Hi Folks,
I can't throw myself any further into this ... I have to get back in the
fray here. However - a couple of quick-ones for Brent and Bruno:
COL
> the instant the
> abstraction happens, from that moment on you know NOTHING about the
> current state of the distal environment...all you have is IO
> measurements.
>
BRENT
And that's all a human scientist has too - IO measurements by his senses.
COL
This is just plain empirically, factually wrong. The whole of
neuroscience for 100 years confirms that perception is a central nervous
system functiona: CRANIAL, CENTRAL. Not peripheral. The CNS qualia are
observation. The peripheral nervous system has/bestows no experienced
qualities whatever. It only feel like it does because of the CNS doing
its thing. I don't want to debate this. I shouldn;t have to. Go and look
it up. There's about a million books and papers on it. In terms of my
discussion of a scientist:
a) CNS qualia are scientific observation
b) The events in the distal world outside the scientist are the
scientific measurement.
In the scenario I painted, the scientist COMP_S has to be hooked up to
IO (you could even put the COMP_S inside a human body). Results are
identical. There is absolutely no point in simulating the distal natural
world! You are there to do SCIENCE: You don't know what the distal
natural world is, by definition! Therefore you can't possibly simulate
it ...and... if you could you wouldn't want to do science on it...
that's the nice point about this scenario. It cuts out the logical holes.
BRUNO:
Please don't get me wrong. I actually hold the universe to be literally
a mathematics...not COMP _OF_ a mathematics. There is no computer. The
universe is the computer. There's no _abstractions_ of it. There's
actual reified symbols interacting with each other, not being forced to
interact by something else..... but that's a whole other story only
indirectly related to the one I told.
In regards to being guided by what the HUMAN_S vs COMP_S scenario tells
us - as is usual in the most Popperian of science... what I get is
merely guidance in my behaviour - my choices - as a designer. I don't
get told by the analysis a 'truth' or what to do. I get told by the
analysis what NOT to do - or better - what choices to refrain from
making. The COMP_S is an _abstract_ symbol manipulator. When you
abstract X, then all relations between X and everything else (through
the parallelness with all the other entities in the universe) - are all
gone.
Classical mathematical thinking is a single stream - a single line of
proof. Our abstractions are the axioms in such a style of proof. But
that's NOT what any entity X (the intended abstracted entity) has. If
there are 10^1234 entities in the universe then X has (10^ 1234 -1)
relationships with everything that is not X. Our models throw all of
them away. I cannot prove that these relationships play no part in
scientific behaviour. Nor can you. So in my design I shall, in the
popperian tradition, refrain from making any choice that eliminates
those relationships. So: abstract COMP is OUT as an option.
You said:
"That we cannot build constructively a scientist is correct. But it is
then very misleading to use the word "false". Also, it is not because
we cannot built constructively a scientist that we can infer that we
are unable to isolate one, or to copy one, (and then: without
constructively proving that we have done so). We just cannot know who
we are.
By using "false" here you change the usual meaning of the word, and it
could lead to add misunderstandings in a field where there are already
many misunderstandings."
OK we're getting to the nub of it. Firstly I disagree that "we cannot
build constructively a scientist". I hold that we can... unless by "We"
you mean "All those people who subscribe to COMP" .. in which case you
are correct. I hold that we CAN. I am not interested in copies or
'simulations' = pretending. My scenario demands authentic original
science done from a point of view of incomplete knowledge of a system
from the vantage point of being literally built inside the system being
scientifically described.
RE: 'falsity', 'refutation' etc.
The BEER definition of
COMP = '...the theoretical claim that a system's behavior derives from
its instantiation of appropriate representations and computational
processes." Beer, R. D. (1995), 'A Dynamical-Systems Perspective on
Agent Environment Interaction'. Artificial Intelligence 72(1-2):pp.
173-215.
Such a definition does not preclude "scientific behaviour" as a "system
behaviour". Therefore COMP predicted that a scientitist can be created
by "instantiation of appropriate representations and computational
processes" , which I take to mean, as most would , a Turing
Machine/digital computing abstraction-based symbol manipulator.
*FACT 1:* COMP cannot deliver one very specific, highly specialised
thing: an authentic scientist. For all the reasons I have supplied in
previous posts.
YES you can simulate a scientidst...! From that point of view COMP is
true. But if you could you wouldn;t want to because all the science must
have been done already... otherwise you wouldn't be able to simulate it!
The logic is compelling.
By the BEER definition (which is only a theoretical claim anyway!),
because FACT 1, COMP is false in the specialised circumstance....
therefore it is methodologically false as a general/universal claim.
I don't give a stuff about the survival of COMP as a concept. All I care
about is making good decisions. A good decision involves NOT choosing
COMP to build something capable of doing authentic original science on
the a-priori unknown....and... something that is provably capable of
such a scientific act is logically arguable to be capable of less
rigorous, everyday problem solving of the kind humans do....
COMP is refuted as a general claim. It doesn't mean it's not useful. It
just means that you can't simulate a scientist. You can still use COMP
to do a whole lot od things.... just not artificial GENERAL intelligence.
The useful way ahead is to forget about 'computing' the universe and
start discussing the universe AS innate computation (there is no
computer!). But this is not COMP as I understand it. You can rearrange
the meaning of the words all you like... there's no intrinsic value in
being right by a definition! What need to do is make correct choices...
so COMP - as defined above - is a dead idea...way past its use-by date.
I don't need to think about UTMs or dovetailers or anything else
designed to make up for the fundamental shortcomings of a COMP approach
... I have an empirically testable, logically arguable, here and now
decision-maker.
In my INORGANIC_S ...there is still 'computation'... but the computation
involves manipulation of natural symbols and the 'rules of
transformation' of it, as a calculus ...are literally Maxwell's
microscopic equations. All relationships with everything else that is
NOT (INORGANIC_S) are conserved and allowed to influence the dynamics.
If you want to redefine COMP to include such natural symbol
manipulation...fine...! It makes COMP true, but changes nothing....I
still decide to do the same thing...NOT to use Church-Post-Turing-based
approaches which force the computing to be dissociated from the computed
through any abstraction process..
So... Marc... you can let Mr Yudkowski know that his ideas are formally
refuted as a way to AGI. I am thinking of going to the AGI conference
next year... in which case I'll tell them all!
Until next time....
cheers
colin
================================
I can;t do any more now... please run with it... I have to dissappear
back into my work... I have an INORGANIC_S to build!
Thanks for the interesting sparring...to be continued another day...
maybe when my COMP refutations come out in print.
cheers
colin
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list.domain.name.hidden
To unsubscribe from this group, send email to everything-list+unsubscribe.domain.name.hidden
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---
Received on Tue Sep 02 2008 - 20:12:00 PDT