From: Jim Ledbury (jim.ledbury@dsl.pipex.com)
Date: Mon Oct 18 2004 - 22:39:54 BST
Hi, Scott, Mark,
In no particular order...
Turing test was coined c. 60 years ago to try to get people to think
about the problem of thinking machines. I think we have to think a
little more specifically now, like can they understand the halting
problem and how do we test for this.
AI was and perhaps still is very naive in its assuptions of what it
takes to produce thought. At first, imbued as we were by the classics,
all we thought was that it took a little bit of image processing,
language parsing, logic and hey bingo! I don't know how long ago Scott
gave up on AI, but certainly stuff like genetic algorithms and neural
nets offer some form of insight into how how random variations turn into
adaptive features and how feedback mechanisms with relatively simple
linkage rules can start to provide meaningful output presented with
complex patterns.
I haven't a clue whether something reasonably AI-ish can be built in my
lifetime, but I won't be asking it into how it feels if this should
happen: perhaps the element of emotion is intrinsic to what we would
accept as intelligence; maybe we won't recodnise something without
emotion. I doubt if something we might wire up in the next decade or so
would have anything like the the complex interplay of quality that are
to us emotions. These patterns to hundreds of millions of years to come
about and the question as to whether this can be answered in terms of
matter or requires *something else* is to say the least somewhat premature.
Adding complexity to a system can have a very great effect. It can
cancel out dominant effects and bring to the fore hitherto unobserved
variations.
Space-time organisation seems to be pretty important to the physical
organisation of matter. I'd say that it probably underlied the workings
of our brains in fact: apart from the odd EPR effect pretty much
everything in the universe occurs due to spaciotemporal proximity.
Okay, there might be some non-local stuff which is important but at best
I suspect these are delocalised feedback loops. Perhaps we can find
this out in AI hardware.
The argument that "if the only channel from one electron to another is
assumed to be a photon, then there can be no gestalt bigger than one
electron absorbing or emitting one photon" is plain silly. Electrons
tend to occur in probability patterns determined by their surroundings.
If the physical juxtaposition of molecules has something to do with
memory, this will affect the response of the electrons. One transistor
produces a switch, two transitors produce an oscillator.
Space and time exist in my dreams because evolution has been hammering
to get something to correspond to physical associations into my thick
skulled mammalian and previous ancestors for millions of years. Either
it worked or we have been incredibly lucky at roulette.
Regards,
Jim.
Scott Roberts wrote:
>Mark,
>
>
>
>>I do have a comment re your rejection of the materialistic theory of
>>consciousness (see below). Personally, I'm on the fence about this
>>because I think that, so far, science has not managed to to come up
>>with a configuration of matter that produces anything close to self-
>>awareness. However, if AI scientists do, in time, produce computers
>>(or more likely thousand or millions of computers running in
>>parallel) that are able to demonstrate truly heuristic decision-
>>making, not just super-sophisticated logic-tree or algorithmic type
>>processing, this would probably convince me that consciousness can
>>emerge from matter, especially if such a machine passed a rigorous
>>Turing test with colors flying.
>>
>>
>
>It's not going to happen within my lifetime, and I doubt it will happen at
>all. I spent 5 years as a grad student of computer/cognitive science, but
>quit after realizing that the whole enterprise was based on fallacious
>premises (see below). So the possibility of a machine passing the Turing
>test is, to me, irrelevant. I have to choose now whether to think about
>consciousness in materialist terms or in non-materialist terms. For the
>reasons given, I have chosen non-materialism. Build such a machine, and
>then ask me again.
>
>
>
>>My questions to you, and anyone else who wants to answer are: Would
>>this scientific development convince you that awareness is
>>fundamentally materialistic?
>>
>>
>
>Probably not. There is no denying that consciousness is affected by the
>workings of the brain. But that does not, and I don't see how anything can,
>discriminate between the idea that the brain creates consciousness and the
>idea that the brain channels consciousness. So if a machine could pass the
>Turing test, how do we know it is not channelling, just as a human brain
>does?
>
> If not, is there any scientific or
>
>
>>rational development that would?
>>
>>
>
>Who's to say? "Any scientific or rational development" is meaningless. And
>what do you have in mind of a rational development that is not scientific?
>My choice on how to think about consciousness is based on applying reason
>to current knowledge, both scientific and non-scientific. The latter
>includes knowing what it is like to be conscious.
>
> If not, would you then agree that
>
>
>>your continued belief in the non-material nature of consciousness is
>>irrational?
>>
>>
>
>As long as there are non-materialist explanations for the same data,
>choosing materialism or non-materialism is always a non-scientific choice.
>Would you say that, based on current knowledge, including your knowledge of
>what it is like to be conscious, a belief in a material nature of
>consciousness is rational?
>
>
>
>>Now, my one comment on Scott's reasoning for rejecting the
>>materialistic theory of awareness...
>>
>>On 15 Oct 2004 at 17:06, Scott Roberts wrote:
>>My reasoning on rejecting a materialist theory of consciousness is,
>>briefly, that we are aware of big things, but materialism supposes
>>that awareness comes about by the brain putting together a lot of
>>little things (like photons or molecules). This is, in my view,
>>impossible, since each little thing is separated in space and/or time
>>from each other little thing. Since the brain is also composed of
>>little things, there could be no awareness of anything bigger than
>>the little things.
>>
>>msh comments:
>>I think a nuerobiologist would say that a human brain and nervous
>>system is considerably more than a mass of "little things" out of
>>spatial and temporal contact with one another, or with the outside
>>world. Jim might want to correct me on this, but my impression is
>>that science regards the brain as an incredibly complex system of
>>billions of nuerons and synapses that have evolved over billions of
>>years to work together to provide, so far, unmatched parallel
>>processing power.
>>
>>
>
>Sure it is incredibly complex. But adding complexity and parallelism has no
>effect on my argument. If one assumes that the brain is solely a
>spatio-temporal mechanism, that is, one assumes that that which "makes
>consciousness" does not require bringing in non-locality, then my argument
>holds. If the whole thing reduces to electrons and photons, and if the only
>channel from one electron to another is assumed to be a photon, then there
>can be no gestalt bigger than one electron absorbing or emitting one
>photon. And as I said, to be aware of even that requires memory. Now it
>could be that this electron/photon business is inaccurate, that there may
>be unknown physical laws that have a bearing. I would bet, though, that
>these unknown physical laws will have no spatiotemporal model, as the
>current ones do not. So it seems pretty silly to me try to build a
>conscious machine based on spatiotemporal models, as all current work in AI
>does.
>
>A functionalist will argue that there is no merging, that perception "just
>is" the highly organized multitude of lowest level events. This is just
>arm-waving. One needs to bear in mind that it is a belief in materialism
>that creates the problem of consciousness in the first place. Back in
>Locke's day was born the idea that all explanations should be in terms of
>what we perceive (empiricism). Locke assumed that space and time existed
>independently of our minds, that in perceiving we add color, taste, and so
>on to spatio-temporal things. But, staying in this scenario, we learn that
>the channels from the things to our minds are photons, air vibrations, and
>molecules, ultimately (according to present theory), photons, electrons,
>and quarks, which we don't perceive, and which we cannot even picture as
>spatio-temporal things (we need mathematics to describe their workings, or
>more accurately, how to predict how they will appear to spatio-temporal
>observation). So, now, what it matter? It is not the big things of our
>perception, since that is all put together by our minds. But, illogically,
>the materialist assumes that the process of perception can be described in
>terms of the little things *as if the little things (electrons, photons)
>exist in the same sort of reality that the big things do*. In other words,
>the empirical dictum that works well with describing how big things work is
>being illegitimately applied to the process that creates the big things in
>the first place. This is an attempt to pick oneself up by pulling on one's
>hair (the Baron Munchausen fallacy).
>
>By the way, if you can't imagine that space and time are also added by the
>process of perception, ask yourself where the space and time of your dreams
>come from.
>
>- Scott
>
>
>
>MOQ.ORG - http://www.moq.org
>Mail Archives:
>Aug '98 - Oct '02 - http://alt.venus.co.uk/hypermail/moq_discuss/
>Nov '02 Onward - http://www.venus.co.uk/hypermail/moq_discuss/summary.html
>MD Queries - horse@darkstar.uk.net
>
>To unsubscribe from moq_discuss follow the instructions at:
>http://www.moq.org/md/subscribe.html
>
>
>
>
MOQ.ORG - http://www.moq.org
Mail Archives:
Aug '98 - Oct '02 - http://alt.venus.co.uk/hypermail/moq_discuss/
Nov '02 Onward - http://www.venus.co.uk/hypermail/moq_discuss/summary.html
MD Queries - horse@darkstar.uk.net
To unsubscribe from moq_discuss follow the instructions at:
http://www.moq.org/md/subscribe.html
This archive was generated by hypermail 2.1.5 : Tue Oct 19 2004 - 01:17:39 BST