From newshub.ccs.yorku.ca!torn!utcsri!rpi!zaphod.mps.ohio-state.edu!moe.ksu.ksu.edu!ux1.cso.uiuc.edu!mp.cs.niu.edu!rickert Tue Jul 28 09:42:02 EDT 1992
Article 6518 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!torn!utcsri!rpi!zaphod.mps.ohio-state.edu!moe.ksu.ksu.edu!ux1.cso.uiuc.edu!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Subject: Memory and store/retrieve. (was: Defining Intelligence)
Message-ID: <1992Jul27.171820.30707@mp.cs.niu.edu>
Organization: Northern Illinois University
References: <4474@rosie.NeXT.COM> <6guJoB3w164w@cybernet.cse.fau.edu>
Date: Mon, 27 Jul 1992 17:18:20 GMT
Lines: 140

In article <6guJoB3w164w@cybernet.cse.fau.edu> justin.bbs@cybernet.cse.fau.edu writes:
>> 
>> Could we instead say that "intelligence requires the ability to use
>> information from past experience to influence behavior"?  We could

>> Paul
>> Paul_King@next.com

>Hmmm... how can one not equate a device that uses information from the 
>past to influence [present] behavior as a storage/retrieval system?

 Since I was the one who questioned whether we possess a storage/retrieval
memory system, perhaps I should clarify further.  There is apparently
widespread misunderstanding of my point.

 Let's start with Paul King's definition of memory above as "the ability
to use information from past experience to influence behavior"?

	AN EXAMPLE OF A MEMORY WHICH IS NOT A STORE/RETRIEVE SYSTEM

 Now consider the lowly tree as an example of a system which meets Paul's
definition of memory, yet which most people would not consider to be a
store/retrieve memory.  Indeed most people do not wish to consider the
tree as possessing any memory at all.

 Before I go further, let me emphasize a few things I am NOT saying.
I am not claiming that the tree is intelligent (although it is surely
more intelligent than a rock).  And I am not claiming that human memory
is constructed the same way as a tree's memory.

 Consider now the way a tree uses information from the past.  A tree in
the open plains can be most productive if it develops leaves all around.
But a tree in a dense forest needs to develop its leaves near the top
of the tree, since the forest produces too much shade in lower regions
for productive photosynthesis.  However, when the leaves break in the
spring there is plenty of light all around, since the competing trees
in the forest have not yet leafed out to provide the shade.

 If the tree were to grow as a response to the stimulus of sunlight, it
would form leaves all around, since the light is widespread in early
spring.  Then the lower leaves would die.  However, in the forest, most
of the new leaves of early spring form near the top of the tree where
light will be available all summer.

 The new growth of the tree depends on the structure and buds formed in
the prior growing season.  Thus experience in one growing season influences
behavior in the next.  This meets Paul's requirements for memory.

	WHY THE TREE IS NOT A STORE/RETRIEVE MEMORY.

 Now back to the question of store/retrieve memories.

 When we think of storage and retrieval, we tend to think of storage as
an essentially atomic or singular event.  In the case of the tree there
is no such storage event.  Instead there is a gradual accretion of
information over an extended period of time.  Likewise we tend to think
of retrieval as an atomic event.  Again with the tree, the information
is represented in a way as to directly affect behavior (future growth)
and so is used continuously without any retrieval event.  An experienced
arborist can examine a tree, and tell you a great deal about the
information represented in the structure and buds of the tree.  But in
doing so, the arborist is not really retrieving the information, but is
inferring it from indirect evidence.

	HUMAN MEMORY AS NON STORE/RETRIEVE

 Back to human memory:

 With human memory there is much to suggest that there is a steady
accretion of information, rather than atomic storage events.  To be
sure, the rate of accretion is much faster than with the tree.  But even
simple factual knowledge, which we supposedly either know or don't know
seems to require steady accretion.  In learning we go through stages
where we can remember the supposedly learned information only after
suitable prompting and hints.  If there were a store operation, learning
should not work this way.

 Likewise, with retrieval, there is much about the way we remember to
suggest that we are really inferring the information rather than using
an atomic retrieval event.  We talk about searching our memory, but the
search time is really the time we use to make the inference.  It is well
known that several witnesses to an event will remember it quite
differently, just as you would expect if their memories were really the
result of an inference rather than a retrieval operation.  Similarly
one individual may remember something differently today from the way he
remembered it yesterday, exactly as we would expect if the memory were
really an inference from inconclusive information.

	MEMORY AS A RECOGNITION SYSTEM

 If you try to use human memory as a store/retrieve system, you quickly
discover it has great inadequacies.  The inventions of writing,
phonographs, motion picture films, magnetic tape recorders, and computer
memories, are all compensations for the fact that human memory does not
serve the store/retrieve function at all well.  But the human memory is
a marvel when it comes to recognition.  You can pick out one face in a
crowd.  You can learn all kinds of new patterns, and learn to recognize
them.  This is a remarkable and important ability, and may be one of
the most important aspects of our intelligence.

 Learning to recognize patterns is a type of memory which requires a
steady accretion of information.  Experience permits ever more precise
recognitions based on ever more minimal evidence.  A recognition system
also allows inference of information contents based on testing what can
be recognized, and if so used can give the illusion of a retrieval
system.

	THE EVOLUTION OF MEMORY:

 Imagine yourself to be the controller of evolution.  As Bill Skaggs
recently pointed out, evolution is not a creative designer.  Rather,
evolution is an incremental engineer, making small improvements at every
stage.

 We evolved from primitive ancestors.  If you look far enough you find
fish, and even further back you find invertebrate chordates.  Our
intelligence started its evolution a long time ago.  Now, in your role
of master of evolution, will you enhance the ability of a fish to store
and retrieve information?  Or will you better serve the needs of survival
if you enhance the ability to learn and recognize.

 It seems to be the functions of learning and recognition which win out
all the way up the evolutionary tree.  This is even true when you reach
homo sapiens.  Keep in mind that civilized man is not the product of
biological evolution.  It is primitive hunter-gatherer man which evolved.
And even for hunter-gatherer man, the ability to learn and recognize is
far more important than any ability to store and retrieve.

 We are left then with the problem that a store/retrieve memory offers
very little benefit that would drive evolution, while a learn/recognize
memory offers tremendous advantages.  It seems apparent that a
store/retrieve memory could not have evolved.  A learn recognize memory
appears certain to evolve, although it may well be an incredible fluke
that it evolved as extensively as it has.

 In the circumstances, if we believe we evolved, we must conclude that
we almost certainly do not have store/retrieve memory systems, and that
what we interpret as storage/retrieval is really a side effect of
something else, probably related to our learning/recognition system.



