question from a solver

[bkl...@gmail.com]

[from anonymous, 23 June 2008]
I've been thinking about open-ended results in a closed system, from a
slightly different direction. The innocentive challenge fits well with
this.

Here is a question: can the innovation be specified, given a certain
number of constraints?

I see two different innovation possibilities: environmental and
'internal'

Internal example: (viz. Lenski's report) say you delete all genes that
are related citrate metabolism in an organism and for extra safety,
you delete all 'tri-carboxylicacid' metabolic genes. You then supply a
regular energy source and citrate as an energy source let the system
run and test for citrate metabolism. Would the innovation challenge be
satisfied if you had citrate replete conditions and deplete in other
energy sources? This would be applying adaptive pressure to the system
to start to metabolize citrate.

environmental example: say you have a functional system that is
geographically separated from an energy source that is 'better' than
the energy source available locally. To get to the energy source,
there are contraints of distance, safety, mortality, generational
propagation, horizontal adaptation info transfer, etc. In addition,
the system has an internal or external random number input. The system
is run and the innovation checked for would be the discovery of, and
utilization of the new energy source. The use of this energy is not
specified, but the system has all the possibility of using it, it is
just waiting for the "random' chance to find and use it.

Okay, it seems there may be a difference in environmental adaptation
versus a genetic adaptation to an already specified condition.

What is my question? I'm not quite sure... Can a specified innovation
be offered to constrained system to see if adaptation is possible?
This seems to represent an 'evolutionary pressure'...is this allowed?

Hasn't this kind of stuff been demonstrated in computer simulations?

[Brig Klyce replies 23 JUne 2008]
A specified innovation can definitely be offered. One could probably
list a thousand innovations to try for. I have been reluctant to
specify anything because, in my observation, here's what happens --

(Example I) An experimenter or team lists a target innovation. Then
the means to reach it are supplied in a handful or few dozen pieces.
The computer has time and capability to try a significant fraction of
the combinations (puzzle-solve) without any danger of failure. It
finds a solution. The experimenter or team claims success. The public
is told that something like OEEI-QS has been demonstrated.

But in fact, there is absolutely nothing open-ended about it. The
computer was given a specific task, and the tools to accomplish it,
and it did. Nothing else will ever come from that setup.

But suppose (Example II) a computer has modem hardware, but no modem
software. An artificial life model is under pressure of some kind.
Without prompting, it writes the modem software and "escapes." That
would demonstrate OEEI-QS, I believe. The model has shown creativity
and could probably do more.

But if I set up modem-software as the task, the danger is that modem-
software will be targetted as in Example I, the task will be
accomplished, nothing else will ever emerge, but OEEI-QS will be
claimed.

[...I'm gonna post this on the Google discussion forum. I think it is
useful. Thanks. Brig]