Problem Definition, Discussion, and Question

[Bud]

In trying to set up a definition of this problem and also define the
criteria for a solution, I have run into some issues.

First, from the discussion earlier about Lenski's results, it seems
clear that no biological system can provide an acceptable proof. This
is unfortunate because most of the terms used in stating the problem
are defined and well known in the realm of biology, but are vague or
undefined in the area of computer software.

For example, when a quarantined system is defined as one into which no
new instructions are introduced, this can be understood in the realm
of biology as meaning no additional genetic material is introduced.
However, with computer programs, the meaning is far from clear. At
the processor level, there can be no new instructions because all of
the possible machine level instructions are defined by the hardware,
and at least for modern processors, are pretty well known and
documented. Furthermore, the set of instructions is provably
complete, in the sense that anything which can be computed can be
computed using only the instructions already available. So from this
perspective, in a computer model, the requirement prohibiting the
introduction of new instructions is meaningless.

Evolution is understood to occur in the biological realm in response
to environmental pressures. Evolution in genetic algorithms is driven
by some preprogrammed scoring system. The current population of
programs is copied with some random mutation. All versions are run
and scored, and the lower scoring instances are eliminated. In this
manner, only the better performing programs are retained from each
generation. The “environmental pressure” is part of the supervising
program.

If the scoring system is set up to favor programs which perform a
fixed task faster, it may be said that this is an example of
optimization, and therefore does not qualify as OEEI. However, it is
conceivable that the scoring system could be constructed to favor a
“new” capability such as SpellCheck. For example, a simple program to
compose “sentences” containing random combinations of letters in
groups (“words”) would be straight-forward. If this program were
incorporated into one of the genetic programming systems and scored on
the basis of how many words were spelled correctly, it could
conceivably evolve into a program which checked its own spelling. It
would include SpellCheck. As a result, all of the words in the
generated “sentences” would be spelled correctly.

However, the dictionary for the SpellCheck function would be
incorporated into the scoring system, so in one sense, it would
already be in the program. Question: Would the evolution of
SpellCheck under the conditions described constitute OEEI?

[bkl...@gmail.com]

Re: Lenski -- Yes, biological systems are problematical because 1) it
is hard to know what you are starting with and 2) contamination is
rampant -- worse than I knew.

Re: in a computer model, the meaning of "quarantined" -- All I mean is
that once the experiment starts, no more keystrokes, discs, downloads,
human input, signals, etc. etc may be introduced.

Re: Spellcheck -- I think this question is answered in another topic:
"Question from a solver."

[Bud]

I am sorry, but having just read your post to "Question from a solver"
I do not think you answered my question. Would the evolution of
SpellCheck under the conditions I described be OEEI?

> > SpellCheck under the conditions described constitute OEEI?- Hide quoted text -
>
> - Show quoted text -

[bkl...@gmail.com]

No. See latest discussion under "Didn't Lenski...?" (The supplied
dictionary would be a monstrous cryptic program.)