Christopher Eliot (eliot@giane.cs.umass.edu)
Fri, 11 Dec 1998 13:18:38 -0500 (EST)
Date: Fri, 11 Dec 1998 13:18:38 -0500 (EST) From: Christopher Eliot <eliot@giane.cs.umass.edu> Subject: Re: [ifets] Technology and Brain
On Thu, 10 Dec 1998, Martin Owen wrote:
> Psychologists constructed notions of long, medium and short term memory to
> hypothesize and perform some experiments to undertand some phenomena. The
> brain doesn't have a "long", "medium" and "short" term memory devices in
> the meat machine, but they were handy constructs for that particular
> community at the time.
Plato clearly distinguishes long term and short term memory in the
Theatetus. First he compares memory to a wax tablet and then dismisses
this view as simplistic. He goes on to compare memory to an aviary,
where there are many birds (memories) in a large cage but only a small
number are in your hands (short term memory or focus of attention) at
any instant.
I don't think psychologists were the first to distinguish long term
and short term memory.
> The issue is that humans are more than meat machines.
This is a bold assertion that is at once true and also not true.
In the most important sense the claim that humans are more than meat
machines is absolutely false. An automobile is just metal and plastic
etc., even though it is part of a larger system of transportation. A
book is just paper and ink even though it is part of the record of
human knowledge. A human is just a meat machine even if he is part of
humanity and exists in a social context.
A human being in complete isolation, perhaps in a sensory deprivation
chamber will quickly become insane. But being insane does not make a human
cease to be a human. So, the social context is not a necessary element
of being human and should not be considered as a part of us.
On a practical level, we do of course live in society and our thoughts
and actions relate to the things and other people around us. We are
part of the social context and contain simplified mental models of
that context. But, I think it is important to keep track of what is a
part of what and not try to turn common sense on its head.
>They live in a social
> environment. Describing the levels of nitric oxide in the brain at any
> given time tells us nothing about what a human knows ( and never will!).
Bill Gates once said that 640K would be more than enough memory for anyone.
I do not think it is wise to predict what will not be discovered.
> It
> is only through action in the world does that happen, and thanks be, we
> have evolved into language users.
This is a confusion between evidence, computability and definition.
Currently, we don't have the measurement ability, theoretical
understanding or computational ability to figure out what a person is
thinking based upon low level measurements of brain activity.
Currently, the only evidence we have about a person's mental state is
based upon his actions and speech acts. But it is wrong to define
mental state in terms of actions and speech acts. Every person has
private thoughts which are quite real, even if they produce no behavior
and are not vocalized.
> AI scientists make too great a claim for their constructs ( unlike me!) .
I might dispute this suggestion except that it is an empty statement.
It would be helpful if you would indicate which AI scientists you
think are overstating which claims.
Or, if you prefer, I will make a claim and you can decide if this is
overstated.
I claim that a software human could be built (in principle). It could
be built to run on a computer that is not fundamentally different from
any personal computer you are familiar with. (Technically, I am saying
that a Von Neuman architecture is sufficient.) It would require vastly
more memory. Depending upon the speed of the processor it would be a
human with extremely slow reactions. But it would be a human, and I
don't mean a simulated human. It would not be a homo-sapiens, of
course, but it would be human.
I have no idea how to build this software and neither does anyone
else. Funding is probably not the major problem. 50 billion dollars
seems adequate and that amount of funding has been made available for
space exploration. It would require the cooperation of huge numbers
of top quality scientists and this may be the most difficult part
of the problem.
In case anyone chooses to dispute my claim, I will outline my major
justifications.
1) Socrates said that he knew nothing except that he was in all
other ways without knowledge. In this he claimed to be wiser than
most because at least he knew the limitations of his own knowledge.
AI research has demonstrated in many ways that human expert knowledge
is actually not that complex. (While common sense knowledge is more
complex than originally thought.) Building computer software to
achieve human levels of performance is simpler than some think
because humans do not have much knowledge.
2) After much thought, I cannot see any viable alternative to my
claim. There are many people who do not want to agree with any claim
that human thought could be equivalent to computer processing.
However, there is no other viable theory. Discussion is difficult
because there is no truly adequate definition of the basic terms
involved, thought, awareness, mind etc. However, homo sapiens have
minds and think and are aware. This is not a function of the digestive
tract or the skeletal system but all evidence we have indicates it
is because of our nervous system, in particular our brains. These
operate with neurons and chemical transmitters instead of transisters
but still perform a computation.
-Chris eliot
[...]
Christopher R. Eliot, Senior Postdoctoral Research Associate
Center for Knowledge Communication, Department of Computer Science
University of Massachusetts, Amherst 01003. (413) 545-4248 FAX: 545-1249
ELIOT@cs.umass.edu, <http://www.cs.umass.edu/~eliot/>
---------------------------------------------------------
Forum website: http://zeus.gmd.de/ifets/
Email address for sending message to everyone on the list ifets@gmd.de
Forum's contact person ifets-info@gmd.de
Join/Leave Digest or Normal List http://zeus.gmd.de/ifets/maillist.html
---------------------------------------------------------
This archive was generated by hypermail 2.0b3 on Fri 11 Dec 1998 - 19:42:21 MET