[MD] Metabiology

[Andy Skelton]

ADRIE:
> strange , this points out towards digital evolutive patterns being unknown,
> uncomputable but very rapidly evolving in quality,
> like in digital evolution is unlimited.
> this is completely congruent with biological evolution,only lacking
> interaction with the environment.
> completely congruent with quality(fittest) is undefinable

Omega (Chaitin's constant) is the probability that a program will halt
(if run on a theoretical everlasting computer). The halting problem is
interesting:

"Alan Turing proved in 1936 that a general algorithm to solve the
halting problem for all possible program-input pairs cannot exist. We
say that the halting problem is undecidable over Turing machines." -
http://en.wikipedia.org/wiki/Halting_problem

This means there is no possible program that can tell whether another
program+input will halt. When you are dealing with randomly-evolved
programs you are going to produce some that don't ever halt, e.g.
infinite loops. If you are trying to evaluate a program by its
complete output, you must run it until it stops producing output---it
halts. Waiting forever would effectively freeze evolution at that
point.

But that's not what makes this interesting in the light of the MoQ. We
could be dealing instead with programs that are intended never to
halt, and we could be evaluating them by their ongoing behavior.

The oracle measures fitness. If you started evolving programs with the
goal of a program that always produces the string "Hello, world" and
then halts, your oracle would simply compare the program's output to
the intended output and favor those programs that more closely
approximated the desired behavior. But that's evolution toward a goal;
toward a static pattern already decided.

That is great for many things. For instance, we have used such
techniques to evolve voltage regulators that work better than any
previous designs. (The best of these circuits contain components
arranged in ways we had never thought of, as well as vestigial parts
that have no functional effect on the output voltage.) So we got
better voltage regulators by testing each mutation, finding those that
regulated voltage better, and mutating them further. But this doesn't
match the kind of pattern evolution we know as life.

In my experience, all patterns are either persisting against
opposition or failing to persist. To what end? No end! Only
persistence against opposition. A pattern will evolve as long as it
has to in order to persist. If opposition disappears, the pattern may
safely remain static. Without a challenger it has nothing to be pitted
against, nothing to be measured against.

I am very much interested in evolving digital life forms but now I am
even more interested in the purpose for such a creation. I am starting
to believe that it will be most successful if its existence is purely
for its own sake. It must have selfishness at its core. (Reminds
Dawkins.)

At first, it seems like something we would not want to create. A class
of programs that simply exist for their own sake, if successful, would
take over the whole internet and crash the world economy. Right?

Actually, no. That would not be success; the infected computers would
be quarantined and shut down and the programs would have failed. At
worst, power generation would halt and the programs would have failed.

It would be more successful if it evolved a sense of moderation.
"Don't take over the whole internet; just take as much as we need to
prosper."

While some variants would evolve to avoid detection by system
administrators, others could flourish by evolving traits that
sysadmins find beneficial, thereby making them desirable.

So, yes, I think people should be releasing computer viruses into the
wild, subject to a certain code of ethics. And remember to back up
your documents and don't do all your banking online. :-)

Andy