All,
For reasons similar to those mentioned by others, I have found the
phrase "artificial intelligence" to be less than adequate to convey my
interests in this domain. And after considerable time, I came up with a
term that I prefer; "synthetic awareness". It comes from having
interests in several different domains which feed into my interest in
fabricating non-homo-sapien memetic propagation.
First, synthetic is more inclusive. It means that borrowing and
incorporating specialized awareness/knowledge from organic/memetic
domains is included and acceptable. It also means that fabrication of
new awareness/knowledge from strictly computation domains also works.
Secondly, awareness is more expansive than knowledge. Boolean
mathematic frames (proof focused rule based systems) and the symbolic
efficiencies around linguistics (must be able to be articulated
accurately) have most "intellectuals" fixated on producing "idealized"
knowledge. And while there is significant value in the results
produced, the results (to me) are too sequential and fragile to be
expected to scale up to extremely high levels of complexity. This is
why computer Checkers/Draughts is solved, computer Chess is not solved
but beat the highest skilled humans, and computer Go is not even
effectively beating low ranking amateurs. Awareness covers much more
complex notions like the subtleties implied in "intuition" and "creativity".
Here is my reframing of a statement by a psychology author, Nathanial
Branden:
<STATEMENT_REFRAME>
The need to create "synthetic awareness" has acquired a new urgency in
the computational age. The more rapid then rate of change, the more
fragile and dangerous it is to operate computers mechanically, relying
on routines of Boolean software and Boolean behavior that may be
irrelevant or obsolete.
</STATEMENT_REFRAME>
As has been discussed ad infinitum here on computer_go, I don't see
computer Go be solved meaning like Checkers/Draughts has been "solved".
I do think it is achievable to generate some sort of computational
result which can eventually outplay the humans of the highest skill. I
also think some significant breakthroughs are required around the move
away from booleans (perfect move-vs-imperfect move) and towards scalars
(probability of each available move will lead to overall increased value).
While the domain of the rules of Go feel very "rigid", the complexity is
so vast that any idealized solution is going to turn out to be a local
optima, i.e. with enough effort and exploration, it will be discovered
the idealized solution, too, has weaknesses which can be exploited and
eventually cause it's failure. As such, the more dynamic and creative
the nature of the resulting "entity", the more likely it will be the
entity can eventually hop out of the local optima in search of an even
higher optima. The more reserved, risk averse and rigid the "entity",
the more likely it will be unable to move forward and the sooner it will
succumb to another "entity's" discovering it's weaknesses and eventually
out-playing it.
Go is the perfect game for demonstrating that even with a perfectly
rigid foundation, the solution space is vastly more effectively searched
via dynamic evolving mechanisms than via static rigid mechanisms. And
as can be seen with the recent UCT/MC results, we are still just barely
above randomness in terms of discovering and inventing solutions.
Jim
Erik van der Werf wrote:
On 7/21/07, Weimin Xiao <[EMAIL PROTECTED]> wrote:
Intelligence is the ability to adapt or learn.
A hypothetical almighty oracle that already knows the correct answer
to every question and the right response in every situation would
never have to adapt. Hence evidence of intelligence according to your
definition would not be observed.
IMO the adaptation is just a means to an end. The end (Intelligence,
whatever it is) does not necessarily require adaptation.
Erik
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
