вход по аккаунту


Origins of the Cognitive Revolution

код для вставкиСкачать
Origins of the Cognitive
October 7, 1998
The Cognitive revolution in
Computer science
Anthropology, sociology...
Cognitive metatheory (Baars)
What are we talking about (when we talk about cognitive
science)? Cognitive metatheory: (Baars) "…a belief
that psychology studies behavior in order to infer
unobservable explanatory constructs, such as
"memory," "attention," and "meaning." (144). "The
cognitive revolution took place in many places at the
same time, and involved a number of areas, including
memory, language, imagery, and attention. (147)…a
metatheory that encourages one to infer unobservable
theoretical constructs from empirical observations.
Howard Gardner
"a contemporary, empirically based effort to
answer long-standing epistemological questions
-- particularly those concerned with the nature
of knowledge, its components, its sources, its
development, and its deployment."(6)
Gardner's list of key components of
cognitive science:
1. Mathematics and computation: by the
1950s, scientists were comfortable with
the idea of an algorithm that could be
specified in very general terms, and
which could in principle be computed
automatically. Mathematical proofs were
themselves now something that could
be studied mathematically (David
Hilbert, Kurt Gödel); mathematical truth
could be viewed as formal consistency.
David Hilbert 1862 -1943
Kurt Gödel 1906 -1978
John von Neumann
Born in Hungary; 1903 - 1957.
Early work on mathematical foundations
of quantum mechanics (operators in
Hilbert space).
Working on Gödel's problem when he
cracked it.
Credited with the design of the modern
John von Neumann
Gardner's List: 2. The neuronal model
McCulloch and Pitts showed that
"anything that can be exhaustively and
unambiguosuly described by a suitable finite
neural network." (von Neumann).
Claude Shannon's MA thesis: similar
property of relay circuits.
Thus: binary circuits can embody logical
Gardner's themes:
3: The Cybernetic Synthesis
The core idea: the nervous system
operates in a continuous relationship of
feedback with the environment,
modifying its activity in order to best
satisfy achievement of the current goalstate (a future, not-yet-achieved state).
Gardner's Themes: 4. Information theory
Claude Shannon (electrical engineer at
MIT and Bell Labs):
showed that there was a quantifiable
notion of information. Information is
what is not redundant in a message.
What was critical was showing that
these were hard, cold items submissible
to mathematical analysis.
Shannon suggested that the information
content of a communication channel
was equal to
S pi log (pi)
Norbert Wiener: "Information is
information, not matter or energy. No
materialism which does not admit this
can survive at the present day." (1961).
Gardner's themes:
5. Neuropsychological syndromes: the
study of aphasias, and many other
function disruptions caused by brain
Howard Gardner's "key features of
cognitive science":
пЃ® Representations
пЃ® Computers
пЃ® De-emphasis of affect, context, culture,
and history
пЃ® Belief in interdisciplinary studies
пЃ® Rootedness in classical philosophical
Cognitive revolution
dealing with problems of mind as
problems of information-processing.
= an algorithm (an explicit set of formal
steps) to modify digital information.
The impact of technology on the
metaphors that guide our thinking
about the mind: Daugman 1990
"...the water technology of antiquity
underlies...the Greek pneumatic
concept of the soul...
"...the clockwork mechanism proliferating
during the Enlightenment are ticking
with seminal influence inside"
la Mettrie's L'Homme Machine (1748);
"...Victorian pressurized steam engines
and hydraulic machines are churning
underneath Freud's hydraulic
construction of the unconcsicous and its
libidinal eocnomy;
"the arrival of the telegraph network
provided Helmholtz his basic neural
metaphor, as did reverberating relay
circuits and solenoids for Hebb's theory
of memory...
" would be folly for us to regard the
recent computer bewitchment of
theoretical work ... as an entirely
different kind of breakthrough in the
history of ideas....
"Yet there are many ...who ask precisely
that we not think of computation as just
the contemporary metaphor, but instead
that we adopt it as the literal description
of brain function..."
"Thus, for example, Zenon Pylyshyn
complains that 'there has been a
reluctance to take computation as a
literal description of mental activity, as
opposed to being a mere heuristic
"It might be said that a cornerstone of
Western thought ... is the notion that
persons are embodied
spirits....Michaelangelo's Sistine fresco
of Adam...Descartes...Pygmalion...
Hydraulic and mechanical metaphors:
Began in pre-Socratic thought, with four
humours (Hippocrates): phlegm, bile
(black, yellow), and blood. Evolved into
Galen's animal spirits.
Clockwork: Descartes:
I wish that you would consider all of these
as following altogether naturally in this
Machine from the disposition of its
organs alone, neither more nor less
than do the movements of a clock or
other automaton from that of its
coutnerweight and wheels...
And the best-known of all,
de la Mettrie (L'Homme Machine):
the human brain and body: "a machine
that winds its own springs -- the living
image of perpetual motioin is an
assemblage of springs that are
activated reciprocally by one another."
The hydraulic image reemerges in
Freudian terms: the urges which can, or
cannot, be blocked or rechanneled by
the conscious Ego.
Electrical switching of circuits. Remember
that a circuit is a linear structure that
must complete a loop: electricity doesn't
do this in nature -- it's an
accomplishment of human engineering.
Circuits for power and circuits for
telegraphs and telephones.
"The computational brain...notion was
originally McCullo[ch]
and Pitts (1943) [University of Chicago,
as we'll see] that nervous activity
embeds a logical calculus...
"further explored ... by John von
Neumann (1948)...Alan Turing had
proposed in 1950 the famous "Turing
test [cognition can be tracked by
language facility]."
Turing earlier had shown that any
algorithm can be implemented on a
universal Turing machine, suggesting
that one can study properties of
algorithms independent of where they
are implemented.
Wilfrid Rall "Some historical notes" (from
Schwartz collection)
McCulloch and Pitts (1943) "A logical
calculus of the ideas immanent in
nervous activity," written while both
were at the U of C (both moved to MIT
in the years after WWII); both were
during the war in the mathematical
biology community led by Nicholas
Rashevsky at the U of C.
Pitts was a grad student in mathematical
biophysics,* also worked with Rudolph
Carnap in philosophy. At MIT he worked
with Wiener (he never finished his PhD).
Work during the early 1940s including
"parallel interconnected neurons,
dynamics of simple circuits, the general
neural net, fluctuations of threshold..."
.."a statistical consequence of the
logical calculus of nervous nets (Dec
Pitts a graduate student? That's what Rall
says. But Jerry Lettvin, his best friend at
the time, says Pitts was a perpetual
outsider befriended by brilliant faculty,
like Carnap and McCulloch; and that
Pitts was 18 years old, and had been
kicked out by his family. (see Talking
Nets, Anderson and Rosenfeld, MIT
Press, 1998, p 3ff).
After the war, many physicists switched to
biophysics. "One interesting and
important topic presented in [a course in
the late 1940s] was the concept of
nonequilibrium steady states...."
September 1948: Hixon Symposium at
Cal Tech:
Major addresses by John von Neumann
on the digital computer (which he had
been designing);
Warren McCulloch (of whom we have
Karl Lashley: "The problem of Serial
Order in Behavior".
Karl Lashley: "The problem of Serial Order in
Behavior". "The problems raised by the
organization of language seem to me to be
characteristic of almost all other cerebral
activity." To wit: spotlight on the complex
organization of behavior. This complex
behavior requires advance planning, of a
hierarchical sort; it cannot be analyzed as a
series of acts, each caused by the
environment and the previous act....
Lashley: "Attempts to express cerebral
function in terms of the concepts of the
reflex arc, or of associated chains of
neurons, seem to me doomed to failure
because they start with the assumption
of a static nervous system. Every bit of
evidence available indicated a dynamic,
constantly active system, or, rather, a
composite of many interacting
Summer 1956 Dartmouth conference
Early lights in computer science:
пЃ® Marvin Minsky
пЃ® John McCarthy: LISP, MIT then
Stanford AI labs
пЃ® Allen Newell
пЃ® Herbert Simon-- Newell and Simon
wrote Logic Theorist (1955).
Newell and Simon strong functionalists:
With Shaw, they wrote in 1964:
We do not believe that this functional
equivalence between brains and computers
implies any structural equivalence at a more
minute anatomical level...Discovering what
neural mechaisms realize these information
processing functions in the human brain is a
task for another level of theory construction.
Our theory is a theory of the informaiton
processes involved in problem-solving and
not a theory of neural or elctronic mechaisms
for information processing.
Influential writing of Konrad Lorenz and
Niko Tinbergen coming out of Europe
on ethology: biological determinants of
animal behavior.
Discovery of critical periods in animal
development. (This influence is palpable
in Chomsky's review of Skinner's Verbal
Behavior, in Language 1956)
Getting ahead of ourselves, to Newell and
Simon's view:
all intelligent systems involve physical
symbol systems: a control, a memory, a
set of operations, input and output.
Involves production systems -- an
operation which is carried out if a
certain specific condition is met.
"Programs consist of long sequences of
such production systems operations on
the data base." (Gardner).
September 11 1956:
MIT Symposium on Information Theory
Alan Newell and Herbert Simon "Logic
Theory Machine" (proof generator)
Noam Chomsky "Three Models of
George Miller: Magic number 7 plus or
minus 1.
Newell and Simon wrote,
One can date the change roughly from
1956: in psychology, by the appearance
of Bruner, Goodnow, and Austin's Study
of Thnking and Miller's "Magical number
seven"; in linguistics, by Noam
Chomsky's "Three models of language";
and in computer science, by our own
paper on the Logical Theory Machine.
Also in this period:
von Neumann's (posthumous, 1958) The
Computer and the Brain
Major influence of Noam Chomsky
starting in the 1960s: graduate program
begins in 1962 at MIT in linguistics, with
Chomsky and Morris Halle.
Rapid growth of transformational syntax
and phonology:
1965 Aspects of the Theory of Syntax
1968 Sound Pattern of English
1965 presented what Chomsky called the
Standard Theory -- the Aspects model -which many took to be a statement
about semantics:
Semantic interpretation
Deep Structure
Phrase structure rules
Surface structure
The Standard/Aspects model
Two conceptions of what doing grammar
is (Huck and Goldsmith 1995):
Mediationist view: Grammar is the
component that links the order of words
to the logical form, and the study of
grammar is the decoding of that
translation system.
Distributionist view: The study of grammar
reveals the principles governing where
the morphemes of a language may
This led to a major split in the area of
syntax, pitting Chomsky and many
students at MIT against George Lakoff,
Haj Ross, Paul Postal, and Jim
When the dust had settled, all five were
doing different things -- roughly
Chomsky did little new syntax between
1967 and 1977, then developed the
principles and parameters/ Government
and Binding approach (first, the Pisa
Jim McCawley continued his work on
logic and syntax.
Haj Ross worked on freezes and poetry;
in 1985 left MIT.
Paul Postal developed Relational
Grammar, Arc-Pair Grammar
George Lakoff developed Cognitive ...
George Lakoff developed Cognitive
grammar; heavily involved with studies
of metaphor and constrution grammar.
(See his description in interview in Huck
and Goldsmith 1995.)
But in the messages that generative
grammar sent to the world were:
1. The real goal is not good grammars of
languages, but explanatory adequacy,
i.e., explanations of particular
languages based on principles that are
intended to be truths about all
languages (=Language).
2. Formal expression was crucial; to
quote Bacon, truth comes more easily
from error than confusion.
3. Deep insights will come from analyses
where surface, or apparent, complexity
is decomposed into a series of ordered
modifications (=derivation), which are
the effects of a series of ordered rules.
4. There is no discovery procedure, no
algorithm that takes data in and sends
out a grammar; rather, there is an
evaluation measure...
In Syntactic Structures, Chomsky
sketched three positions:
1. Data
2. Data
3. Data
Grammar 1
Grammar 2
G1 > G2
These 3 positions demand successively
less of Universal Grammar, but
Chomsky said only the 3rd was
practically doable.
Thus he said UG could assign a
complexity measure (or an evaluation
metric), and a grammar with less
complexity is better than a grammar
with higher complexity.
It was clear that no learning theory
conceivable in the 1950s could learn a
generative transformational grammar.
But if our innate schema contains the
important structure, then learning is less
Размер файла
188 Кб
Пожаловаться на содержимое документа