> From: Counter, SRM <srmc196@soton.ac.uk>
>
> Please could you list the discussion topics that need to be
> covered from Chapters 2,3+4. I have read both chapters 1+2
> and it is damn long-going as I am reading and making notes
> on everything. Is this right or are their particular topics
> that we should be making notes on in each chapter?
Here are the Lecture Notes for Chapter 1. I'll
post them for the other Chapters too. They'll also
be on the Web. They should guide you about what's
important in each chapter.
[Chapter 1 Lecture Notes]
WHAT IS COGNITIVE SCIENCE?
Cognitive Psychology
Neuropsychology
Robotics/Artificial Intelligence
Linguistics
Philosophy of Mind
WHAT IS COGNITIVE PSYCHOLOGY TRYING TO EXPLAIN?
The mind.
HOW CAN WE EXPLAIN THE MIND?
Not by introspecting what is going on in our own minds
(that's folk psychology)
Not by describing our behaviour
(that's behaviourism)
WHAT IS IT TO EXPLAIN SOMETHING?
It is to find a causal model with which the evidence
can be predicted and understood
WHAT IS IT TO UNDERSTAND SOMETHING?
It is to have no further questions to ask about it!
WHAT IS IT ABOUT OUR MINDS THAT WE NEED TO PREDICT AND EXPLAIN?
(1) what it feels like to have a mind
How we feel can be predicted and explained (in part) by studying
our brainsU electrical and chemical activity and the effects of
events or drugs or injuries on our brains. (Cognitive Psychology
does not yet have much
to say about feelings.)
(2) what we can do with our minds
Explaining what we can do is more complicated: What we can do
is what our brains can do. But we do not yet know enough about
the brain to understand how it can do what it does. And studying
its electrical or chemical activity, or the effects of injury has
not yet answered any of the hard how questions. The only way
to answer those questions is
to by modeling.
WHAT IS MODELING?
If there is a system (human or animal or artificial) that has
certain capacities that we think we can explain, then we should
be able to model the system, as well as its capacities, by
simulating them on a computer. This does not mean that the
system we are explaining us a computer! A computer is just a
powerful way of testing causal explanations.
EXAMPLE: Suppose thermostats grew on trees instead of being
built by us. We would either have to build them or model them in
order to explain them.
Some discussion topics from Chapter 1:
WHAT IS A REPRESENTATION
The birth of Cognitive Science from Noam
Chomsky's famous review of BF Skinner's
book on language Verbal Behavior
What are symbols and what is computation?
What is Searle's Chinese Room Argument
What is the Symbol Grounding Problem?
What are "mental images"? What was the debate over whether or not they
exist about?
3 Levels of explanation:
Behavioural capacity
Algorithm
Implementation
What is a computer? (automata, Turing Machines)
Computation = rule-based symbol manipulation
An Algorithm is a rule for manipulating symbols
Symbols are just arbitrary shapes that we agree to interpret in a
certain way, for example:
cat mat Matt + 1 = egg mix
Symbols vs. Nets:
Symbol systems are better at:
language, logic, calculation, reasoning, problem-solving
Neural Nets are better at:
learning, pattern recognition + more brainlike
You should know what the following mean:
Connectionism (Neural Nets)
Nodes, activations, weights
Perceptrons
Multilayered nets
localist vs distributed networks
(PDP parallel distributed processing)
Hebb Rule
Gradient descent + Local Minima
XOR
Past tense learning
Hybrid Models
This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:51 GMT