Fw: The Facts of Life and Hard AI

From: CMR <jackogreen.domain.name.hidden>
Date: Sun, 18 Jan 2004 20:29:02 -0800

> The "Emergence of Life" paper is talking specifically about those sorts
> of life that can emerge
> That's why that kind of life ("natural" life) is a truly emergent or
> (emergent from less-order) system.

Well, I'm an agnostic, but your point is well taken here. That said who was
that talking about a "god program", Hal?

> One way of looking at A.I. is that it may become in some attributes
> life-like (I prefer just to say
> it will become a true cognitive agent i.e. a true thinker (active
> modeler) without NECESSARILY
> also independently being a fully self-sufficient life-form. If WE can be
> considered part of the environment
> of AIs, then they are a life-form that uses US to reproduce (at least
> initially).

Paraticism and symbiotic raltioanships are common amoung "natual" life

> It's traditional to think of the environment of a lifeform as less
> ordered than the lifeform itself, so this
> AI case, where the environment includes extremely ordered self-emergent
> SAS's (ourselves)
> is a little bit strange situation and it's hard to categorize.

Well, speaking of symbiosis, is my gut less "ordered" than the critters like
ecoli that make a home there? I'm not sure I buy that generalization. I'm
more of the Starr hierarchal ecology ilk (by way of Koestler) or perhaps
Joslyn's meta-transition model; adaption is as much about feedback between
hiearchal scales as within them...

> With AI, it's probably best just to say that there is another emergent
> system emerging, which is
> (at this stage) a combination of humans (the human-species pattern and
> its behaviours) and the software
> (informational) and computing hardware technological/cultural artifacts
> we produce, acting together
> to form the new emergent system.

No issues with this view; I wouldn't be at all surprised if cyborgs inherit
the earth.

> People do talk about AI computers/robots and nano-tech, in combination
> perhaps, becoming self-sufficient
> (self-replicating and self-advancing/adapting independent of their human
> creators.)
> I have no trouble believing that this is in-principle possible. I just
> want to point out that
> the properties for true long-term sustainability of pattern-order are
> HARD (difficult, onerous)
> requirements, not easy ones. Natural life (in the admittedly single case
> we know) is highly constrained
> because of the constraints on its long-term survival and incremental
> improvement in a less-ordered
> environment.

"Hard" may not be the most useful term here; highly constrained, yes; Once
the conditions were sufficient, I think the rest was inevitable. Here I have
to play the "nano-tech" card; one can imagine some uber termite mound
encompassing the globe (and beyond) custom designed (grown?) from the atomic
scale up to our progeny's specs;

The prototype of just such an ecology may well already be in place (were
leveraging it now); with evermore bandwidth, interconnectivity, agency and
now the advent of "grid" distributed computation it's conceivable that
"something" is already in gestation. But I haven't a clue about that..

> It seems easier (but is it much easier really?) to get AIs to
> self-improve/self-sustain purely as virtual (informational) patterns
> or entities (i.e. as software and data ie. pure-informational
> entities/thinkers/knowledge-bases) rather than as informational/physical
> hybrids as we are. I suppose some of the people on the everything-list,
> myself included, may see the
> distinction between informational and physical as more just a matter of
> degree than of substance,
> so this is a puzzling area. Certainly both human-built computers and
> physical machines (robots eg mars rovers,
> nanobots etc) have a long way to go, not only in their basic FUNCTIONAL
> development, but
> perhaps more significantly and certainly more difficultly in their
> ROBUSTNESS (lack of brittleness)
> AND EVOLVABILITY (& META-EVOLVABILITY?) criteria, and their raw-material
> choice
> (natural life uses primarily the most commonly occurring-in-the-universe
> chemically-bondable elements
> (hydrogen, carbon, oxygen, nitrogen etc) for good reason), before they
> could hope to be very self-sustainable.

Lanier's "phenotropics" speaks to that brittleness factor. The promise and
the danger may both lay in unleashing genetic programming type strategies on
the problem; "evolving" our broods towards those goals of robustness and
self sustainability w/o really having a handle on the process. Might be
prudent to review Asimov!

> It is interesting to speculate that the mechanisms available to a future
> AI robot/nanotech-conglomerate/web-dweller
> for self-adaptation might be far more flexible and wide-ranging than
> those available to early natural life on Earth,
> because we are building AI's partly in our image, and
> we, after all, by becoming general thinker/planners (information
> maestros if you will) have managed
> to increase enormously the range of ways we can adapt the environment to
> our needs. (Caveat: As an eco-aware
> person however I can tell you the jury's out on whether we're doing this
> to system-survival-levels of sophistication,
> and the jury's leaning toward "guilty" of eco-cide - or more precisely
> guilty of severe eco-impoverishment and disordering).

I'm no extropian; at some level I'm scared to death (a state of being I
apparently share with Sun's former CTO); Just because we managed not to nuke
ourselves into oblivion ( so far), it doesn't necessarily follow that we'll
weather the coming info-nano-genetic age; far from it. But I'm also
fascinated by these musings. The combination of being a reformed sci-fi buff
and having background in evolutionary theory, I suppose.

> BTW I'm most excited today in the AI field by the possibilities that the
> combination of the WWWeb's
> information as accessed via google (and similar) and AI
> insights/technologies will have. The web is
> not a big distributed brain yet, but it could get there.
Then you gotta be finding the semantic web project intriguing, I would
Received on Sun Jan 18 2004 - 23:32:28 PST

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:09 PST