      Telmo,
       

      I would argue you're putting too much emphasis on the separation between DNA and organism. In truth, one could not exist without the other. This leaves us with a "chicken and egg" problem, or, perhaps more profoundly, a strange loop.

      The separation of phenotype from genotype was a very big step in biological evolution (see The Major Transitions in Evolution by John Maynard Smith and Eörs Szathmáry (Oxford University Press, 1995)).  In the RNA world scenario, early replicators were themselves enzymes (ribozymes) with an efficiency about 1/100th that of protein enzymes.  Getting from there, to RNA holding an amino acid as a "tool", to separate "holders" and "sequence coders", and finally to a primitive genetic code, required a lot of evolution and a significant increase in complexity, but provided compelling advantages.

      I think the same principles hold in evolutionary computation.  See for example F. Gruau's "cellular encoding" of neural networks.  Directly evolving networks for, say, the XOR problem is very inefficient with classical GP and the run time blows up exponentially with input size.  But if you evolve programs to generate networks, it isn't hard to reach a program which, given an integer N as input, produces a network which solves the N-input XOR problem.  In other words, solving the problem completely for all N in a separated framework is easier than solving it for (say) N=7 in a non-separated framework.

          Howard A. Landman
    • William Tozier
      ... Most of what you re asking for reminds me of Walter Fontana s abstract chemistry work (with Leo Buss) in the early 90s. My intuition is that if you head
      Message 2 of 14 , Apr 20, 2010
      • 0 Attachment

        On Apr 19, 2010, at 3:05 PM, Russ Abbott wrote:

        GP has been most successful (as far as I can tell) in evolving designs rather than programs. So the question would be whether GP can be used to generate a design of a self-reproducing entity, where one of the requirements is that the entity include its own "interpreter" for any symbolic information that it must pass form one generation to the next.

        Most of what you're asking for reminds me of Walter Fontana's abstract chemistry work (with Leo Buss) in the early 90s.  My intuition is that if you head much farther down this road, we'll be quickly discussing what "genetic programming" actually is, as well as revisiting old but often unremarked problems in the philosophy of biology.

        That would be a good thing, by the way.

        For example, by switching attention from λ calculus to a concurrent π calculus, all sorts of changes are available: more maybe than most of us can follow, or probably want. From some small experience in the matter, it may be rewarding to focus not on what kinds of things are possible, but what would be of simplest change of most use towards addressing your curiosity.


        Bill Tozier
        Line
        Vague Innovation, LLC
        Business Development for Independents

      • Russ Abbott
        Although this doesn t respond directly to Bill s suggestions, it occurred to me that the distinction we are picking at resembles that between biological
        Message 3 of 14 , Apr 20, 2010
        • 0 Attachment
          Although this doesn't respond directly to Bill's suggestions, it occurred to me that the distinction we are picking at resembles that between biological viruses and biological cells.

          Viruses can neither synthesize proteins nor store energy on their own. The are completely dependent on their host cells for these functions. Viruses are similar to quines in that they cause their host interpreting agents to make copies of themselves.

          Cells on the other hand can both store energy and synthesize their own proteins from more basic materials. They have their own interpretation machinery built into them.

          Given that, I'd say that I want a GP system that evolves the design of a cell capable of reproduction rather than one that evolves a quine.


          -- Russ Abbott
          ______________________________________

           Professor, Computer Science
           California State University, Los Angeles

           cell:  310-621-3805
           blog: http://russabbott.blogspot.com/
           vita:  http://sites.google.com/site/russabbott/
          ______________________________________



          On Tue, Apr 20, 2010 at 4:43 AM, William Tozier <bill@...> wrote:
           


          On Apr 19, 2010, at 3:05 PM, Russ Abbott wrote:

          GP has been most successful (as far as I can tell) in evolving designs rather than programs. So the question would be whether GP can be used to generate a design of a self-reproducing entity, where one of the requirements is that the entity include its own "interpreter" for any symbolic information that it must pass form one generation to the next.

          Most of what you're asking for reminds me of Walter Fontana's abstract chemistry work (with Leo Buss) in the early 90s.  My intuition is that if you head much farther down this road, we'll be quickly discussing what "genetic programming" actually is, as well as revisiting old but often unremarked problems in the philosophy of biology.

          That would be a good thing, by the way.

          For example, by switching attention from λ calculus to a concurrent π calculus, all sorts of changes are available: more maybe than most of us can follow, or probably want. From some small experience in the matter, it may be rewarding to focus not on what kinds of things are possible, but what would be of simplest change of most use towards addressing your curiosity.


          Bill Tozier
          Line
          Vague Innovation, LLC
          Business Development for Independents


        Your message has been successfully submitted and would be delivered to recipients shortly.