John,

On Thu, Jul 12, 2012 at 4:07 AM, John G. Rose <[email protected]>wrote:

> I thought that self-organizing systems are routinely created and hosted
> within software environments. An example would be a multi-agent host where
> the agents congregate and spontaneously emerge a relatively coordinated
> behavior that solves some individually unrelated goal.
>

The problem with present learning and self-organizing systems is that they
are SO slooooow. My own theory is that this comes from discarding temporal
clues, which is a normal part of chopping time into slices. Some good
mathematicians have worked on what is left after the chop-job, and it just
doesn't seem to be enough to self-organize and learn at a useful rate.

> ****
>
> ** **
>
> A problem is taming the self-organization across complexity barriers in
> such a way as to reach higher thresholds of sophistication, perhaps going
> through multiple stages of self-organization to optimize problem solving
> abilities using computational expense minimizations awarded though the
> coordinating of the self-organization superstructure.
>

I suspect that this path will become clearer, once we figure out how to do
one stage of self-organization.

That superstructure being a systems automata which imposes feedback through
> itself, across internal complexity regions, while learning symbiotically
> with its informational environment.
>

I suspect that chopping things into regions will have problems, because we
don't know where to chop, and there is probably a vast hierarchy of
sub-regions, each of which needs its own separate direction to quickly
converge.

> ****
>
> ** **
>
> In this particular construct I described maybe the overall superstructure
> behavior and coordination is analogish, versus the individual
> self-organizing being more constructed digitally. Analog control itself
> being a self-organized resultant, a fine tuner.
>

At the very bottom, a design decision will have to be made where "analog"
is handled by some continuous quantity like voltage, current, or charge; or
whether everything will be digital. Both approaches have their considerable
problems.

> ****
>
> ** **
>
> Just some thoughts…
>

... to which I have added my own.

BTW, I can't tell whether the others here really grok the distinction
between the unidirectionality of Von Neumann computers, and the
bi-directionality of typical analog computing, and how it relates to
formulas vs. equations. I thought that you might take a shot at this, as
you obviously have the background to do so, and you seem to better "speak
the lingo" here than I do. I liken this to comparing arithmetic to
calculus. Perhaps you have better words?

Thanks John.

Steve



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to