Copy Link
Add to Bookmark
Report
Alife Digest Number 046
ALIFE LIST: Artificial Life Research List Number 46 Saturday, November 3rd 1990
ARTIFICIAL LIFE RESEARCH ELECTRONIC MAILING LIST
Maintained by the Indiana University Artificial Life Research Group
Contents:
Re: what is "emergent"?
Paul Churchland's definition of "emergent" and ad hoc definitions
what is "emergent"?
Philosophy, Dogs, and neural networks
----------------------------------------------------------------------
Date: Fri, 26 Oct 90 18:45:02 PDT
From: schraudo%cs@ucsd.edu (Nici Schraudolph)
Subject: Re: what is "emergent"?
The only precise definition of emergence I have come across is the one I
recently mentioned on comp.ai.philosophy, due to Paul Churchland:
"A property P specified by its embedding theory T1 is emergent with respect
to the properties of an ostensibly reducing theory T2 just in case
1. P has real instances,
2. P is co-occurrent with some property or complex
feature recognized in T2, but nevertheless
3. P cannot be reduced to any property postulated by
or definable within T2."
(Paul Churchland, "Reduction, qualia, and the direct introspection of brain
states", Journal of Philosophy 82:8-28, 1985)
(also quoted and discussed in "Neurophilosophy" by Patricia Churchland)
By this definition at least, emergence is not a matter of degree. The
notion of degrees of emergence may sneak in when there is only a fuzzy
idea of what theories T1 and T2 the emergence is relative to. Many ad-hoc
definitions of emergence in the recent discussion did not take into account
at all that emergence is a _relation_ between a phenomenon and two theories
describing it, so it comes as no surprise that they define a fuzzy kind of
emergence.
--
Nicol N. Schraudolph, C-014 nici%cs@ucsd.edu
University of California, San Diego nici%cs@ucsd.bitnet
La Jolla, CA 92093-0114 ...!ucsd!cs!nici
------------------------------
Date: Sun, 28 Oct 90 18:29:33 -0500
From: Marek W. Lugowski <marek@iuvax.cs.indiana.edu>
Subject: Paul Churchland's definition of "emergent" and ad hoc definitions
In response to Nici Schraudolph's note quoting Paul Churchland's precise
(as opposed to ad hoc) definition of "emergent", I would like to confess
a sense of ambivalence and exasperation. Perhaps we could have a short
discussion on the subject. Gently:
1. I am unsure of the distinctions made by Paul's theory. Does it mean
I can make anything "emergent" by supplying matching T1 and T2? Why?
Why not?
2. Why is important that "emergent" not be a matter of degree? Entropy
is. I sort of thought of them using similar, thus linking, metaphors.
3. What advantage does a theory have which is not "ad hoc" if the
field is necessarily ad hoc? I think the field *is* necessarily ad
hoc because the field's theories and constructions appear to me to
rest on metaphors of which choice was arbitrary or at best inherited
from other scientific traditions. If metaphor does structure perception
and experience, why isn't every theory of "emergent" ad hoc in
methodological sense? (My relativism is showing! :)
4. I don't understand Paul's definition of "emergent". I mean by that
in cases where I want to think about the "emergent", Paul's definition
"leaves me cold."
I am eager to hear more about this business of "emergent" and the defense
of "precise definitions" in alife at this point in alife's development.
-- Marek
------------------------------
From: Andy Holyer <and@ux.rfhsm.lon.ac.uk>
Date: Wed, 31 Oct 90 16:42:04 GMT
Subject: what is "emergent"?
I did some work on just this distinction a few years ago with a view
to submitting it as a thesis subject. The definition I arived at was
that an emergent property was epiphenomenal to the system which
produced it.
The easiest example of this is the time-sharing performance of
the machine that (I presume) you're sitting in front of now. Most of
the behaviour of the machine is not emergent (at least with regard to
the formal system consisting of the semantics of the OS and any user
programs you are running. You type "who" ; you get a list of users.
this is an effect which can be altered by direct manipulation of your
system (log someone out). The point is, that there is a direct
relationship between the stimulus and the response (to be entirely
strict, typing "who" to a Unix Shell will cause the file "/bin/who" to
be run which in turn prints out the infirmation stored in a table in
/dev/kmem... you didn't really want to know all that, did you?)
Compare and contrast this with an emergent property, say "The
length of time it will take to FTP a 100k file from simtel20.army.mil
(In Jonathan's case it's probably more apt to say "the time to hhcp a
100k file from lancs.pdsoft"). There is no way to *directly* control
the speed of transfer (you can indirectly affect these by up-gunning
the bandwidth of your line, the performance of the switching gear, of
simply removing other connections from the line... but nowhere in in
the OS is a register affecting the gross line speed over a long
connection. (There's a Hofstadter article which makes a similar point
about "the number of users above which an OS will thrash".)
------------------------------------------------------------------------------
| && | |
| & & | &ndy Holyer |
| & & | Snail: Dept. of Medical Informatics & Computing, |
| && | Royal Free Hospital School of Medicine, |
| & & & | Rowland Hill St. |
| & & & | London NW3 2PF |
| & & | England |
| & & & | JANET: and@uk.ac.lon.rfhsm.ux |
| & & & | Voice: (+44) 1 794 8673 |
| &&& & | |
|----------------+-----------------------------------------------------------|
| "It's not easy having a good time. Even smiling makes my mouth hurt" |
| Frank N. Furter |
- -----------------------------------------------------------------------------
------------------------------
Date: Wed, 31 Oct 90 21:03:48 PST
From: gary%cs@ucsd.edu (Gary Cottrell)
Subject: Philosophy, Dogs, and neural networks
SEMINAR
Approaches to the Inverse Dogmatics Problem:
Time for a return to localist networks?
Garrison W. Cottrell
Department of Dog Science
Condominium Community College of Southern California
The innovative use of neural networks in the field of Dognitive
Science has spurred the intense interest of the philosophers of
Dognitive Science, the Dogmatists. The field of Dogmatics is devoted to
making sense of the effect of neural networks on the conceptual
underpinnings of Dognitive Science. Unfortunately, this flurry of
effort has caused researchers in the rest of the fields of Dognitive
Science to spend an inordinate amount of time attempting to make sense
of the philosophers, otherwise known as the Inverse Dogmatics problem
(Jordan, 1990). The problem seems to be that the philosophers have
allowed themselves an excess of degrees of freedom in conceptual space,
as it were, leaving the rest of us with an underconstrained optimization
problem: Should we bother listening to these folks, who may be somewhat
more interesting than old Star Trek reruns, or should we try and get our
work done?
The inverse dogmatics problem has become so prevalent that many
philosophers are having to explain themselves daily, much to the dismay
of the rest of the field. For example Gonad[1] (1990a, 1990b, 1990c,
1990d, 1990e, well, you get the idea...) has repeatedly stated that no
connectionist network can pass his usually Fatal Furring Fest, where the
model is picked apart, hair by hair[2], until the researchers making
counterarguments have long since died[3]. One approach to this problem
is to generate a connectionist network that is so hairy (e.g., Pollack's
RAMS, 1990), that it will outlast Gonad's attempt to pick it apart.
This is done by making a model that is at the sub-fur level, that
recursively splits hairs, RAMming more and more into each hair, which
generates a fractal representation that is not susceptible to linear
hair splitting arguments.
Another approach is to take Gonad head-on, and try to answer his
fundamental question, that is, the problem of how external discrete
nuggets get mapped into internal mush. This is known as the *grinding
problem*. In our approach to the grinding problem, we extend our
previous work on the Dog Tomatogastric Ganglion (TGG). The TGG is an
oscillating circuit in the dog's motor cortex that controls muscles in
the dog's stomach that expel tomatoes and other non-dogfood items from
the dog's stomach. In our grinding network, we will have a similar set
up, using recurrent bark propagation to train the network to oscillate
in such a way that muscles in the dog's mouth will grind the nuggets
_______________
[1]Some suspect that Gonad may in fact be an agent of reactionary
forces whose mission is to destroy Dognitive Science by filibuster.
[2]Thus by a simple morphophonological process of reduplication, ex-
haustive arguments have been replaced by exhausting arguments.
[3]In this respect, Gonad's approach resembles that of Pinky and
Prince, whose exhausting treatment of the Past Fence Model, Rumblephart
and McNugget's connectionist model of dog escapism, has generated a sub-
field of Dognitive Science composed of people trying to answer their ar-
guments.
into the appropriate internal representation. This representation is
completely distributed. This is then transferred directly into the
dog's head, or Mush Room. Thus the thinking done by this
representation, like most modern distributed representations, is not
Bayesian, but Hazyian.
If Gonad is not satisfied by this model, we have an alternative
approach to this problem. We have come up with a connectionist model
that has a *finite* number of things that can be said about it. In order
to do this we had to revert to a localist model, suggesting there may be
some use for them after all. We will propose that all connectionist
researchers boycott distributed models until the wave of interest by the
philosophers passes. Then we may get back to doing science. Thus we
must bring out some strong arguments in favor of localist models. The
first is that they are much more biologically plausible than distributed
models, since *just like real neurons*, the units themselves are much
more complicated than those used in simple PDP nets. Second, just like
the neuroscientists do with horseradish peroxidase, we can label the units
in our network, a major advantage being that we have many more labels
than the neuroscientists have, so we can keep ahead of them. Third, we
don't have to learn any more than we did in AI 101, because we can use
all of the same representations.
As an example of the kind of model we think researchers should turn
their attention to, we are proposing the logical successor to Anderson &
Bower's HAM model, SPAM, for SPreading Activation Memory model. In this
model, nodes represent language of thought propositions. Because we are
doing Dog Modeling, we can restrict ourselves to at most 5 primitive
ACTS: eat, sleep, fight, play, make whoopee. The dog's sequence of
daily activities can then be simply modeled by connectivity that
sequences through these units, with habituation causing sequence
transitions. A fundamental problem here is, if the dog's brain can be
modeled by 5 units, *what is the rest of the dog's brain doing?* Some
have posited that localist networks need multiple copies of every neuron
for reliability purposes, since if the make whoopee unit was
traumatized, the dog would no longer be able to make whoopee. Thus
these researchers would posit that the rest of the dog's brain is simply
made up of copies of these five neurons. However, we believe we have a
more esthetically pleasing solution to this problem that simultaneously
solves the size mismatch problem. The problem is that distributed
connectionists, when discussing the reliability problem of localist
networks, have in mind the wimpy little neurons that distributed models
use. We predict that Dognitive neuroscientists, when they actually
look, will find only five neurons in the dog's brain - but they will be
*really big* neurons.
------------------------------
End of ALife Digest
********************************
