June 25, 2012 12:45 PSP Book - 9in x 6in 09-Junichi-Takeno-c09
My Standpoint 183
Figure 9.9. Recursive retreat.
The author believes that the recursive retreat problem can be
somehow avoided by using infinite repetition (Fig. 9.10). The state
of becoming aware of the self is understandable, but it is impossible
to actually feel the self who retreats multiple times in the realm of
phenomenology. I believe that the self who retreats multiple times
is simply an imaginary object of symbolic logic created in the brain
because we are already handling the concept of infinity as a symbol.
The author, therefore, believes that it suffices if we can create a self
who becomes conscious of the self on a computer, and the rest is
considered to be the result of symbolic logic within the brain.
9.3.6 Quantum Consciousness
Quantum theory would provide an effective means of research if
we were to analyze the brain at the atomic level. There are many
things we can do, however, before we resort to quantum theory. I
believe that quantum theory is not necessary if one is interested in
elucidating the mechanism of consciousness.
June 25, 2012 12:45 PSP Book - 9in x 6in 09-Junichi-Takeno-c09
184 New Architecture of Robot Consciousness and the Robot Mind
Figure 9.10. Infinite repetition.
9.4 Computational Model of the Consciousness Module
This chapter describes the MoNAD, the consciousness module that
is the basis of the consciousness system.
9.4.1 Overview of Computation Model
A MoNAD is an artificial neural network with a slightly complex
structure that is built upon the concept model of consciousness
described in the previous chapter (Fig. 9.11). The first feature of
the MoNAD is that at least two closed loops (c1 and c2) exist and
both are connected to a common neuron group K called primary
representation. The second feature is the somatosensory nerves Y
extending in both common neuron group K and the closed loop c2.
The MoNAD has at least one input terminal Input L. Closed loop
c1 has two symbolic representations Z and Z
. Z is called cognitive
representation and Z
behavioral representation. A MoNAD learns
the neural networks so that the values of Z and Z
are equivalent.
June 25, 2012 12:45 PSP Book - 9in x 6in 09-Junichi-Takeno-c09
Computational Model of the Consciousness Module 185
Figure 9.11. Computational model of consciousness module.
When a behavioral representation Z
is decided, the MoNAD executes
behavior Y from output L, and the information X obtained from input
L is changed to make the value of Z coincide with that of Z
.ZandZ
also serve as I/O terminals (output H and input H, respectively) for
exchanging information with higher-level MoNADs. Information on
the current behavior is input again into K via somatosensory nerve
Y
. The MoNAD, thus, cognizes not only the external environment but
also the behavior of the self. The result of this cognition is reflected
on the information of Z, and this information represents both the
external environment and the behavior of the self.
9.4.2 MoNAD Functions
The MoNAD is designed to satisfy the consistency of cognition and
behavior, which is the author’s definition of consciousness. This
means that in the MoNAD information runs consistently through
the neural networks. In this model, the information flows through
Z
of the closed loop c1 and common neuron group K, and then
reaches Z. The value of Z is copied to Z
and is also input into the
input L of a higher-level MoNAD. Information from Y runs through
June 25, 2012 12:45 PSP Book - 9in x 6in 09-Junichi-Takeno-c09
186 New Architecture of Robot Consciousness and the Robot Mind
Y
of the closed loop c2 and K and then returns to Y. Information
basically circulates in a single MoNAD through K, Z, Z
and returns
to K, from which the information further flows through Y and Y
and then returns to K. Information X from input L is added to this
circulating information all the time. If the internal circulation of
information is maintained normally despite the input of incessantly
changing external information X, the MoNAD is said to be keeping
the “consistency of cognition and behavior.” This is because the
information flow of (Z
,X,Y
), K, Y decides the next behavior to be
taken and the information flow of (Z
,X,Y
), K, Z decides cognition.
Information from Y output from the terminal output L is also a new
order for the lower-level MoNADs to engage in a behavior because
Y changes the behavior of the lower-level MoNADs. When the value
of Z
is changed by a higher-level module, the MoNAD executes an
action from output L that will eventually make the value of Z coincide
with the value of Z
. The author calls this state, in which the value of
Z is equal to that of Z
, a state in which cognition and behavior are
consistent.
9.4.3 Investigating Husserl’s 10 Properties
Husserl enumerated 10 properties of consciousness. Does the
MoNAD proposed by the author satisfy them?
(1) The first-person property is the assertion of the existence
of the self based on mind–body monism. The representation
area of the MoNAD distinguishes between the self and others
and represents the behavior of both. This means that the
MoNAD recognizes the behavior of others based on the
behavior of the self. This is a sort of mind–body monism.
(2) Orientation means that consciousness is always directed
toward something. The MoNAD self-learns to behave, by
output L, to make cognitive representation Z coincide with
behavioral representation Z
. As such, the MoNAD satisfies
the second requirement of Husserl.
(3) This is the same as point 2. In addition, the duality of
consciousness is that the self becomes aware of the self. The
MoNAD is capable of cognizing the behavior of the self using
June 25, 2012 12:45 PSP Book - 9in x 6in 09-Junichi-Takeno-c09
Computational Model of the Consciousness Module 187
its somatic sensation Y
. The MoNAD is therefore able to
become aware of the behavior of the self by itself.
(4) Expectation: The MoNAD performs a behavior (output L)
according to point 2 such that the cognitive representation
Z agrees with the behavioral representation Z
.Thismeans
that the behavioral representation Z
expects the forthcom-
ing cognitive and behavioral state of the MoNAD.
(5) Conviction of the existence of things: The MoNAD always
calculates the cognitive representation Z using external in-
formation (input L) and internal somatosensory information
Y
, .i.e., Z results from the cognition of the external and inter-
nal environments. In the area of the MoNAD where external
information is directly processed, information X about the
external environment inevitably includes uncertainty but Z,
as a representation, always converges on a certain definite
value. The MoNAD can, thus, discern a positive existence in
the midst of uncertain information.
(6) The body is part of the self: The MoNAD cognizes the
behavior of the self using its somatic sensations. This means
that the MoNAD cognizes and is convinced that the present
behavior is being done by the self, i.e., the MoNAD is
confident that the behavior of the self is being done by the
body of the self.
(7) Distinction between the self and others: As stated in point
6, the MoNAD continually cognizes things by obtaining
somatosensory information and information about the
external environment via input L. This means that the
MoNAD always cognizes information about the behavior of
the self and the external environment. As such, I conclude
that the MoNAD distinguishes between the self and others.
(8) Property in which reason is related to emotion: With
the properties 5 through 7, the MoNAD has functions to
represent the internal and external environments. If this is
true, the MoNAD can represent information about emotions
arising in the body. Representation is shown on a Pleasant or
an Unpleasant MoNAD. This means that reason and emotion
can be achieved as a representation function using MoN-
ADs. Details will be described later. As Antonio Damasio,
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.