Symbolic
vs Connectionist
A
symbolic model is based on concrete algorithm for the situation. A
symbolic model may have mapped this task to a number of if-then
statements which tell the robot how to act given a situation e.g. If
there is an enemy ahead with a better gun, then run away. In a
symbolic model, the robot explicitly tries to figure out what next to
do based on the determined algorithm.
On
the other hand, if we use the connectionist model to approach this
task, we would have a lot of connection weights between the layers
which would determine the actions that we take and learning can take
place by reinforcement based adjustment of these weights and this can
happen without any explicit algorithm to play quake.
I
think that the connectionist model more closely resembles the
processes in the mind because
-
There is no definite optimal way (concrete algorithm) to do the
things that we do in our daily life.
-
the neurons form the basic components of the brain and when we gain
implicit knowledge, it is in the form of connections between them
(according to ‘My Brilliant Brain’) and this closely resembles
the mechanism by learning by adjusting weights in connectivist
model.
Also,
if a robot is to become an expert in a field, since experts use
implicit representations and intutions which are better modelled by
the connectionist model, using a connectionist model may be more apt
for this scenario.
A Robot with a mind but without a body is essentially like a Blackbox which takes input and returns an output without any realisation of the external environment or the physical constraints.
In
a disembodied/disjoint system, the brain is essentially a function of
its own parameters however for learning, an interaction between the
mind and body is almost essential. Various aspects of the cognition
such as thoughts, ideas and concepts are indirectly influenced by the
body.
Any
implemented, interactive system is embedded in the world in at least
three ways: temporally, physically and socially [R. Picard. Affective
Computing. MIT Press, Cambridge, MA, 1998 ]. It is observed that a
robot while trying to learn instructions is dependent on time which
it utilizes, the space which it occupies and the objects it
physically interacts with.
Some
examples :
Space
utilization
1.
In order to adjust with the difficulty level of the game, the Robot
can switch to a different position of its fingers in an orientation
that brings out more swift movements.
2.
A Robot can adjust his orientation relative to the position of the
Laptop or for that matter a keypad.
But
when is this possible????
Only
when there is an external medium( a Body with hands and fingers in
this case)which interacts with the space around and helps in
improving his actions.
Time
utilization
1.
Perceiving relative moments of objects and reacting differently in
various situations requires a ‘sense of time’ which can be not be
alone perceived by the mind.
Viewing
from the perspective of neuroscience, the phantom sensations
perceived by amputated humans how integral body is to mind. This
shows that the learning may go beyond the mind and it is the whole
system (Mind and Body) which learns.
Though
the ‘unused’ feedback and sensory data perceived by the mind
through this external body in which it is embedded may not play a
‘direct’ role in ‘task execution’ however they definitely
contribute to attaining expertise over the task by providing an
‘intrinsic feedback’.
Our
thought process is not an abstract entity that is distinct from
perception and action. It exists conjointly with the material aspects
of the physical realities surrounding us. The instructions provided
to the Robot form a part of the ‘explicit memory’ in it. However
while trying to utilize space and time, the Robot is essentially
learning instructions ( of which he is not consciously aware of )
implicitly. These adjustments ( stored in his implicit memory )
complement his external input and help him master the game.
Thus,
it can be concluded that apart from the basic input given to a robot
to perform the task, following two aspects of cognition play a vital
role in learning :
1.
Embodiment with the outer space and time through a materialistic
entity.
2.
Formation of Mental Representations and Patters( in chunks ) stored
in the implicit memory that help him quickly identify the best
possible ‘move’