122 views9 pages
1 Feb 2011
Artificial Intelligence, Connectionism
Artificial intelligence: manufactured cognitive agency
Computation: first, info is encoded into propositions with logical syntactic
structure. Then, it is manipulated/processed inferentially & in a structure-
sensitive manner.
Computers: things that propositionally represent information using a logical
syntactic structure.
Interpreted automatic formal system: haugelands definition
Cognition as computation: around for a long time as an idea, central to
scientific revolution, necessary to explain theory of mind & to fulfill the natural
Scientific Revolution: opposite of Aristotleian point of view
Conformity Theory: (Aristotle) how the MIND works (see below) “patterns of
Parallel Decomposition: The idea that a complex machine can be broken down
into a series of smaller & smaller algorithms
Interpreting: making sense of decoding1 string of x, by translating it to
another, OR understanding
Implementation: functions vertically- between higher & lower levels- within itself
(a computer can be implemented entirely algorithmically)
Use: horizontally, using data from the rest of the world (often things are used
Equivocation: for example, the valid argument “nothing is better than long life
and happiness. A pbj sandwich is better than nothing. Therefore, a pbj sandwich
is better than long life & happiness
Design stance: assuming that emulating the structure of something will allow us
to understand or recreate its function. (see Engineering Endrun under ISSUES)
Unlock document

This preview shows pages 1-3 of the document.
Unlock all 9 pages and 3 million more documents.

Already have an account? Log in
FORMAL SYSTEMS: medium-independent, self-contained
-require positive & reliable read/write cycles/techniques
Writing: manipulating a token
Reading: re-identifying a tokens type
Reliable: succeeds frequently
Positive: succeeds absolutely, totally, without qualification
Finite playability: players are not infinite beings
Players must:
Have a repertoire of primitive operations they can perform, positively &
reliably & finally.
be able to determine the legality of a move
either PRODUCE a legal move, or PROVE that there are none.
DIGITAL SYSTEMS: sets of positive & reliable techniques/methods/devices for
producing & reidentifying tokens themselves, or configurations of tokens, from
a pre-specified set of types.
Tokens: inherently meaningless objects which are used as symbols
Meaning: the relationship between tokens and the world
Formal Properties: properties between tokens (manipulations?)
Manipulations: actions done to tokens… they can be moved around, added or
removed, altered or replaced
Types: functions that tokens can perform. Determined by rules. NOT
PHYSICALcan apply to ANY token as long as it is
-agreed upon ,
- reliably physically identifiable
-consistently/reliably manipulable
Formal Equivalence:
-for each distinct position of tokens in 1 system, there is 1 corresponding
position of tokens
-whenever any move is legal in 1 system, the corresponding move in
corresponding systems are also legal
Unlock document

This preview shows pages 1-3 of the document.
Unlock all 9 pages and 3 million more documents.

Already have an account? Log in
-multiple realizability: the idea that the same mental property, state or
event can be implemented by any number of different physical properties,
states or events (goes with the idea that formal systems are medium-
-Identify tokens by their physical properties
-Indicate conditions under which we can carry out manipulations
-Indicate starting positions, legal manipulations, given legally arisen
configurations of tokens
Philosophical Zombie: a hypothetical being that is indistinguishable from a
normal human being, in appearance and in action. Its only difference being that it
lacks qualia: conscious experience or sentience.
Connectionism: how are thoughts shared using propositional representations?
We assume that cognition involves encoding info syntactically , referentially &
with propositionsbut this is just because we verbalize thoughts.
Alternative to GOFAI, modeling cognition on plastic neural networks
Neural networks: series of nodes in which connections are assigned different
strengths and frequencies of signals.
Microprocessors: nodes & connections between them, with a symmetrical,
differential weighting system. (stronger connections allow for a great ease of
transfer, BUT tons o signal through a weakly weighted variable can override
signal through a heavily weighted connection)
Back-propagation of Error: nodes can alter each others paths of activity… but
how can they LEARN things about the worldhomunculus-free?
Assume 3 layers of nodes: input, hidden, output. When something occurs,
Error value is calculated by subtracting performance from target value,
Then, nodes responsibilities (in percentage of blame) are assigned partially &
probabilistically. Then, connection weights are SLIGHTLY altered accordingly
(weakening connections to nodes deemed blameworthy, etc) This process occurs
again and again until the system accurately identifies mines and rocks alternately,
for example
Unlock document

This preview shows pages 1-3 of the document.
Unlock all 9 pages and 3 million more documents.

Already have an account? Log in

Get OneClass Grade+

Unlimited access to all notes and study guides.

Grade+All Inclusive
$10 USD/m
You will be charged $120 USD upfront and auto renewed at the end of each cycle. You may cancel anytime under Payment Settings. For more information, see our Terms and Privacy.
Payments are encrypted using 256-bit SSL. Powered by Stripe.