Download Cognitive Informatics: Towards Future Generation Computers that

Cognitive Informatics:
Towards Future Generation Computers that Think and Feel
Yingxu Wang, Prof, PhD, PEng, FWIF, SMIEEE, MACM
International Center for Cognitive Informatics (ICfCI)
Dept. of Electrical and Computer Engineering
Schulich Schools of Engineering, University of Calgary
2500 University Drive, NW, Calgary, Alberta, Canada T2N 1N4
Tel: (403) 220 6141, Fax: (403) 282 6855
Email: yingxu(4,ucalgary. ca
Abstract
Computing systems can be categorized as imperative and
This keynote lecture presents a set of the latest advances in
Cognitive Informatics (CI) that leads to the design and
implementation of future generation computers known as
the cognitive computers that are capable of thinking and
feeling. The theory and philosophy behind the next
generation computers and computing technologies are CI.
The theoretical framework of CI may be classified as an
entire set of cognitive functions and processes of the brain
and an enriched set of descriptive mathematics. the
cognitive computers are created for cognitive and
perceptible concept/knowledge processing based on
contemporary mathematics such as Concept Algebra, RealTime Process Algebra, and System Algebra. Because the
cognitive computers implement the fundamental cognitive
processes of the natural intelligence such as the learning,
thinking, formal inference, and perception processes, they
are novel information processing systems that think and
feel. The cognitive computers are centered by the parallel
inference engine and perception engine that implement
autonomic learning/reasoning and perception mechanisms
based on descriptive mathematics.
Keywords:
Cognitive informatics. desc
m,
Keywords:
coiteifrmatics
t
criptive,matem
v
computer architectures, autonomic computing, agent
systems, cognitive models, cognitive processes, natural
intelligence, neural informatics.
1. Introduction
Conventional machines are developed to extend human
physical capability, while modern information processing
machines, such as computers, communication networks,
and robots, are created for extending human intelligence,
memory, and the capacity for information processing.
Proc. 5th IEEE Int. Conf. on Cognitive Informatics (ICCI'06)
Y.Y. Yao, Z.Z. Shi, Y. Wang, and W. Kinsner (Eds.)
1-4244-0475-4/06/$20.OO @)2006 IEEE3
autonomic systems. The former is a traditional and passive
system based on stored-program controlled behaviors. The
latter does not rely on instructive and procedural
information, but is dependent on goal-driven formal
inferences and autonomic perceptions.
Definition 1. An imperative computing system is a passive
system that implements deterministic, context-free, and
stored-program controlled behaviors.
Definition 2. An autonomic computing system is an active
and intelligent system that autonomously carries out robotic
and interactive applications based on goal-driven inferences
and perceptions mechanisms.
Recent advances in Cognitive Informatics (CI) reveal an
entire set of cognitive functions of the brain [4, 12, 17] and
their cognitive process models. The fundamental research
in CI also creates an enriched set of descriptive
mathematics known as Concept Algebra (CA) [12], RealTime Process Algebra (RTPA) [3], and System Algebra
(SA) [ 1], for dealing with the extremely complicated
objects
and neural
problems
in cognitive
natural
computing,
informatics,
knowledge
manipulation,
intelligence,
and software engineering.
2. Theoretical Foundations of Cognitive Computers
The theory and philosophy behind the next generation
computers and computing technologies are CI [4, 10]. The
theoretical framework of CI may be classified as an entire
stoconivfutosofheban[7adannrhd
set of contemporary descriptive mathematics [3, 11, 12].
2.1 Cognitive Informaties (CI)
the neuron level and their abstract mathematical models
[16, 18].
The structure of the theoretical framework of CI [18] is
illustrated in Fig. 1, which covers ten fundamental theories
such as the Information-Matter-Energy (IME) model [4],
the Layered Reference Model of the Brain (LRMB) [17],
the Object-Attribute-Relation (OAR) model of information
representation in the brain [15], the cognitive informatics
model of the brain [16], Natural Intelligence (NI) [2, 18],
Neural informatics (Nel) [18], CI laws of software [13], the
mechanism of human perception processes [8], the
cognitive processes of formal inferences [9], and the formal
knowledge system [14].
Nel is a branch of CI, where memory is recognized as the
foundation and platform of any natural or artificial
intelligence [16].
Definition 4. The cognitive models of memory (CMM)
states that the architecture of human memory is parallel
configured by the Sensory Buffer Memory (SBM), ShortTerm Memory (STM), Long-Term Memory (LTM), and
Action-Buffer Memory (ABM), i.e.:
CMMA
The Theoretical Framework of Cognitive lnformatics (CI)
cl
LTM
Applications(A)
Theories (T)
T2
RB
Tohel
lLmodel
T3
The OAR
model
-
T4
the brain
Csoflaws
algCoeceb(
softvare ILalgebraCA)
T7
ml
Future]generation
Computers
T8
Perception
NE
RTPA
A2
Capacity of human
memory
frontal lobe, i supplementary
motor area in the frontal lobe,
t
and procedural memory in cerebellum [16].
processes
T9
-
T5
Natural
intelligence
Al
M3
s
computing
TIO
human memory onto the physiological organs in the brain
reveal a set of fundamental mechanisms of Nel. The OAR
model provides a generic description of information/
knowledge representation in the brain [15, 16].
A4
Cogrntive proLgeerties
The knowledge
system
A5
Simulation of
cognitive behaviors
A8
Deductve emantics
of softxmre
A7
the
The CMM model and the mapping of the four types of
Autononiic
T6
AS
cortex
A3
System(SA)algebra
processes
Neural
informatics
|Cognitivecomplexity
of software
[16].
The major organ that accommodates memories in the brain
is the cerebrum or the cerebral cortex. In particular, the
association and premotor cortex in the frontal lobe, the
in the motor
cortexprimary
frontal lobe, in
visual
temporal
sensory lobe,
cortex in lobe,
the occipital
Inference
CI model of
(1
ABM
where the ABM is newly identified by Wang
. The ~ahematics
Descriptive
=/
forpatclr
Ti
The
mlM
SBM
STM
A6
Agent
Clfoundations of
software engineering
systemns
The theories of CI and Nel explain a number of important
phenomena in the study of natural intelligence.
Some
areand
as
enlightening conclusions derived in CINel
follows:
(a) LTM establishment is a subconscious process;
Figure 1. The Theoretical Framework of Cl
(b) The LTM is established during sleeping;
Three types of new mathematics, Concept Algebra (CA)
[12], Real-Time Process Algebra (RTPA) [5], and System
Algebra (SA) [11], are created recently to enable rigorous
treatment of knowledge representation and manipulation in
a formal and coherent framework. The three new structures
of descriptive mathematics have extended the abstract
objects under study in mathematics to a higher level, i.e.
concepts, behavioral processes, and systems. A wide range
of applications of the descriptive mathematics in the
context of CI has been identified [14].
(c)
(d)
The major mechanism for LTM establishment is
by sleeping;
The general acquisition cycle of LTM equals to
Xor longer than 24 hours;
o
2.2 Neural Informatics (NeI)
t
2
(e)
The mechanism of LTM establishment is to
update the entire memory of information
represented as an OAR model in the brain;
(f)
Eye movement and dreams play an important
role in LTM creation.
Defilnition 3. Neural Informatics (NeI) is a new
The latest development in CI and NeI has led to the
determination of the magnificent and expected capacity of
human memory as described in [6].
interdisciplinary enquiry of the biological and physiological
representation of information and knowledge in the brain at
4
2.3 Natural Intelligence (NI)
3.1 Von Neumann Machines
Software and computer system are recognized as a subset
of intelligent behaviors of human beings described by
programmed instructive information [5, 7, 17]. The
relationship between Artificial Intelligence and Natural
Intelligence can be described by the following theorem [14,
18].
The key requirements for implementing a stored-program
controlled computer are the generalization of common
computing architectures and the computer is able to
interpret the data loaded in memory as computing
instructions. These are the essences of stored-program
computers known as the von Neumann architecture [1].
Von Neumann elicited the five fundamental and essential
components to implement general-purpose programmable
digital computers in order to embody the concept of storedprogram-controlled computers.
Theorem 1. The law of compatible intelligent capability
states that artificial intelligence (AI) is a subset of natural
intelligence (NI), i.e.:
Al c NI
Definition 6. A von Neumann Architecture (VNA) of
computers is a 5-tuple that consists of five components: (a)
the arithmetic-logic unit (ALU), (b) the control unit (CU)
with a program counter (PC), (c) a memory (M), (d) a set
of inputloutput (II0) devices, and (e) a bus (B) that
provides the data path between these components, i.e.:
VNA ^ (ALU, C, M, 0, B)
(3)
(2)
Theorem 1 indicates that Al is always a subset of NI.
Therefore one should not expect a computer or a software
system to solve a problem where human cannot do. That is,
no Al or computing system may be designed and/or
implemented for a given problem where there is no solution
being known by human being.
Definition 7. Conventional computers with VNA are aimed
at stored-program-controlled data processing based on
mathematical logic and Boolean algebra.
Almost all modem disciplines of science and engineering
deal with information and knowledge. According to CI
theories, cognitive information may be classified into four
categories known as knowledge, behavior, experience, and
skillsriasshownin Tbled
1.
A VNA computer is centric by the bus and characterized by
the all purpose memory for both data and instructions. A
VNA machine is an enhanced Turing machine (TM), where
the power and functionality of all components of TM
including the control unit (with wired instructions), the tape
(memory), and the head of I/O, are greatly enhanced and
extended with more powerful instructions and 1/0 capacity.
Table 1. Types of Cognitive Information
Type Abs. Concept
of
Type of Output
Ways of
Acquisition
Abs. Concept Emp. Action
Direct or indirect
Knowledge Behavior
Input Emp. Action Experience
Skill
Direct only
3.2 Cognitive Machines
Definition 5. The taxonomy of cognitive information is
determined by its types of inputs and outputs to and from
the brain during learning and information processing,
where both inputs and outputs can be either abstract
information (concept) or empirical information (actions).
Definition 8. A Wang Architecture (WA) of computers,
known as a Cognitive Machine as shown in Fig. 2, is a
parallel structure encompassing an Inference Engine (IE)
and a Perception Engine (PE), i.e.:
WA
It is noteworthy that the approaches to acquire
and
are
knowledge/behaviors
experience/skills
fundamentally different. The former may be obtained either
directly based on hands-on activities or indirectly by
reading, while the latter can never be acquired indirectly.
)
(IE 1 PE)
KMU
BMU
EMU
SMU
The Knowledge Manipulation Unit
The Behavior Manipulation Unit
The Experience Manipulation Unit
The Skill Manipulation Unit
( BPU // The Behavior Perception Unit
EPU // The Experience Perception Unit
(4)
)
3. Architecture of Cognitive Computers
As shown in Fig. 2 and Eq. 4, WA computers are not
centered by a CPU for data manipulation as the VNA
computers do. The WA computers are centered by the
concurrent JE and PE for cognitive learning and autonomic
perception based on abstract concept inferences and
The theory and philosophy behind the next generation
computers and computing technologies are CI and the
contemporary descriptive mathematics [10]. It is commonly
believed that the future-generation cognitive computers will
adopt non-von Neumann architectures.
5
acquisition, particularly behaviors, experience, and skills
rather than only focusing on knowledge.
empirical stimulus perception. The IE is designed for
concept/knowledge manipulation according to concept
algebra [12], particularly the 9 concept operations for
knowledge acquisition, creation, and manipulation. The PE
is designed for feeling and perception processing according
to RTPA [3] and the formally described cognitive process
models of the perception layers as defined in the LRMB
model [17].
Corollary 1. All the four categories of information can be
acquired directly by an individual.
Corollary 2. Knowledge and behaviors can be learnt
indirectly by inputting abstract information; while
experience and skills must be learnt directly by hands-on or
r----------------------------------------------_ __ __ __ ----------empirical actions.
~~~~~IE
l
LTM
LTM
3_
ABMulrlW ;
ABM
CM = IE
Stimulii.
3BM
LTM
KMU
BMU
EMU
ABM
LTM
Knoledge -- I
Behaviors
Experience-- |
SMU
ABM
Skills
-
Behaviors
---
of pedagogy.
CI lays anBased
important
for
The
above
theoryand
on thefoundation
fundamental
learning
theories
work, the IE and PE of cognitive computers working as a
virtual brain can be implemented on WA-based cognitive
computers and be simulated on VNA-based conventional
computers.
fftractions
PE
BPU
ABM
LTM
EPU
The Cognitive Machine (CM)
lThe
-*}SB
4. Conclusions
This keynote has revealed that recent advances in Cl have
prepared a reach set of theories, mathematical means, and
discoveries toward the creation and implementation of next
generation cognitive and intelligent computers with nonvon-Neumann architectures and novel inference and
perception engines. The cognitive computers have been
characterized as autonomic and perceptive concept/
knowledge processors rather than imperative data
processors. The new generation computers are founded on
the basis of contemporary descriptive mathematics and
theories developed in CI.
Exience--
Cognitive Machine (CM)
_ __ -_
_
Figure 2. The architecture of a cognitive machine
Definition 9. Cognitive computers with WA are aimed at
cognitive and perceptive concept/knowledge processing
based on contemporary descriptive mathematics, i.e.
Concept Algebra (CA), Real-Time Process Algebra
(RTPA), and System Algebra (SA).
As that of mathematical logic and Boolean algebra are the
mathematical foundations of VNA computers. The
mathematical foundations of WA computers are based on
contemporary descriptive mathematics. As described in the
LRMB reference model [17], since all the 37 fundamental
cognitive processes of human brains can be formally
described in CA and RTPA [3, 12]. In other words, they are
executable and simulatable by the WA-based cognitive
computers.
Acknowledgement
The author would like to acknowledge the Natural Science
and Engineering Council of Canada (NSERC) for its
support to this work. The author would like to thank the
invitation of the ICCI'06 program committee for this
keynote.
3.3 Theory ofLearning and Information Acquisition
References
According to Table 1, the following important conclusions
on information manipulation and learning for both human
and machine systems can be derived.
[1] Von Neumann, J. (1946), The Principles of LargeScale Computing Machines, reprinted in Annals of
History ofComputers, Vol. 3, No. 3, pp. 263-273.
Theorem 2. The principle of information acquisition states
that there are four sufficient categories of learning known
as those of knowledge, behaviors, experience, and skills.
[2] Wang, Y. (2002), Keynote Speech: On Cognitive
Informatics, Proc. 1st IEEE International Conference
on Cognitive Informatics (ICCI'02), Calgary,
Canada, IEEE CS Press, August, pp.34-42.
Theorem 2 indicates that learning theories and their
implementation in autonomic and intelligent systems
should study all four categories of cognitive information
[3]
6
Wang, Y. (2002), The Real-Time Process Algebra
(RTPA), The International Journal of Annals of
Software Engineering, Vol.14, USA, pp. 235-274.
[4]
Wang, Y. (2003), On Cognitive Informatics, Brain
and Mind: A Transdisciplinary Journal of
Neuroscience and Neurophilosophy, Vol.4, No.2,
pp.151-167.
[11] Wang, Y. (2006), On Abstract Systems and System
Algebra, Proc. 5th IEEE International Conference on
Cognitive Inform atics (ICCI'06), IEEE CS Press,
Beijing, China, July.
[5]
Wang, Y. (2003), Keynote Speech: Cognitive
Informatics Models of Software Agent Systems and
Autonomic
Computing,
Proc.
International
Conference on Agent-Based Technologies and
Systems (ATS'03), Univ. of Calgary Press, Calgary,
Canada, August, pp. 25.
[12] Wang, Y. (2006), On Concept Algebra and
Knowledge Representation, Proc. 5th IEEE
International Conference on Cognitive Informatics
(ICCI'06), IEEE CS Press, Beijing, China, July.
[6] Wang, Y., D. Liu, and Y. Wang (2003), Discovering
the Capacity of Human Memory, Brain and Mind: A
Transdisciplinary Journal of Neuroscience and
Neurophilosophy, Vol.4, No.2, pp. 189-198.
[7] Wang, Y. (2004), Keynote Speech: On Autonomic
Computing and Cognitive Processes, Proc. 3rd IEEE
International Conference on Cognitive Informatics
on Systems, Man, and Cybernetics (Part C), Vol. 36,
No.2, March, pp. 161-171.
[14] Wang, Y. (2006), Software Engineering Foundations.
[13] Wang, Y. (2006), On the Informatics Laws and
Deductive Semantics of Software, IEEE Transactions
A Transdisciplinary and Rigorous Perspective, CRC
Book Series in Software Engineering, Vol. 2, CRC
Press, USA.
[15] Wang, Y. (2006), The OAR Model for Knowledge
Representation, Proc. 2006 IEEE Canadian
Conference
on Electrical
and Computer
Engineering (CCECE'06), Ottawa, Canada, May, pp.
1696-1699.
[16] Wang, Y. and Y. Wang (2006), On Cognitive
Informatics
Models of the Brain, IEEE
Transactions on Systems, Man, and Cybernetics
(Part C), Vol. 36, No. 2, March, pp. 203-207.
[17] Wang, Y., Y. Wang, S. Patel, and D. Patel (2006), A
(ICCI'04), Victoria, Canada, IEEE CS Press, August,
[8]
pp. 3-4.
Wang, Y. (2005), On the Cognitive Processes of
Human Perceptions, Proc. 4th IEEE International
Conference on Cognitive Informatics (ICCI'05), IEEE
CS Press, Irvin, CA, USA, August, pp. 203-211.
Wang, Y. (2005), The Cognitive Processes of
Abstraction and Formal Inferences, Proc. 4th IEEE
International Conference on Cognitive Informatics
(ICCIO'0), IEEE CS Press, Irvin, Califomia, USA,
Layered Reference Model of the Brain (LRMB),
August, pp. 18-26.
IEEE Transactions on Systems, Man, and Cybernetics
[10] Wang, Y. (2006), Invited Plenary Talk: Cognitive
(Part C), Vol. 36, No. 2, March, pp.124-133.
Informatics and Contemporary Mathematics for
Y.
The
Framework of
Knowledge Representation and Manipulation, Proc.
Te
1st International Conference on Rough Set and
Natioal Intelligence
Cognitive Informatics,
Jnall of
Informatics and Natural 10-22.
Notes on
(RST'06, Lecture
LetureNote
onCognitive
KnowledgeKnowedgeTecholog
Technology (RSKT'06),
IGP, Hershey, PA, USA, Jan., pp.
Artificial Intelligence, LNAI 4062, Springer,
Chongqing, China, July, pp. 69-78.
[9]
[18]oWang,
(IJCiNi),
7
(2007),
Theoretical