International Journal of Intelligence Science, 2011, 1, 8-16
doi:10.4236/ijis.2011.11002 Published Online July 2011 (http://www.SciRP.org/journal/ijis)
Copyright © 2011 SciRes. IJIS
Foundations of Intelligence Science*
Zhongzhi Shi
The Key Laborato ry of Intelligent Information Processing, Institute of Computing Technology,
Chinese Academy of Sciences, Beijing, China
E-mail: shizz@ics.ict.ac.cn
Received June 11, 20 1 1; revised July 20, 2011; acc epted July 25, 2011
Abstract
In order to make significant progress toward achievement of human level machine intelligence a paradigm
shift is needed. More specifically, the natural intelligence and artificial intelligence should be closely inter-
acted in Intelligence Science study, instead of separate from each other. In order to reach the paradigm, brain
science, cognitive science, artificial intelligence and others should cross-research together. Brain science ex-
plores the essence of brain, research on the principle and model of natural intelligence in molecular, cell and
behavior level. Cognitive science studies human mental activity, such as perception, learning, memory,
thinking, consciousness etc. Artificial intelligence attempts simulation, extension and expansion of human
intelligence using artificial methodology and technology. All together pursue to explore the mechanism and
principle of intelligence which is the engine of advanced science and technology. The paper will give the
definition of intelligence and discuss ten big issues of Intelligence Science. The conclusion and perspective
will be given in last section.
Keywords: Intelligence, Intelligence Science, Machine Intelligence
1. Introduction
Since 1956 artificial intelligence is formally found and
very impressive progress has been made in many areas
over the past years. Its achievements and techniques are
in the mainstream of computer science and at the core of
so many systems. For example, the computer beats the
world’s chess champ, commercial systems are exploiting
voice and speech capabilities, there are robots running
around the surface of Mars. In well-known TV quiz
show “Jeopardy” IBM super computer system Watson
beats the best of the two bit of human champion Ken
Jennings and Brad Rutter. But all th ese achievements are
not in the realm of human level machine intelligence.
Humans are the best example of human-level intelli-
gence. McCarthy declared the long-term goal of AI is
human level AI [1]. Recent work s in multiple disciplines
of cognitive and neuroscience motivate new computa-
tional approaches to achieving human level AI. In the
book On Intelligence, Hawkins proposed machine intel-
ligence meets neuroscience [2]. Granger presented a
framework for integrating the benefits of parallel neural
hardware with more serial and symbolic processing
which motivated by recent discoveries in neuroscience
[3]. Langley proposed a cognitive architecture ICARUS
which uses means-ends analysis to direct learning and
stores complex skills in a hierarch ical manner [4]. Sycara
proposed the multi-agent systems framework which one
develops distinct modules for different facets of an intel-
ligent system [5]. Cassimatis and his colleagues investi-
gate Polyscheme which is a cognitive architecture de-
signed to model and achieve human-level intelligen ce by
integrating multiple methods of representation, reasoning
and problem solving [6]. Based on the LIDA cognitive
architecture, Franklin et al. proposed an underlying
computational software framework for Artificial General
Intelligence [7].
To make significant progress toward achievement of
human level machine intelligence a paradigm shift is
needed. Artificial intellig ence shou ld chang e the research
paradigm and learn from natural intelligence. The inter-
disciplinary subject entitled Intelligence Science is pro-
moted. In 2002 the special Web site called Intelligence
Science has been appeared on Internet [8], which is con-
*This paper is supported by National Basic Research Programme
(2007CB311004), Key projects of National Natural Science Founda-
tion of China (No. 61035003, 60933004,), National Natural Science
Foundation of China (No. 61072085, 60970088, 60903141).
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
9
structed by Intelligence Science Lab of Institute of
Computing Technology, Chinese Academy of Sciences.
A special bibliography entitled Intelligence Science
written by author was published by Tsinghua University
Press in 2006 [9]. The book sh ows a framework of intel-
ligence science and points out research topics in related
subject. The English version of the book Intelligence
Science is published by World Scientific Publishers in
2011.
Intelligence Science is an interdisciplinary subject
which dedicates to joint research on basic theory and
technology of intelligence by brain science, cognitive
science, artificial intelligence and others. Brain science
explores the essence of brain, research on the principle
and model of natural intelligence in molecular, cell and
behavior level. Cognitive science studies human mental
activity, such as perception, learning, memory, thinking,
consciousness etc. In order to implement machine intel-
ligence, artificial intelligen ce attempts simulation, exten-
sion and expansion of human intelligence using artificial
methodology and technology.
Next section will define wh at is intellig en ce. The basic
issues of Intelligence Science are listed in Section 3. Fi-
nally, conclusion and perspective will be given.
2. What Is Intelligence
Intelligence is a very hot word. At the present it is a new
development trend to intellectualize technology, product,
equipment, such as intelligen t computer, intelligent robot,
intelligent database, intelligent management, intelligent
control, intelligent CAD, intelligent network, intelligent
engineering and so on [10]. Intelligence is most widely
studied in humans, but has also been observed in animals
and plants. Numerous definitions of intelligence have
been proposed with no consensus reached by scholars.
Intelligence has been defined in different ways, including
the abilities for abstract thought, understanding, commu-
nication, reasoning, learning, planning, emotional intel-
ligence and problem solving.
Legg and Hutter list 70 odd definitions of intelligence
from collective definitions, psychologist definitions and
AI researcher definitions [11]. They pull out commonly
occurring features and find that intelligence has follow-
ing features:
1) A property that an individual agent has as it inter-
acts with its environment or environments.
2) Is related to the agent’s ability to succeed or profit
with respect to some goal or objective.
3) Depends on how able to agent is to adapt to differ-
ent objectives and environments.
According to above features they give the informal
definition of intelligence as “Intelligence measures an
agent’s ability to achieve goals in a wide range of envi-
ronments.” [11]
We think that intelligence is a comprehensive ability
to use one’s existing knowledge or experience to adapt
new situations or solve new problems [10]. Since 1956,
traditional artificial intelligence adopts reasoning to do
problem solving. Most of expert systems are imple-
mented through deductive reasoning. Inductive reasoning
is applied in machine learning and data mining. Artificial
neural networks are massively parallel, adaptive, dy-
namical systems modeled on the general features of bio-
logical networks. Through trial and error, the network
literally teaches itself how to do the task. In terms of
situatedness, embodiment, intelligence and emergency
behavior-based artificial intelligence has built more
powerful autonomous mobile robots. It is a good idea
that symbolic, connectionist and behaviorist mechanism
are combined together to develop intelligent systems.
Humans have many remarkable capabilities: first the
capability to reason, converse and make rational deci-
sions in the real world of imprecision, uncertainly, in-
completeness of information; and second, the capability
to perform a wide variety of physical and mental tasks.
Machine intelligence should learn from natural intelli-
gence and more closely in teracted in Intelligen ce Science
study.
3. Basic Issues of Intelligence Science
3.1. How Do Brain Neural Circuits Work?
The brain is a collection of about 10 billion intercom-
nected neurons. Neurons are electrically excitable cells
in the nervous system that pro cess and transmit informa-
tion. A neuron’s dendrites’ tree is connected to a thou-
sand neighboring neurons [12]. When one of those neu-
rons fire, a positive or negative charge is received by one
of the dendrites. The strengths of all the received charges
are added together through the processes of spatial and
temporal summation. The aggregate input is then passed
to the soma (cell body). The soma and the enclosed nu-
cleus don’t play a significant role in the processing of
incoming and outgoing data. Their primary function is to
perform the continuous maintenance required to keep the
neuron functional. The output strength is unaffected by
the many divisions in the axon; it reaches each terminal
button with the same intensity it had at the axon hillock.
Each terminal button is connected to other neurons
across a small gap called synapse. The physical and
neurochemical characteristics of each synapse deter-
mines the strength and polarity of the new input signal.
This is where the brain is the most flexible, and the most
vulnerable. In molecular level neuron signal generation,
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
10
transmission and neurotransmitters are basic problems
attracted research scientists to engage investigation in
brain science.
One of the greatest challenges in neuroscience is to
determine how synaptic plasticity and learning and
memory are linked. Two broad classes of models of
synaptic plasticity can be described by Phenomenologi-
cal models and Biophysical models [13].
Phenomenological models are characterized by treat-
ing the process governing synaptic plasticity as a black
box. The black box takes in as input a set of variables,
and produces as output a change in synaptic efficacy. No
explicit modeling of the biochemistry and physiology
leading to synaptic plasticity is implemented. Two dif-
ferent classes of phenomenological models, rate based
and spike based , have been prop osed.
Biophysical models, in contrast to phenomenological
models, concentrate on modeling the biochemical and
physiological processes that lead to the induction and
expression of synaptic plasticity. However, since it is not
possible to implement precisely every portion of the
physiological and biochemical networks leading to syn-
aptic plasticity, even the biophysical models rely on
many simplifications and abstractions. Different cortical
regions, such as Hippocampus and Visual cortex have
somewhat dif f e rent forms of synaptic plastici t y .
Some important questions about human brain structure
and function remain a puzzle to us. What functions hap-
pen in the left-right and front-back division of cerebral
cortex and how to link each other? Many regions of the -
brain come together to form a dynamic and intricate bio-
logical structure that holds many puzzles for us to un-
ravel.
3.2. What Is Perceptual Representation and
Theory of Perception?
The perceptual systems are primarily visual, auditory and
kinesthetic, that is, pictures, sounds and feelings. There
is also olfactory and gustatory, i.e. smell and taste. The
perceptual representation is a modeling approach that
highlights the constructive, or generative function of
perception, or how perceptual processes construct a
complete volumetric spatial world, complete with a cop y
of our own body at the center of that world. The repre-
sentational strategy used by the brain is an analogical one;
that is, objects and surfaces are represented in the brain
not by an abstract symbolic code, or in the activation of
individual cells or groups of cells representing particular
features detected in the visual field. Instead, objects are
represented in the brain by constructing full spatial effi-
gies of them that appear to us for all the world like the
objects themselves or at least so it seems to us only be-
cause we have never seen those objects in their raw form,
but only through our perceptual representations of them.
Objects of perception are the entities we attend to
when we perceive the world. Perception lies at the root
of all our empirical knowledge. So far there are 3 theo-
ries of perception mainly, that is, direct realism [14], in-
direct realism, Gestalt principles [15]. The fundamental
question we shall consider concerns the objects of per-
ception: what is it we attend to when we perceive the
world?
As you know that the binding problem is an important
problem across many disciplines, including psychology,
neuroscience, computational modeling, and even phi-
losophy. Feature binding is the process how a large col-
lection of coupled neurons combines external data with
internal memories into coherent patterns of meaning.
According to neural synchronization theory, feature
binding is achieved via neural synchronization. When
external stimuli come into the brain, neurons corre-
sponding to the features of the same object will form a
dynamic neural assembly by temporal synchronous neu-
ral oscillation, and the dynamic neural assembly, as an
internal representation in the brain, codes the object in
the external world.
In 1990, Eckhorn and coworkers proposed a Linking
Field Network according to the synchronized neural os-
cillation in the visual cortex of cat [16]. Linking Field
Network can synchronize stimuli evoked oscillations at
different regions in the visual cortex if the regions have
similar local coding properties. Referred to noisy neural
model, Bayesian method and competition mechanism a
computational model for feature binding has been pro-
posed [17].
3.3. How Are Memories Stored and Retrieved?
Memory can be defined as a lasting representation that is
reflected in thought, experience or behavior. Based on
operation time memory can be categorized as sensory
memory, working memory or short-term memory,
long-term memory. Sensory memory is memory from
our immediate sensory. Sensory memory preserves ac-
curate representation of the physical features of sensory
stimuli for a few seconds or less. Working memory holds
information temporarily in the order of seconds to min-
utes. Long-term memory can be considered a warehouse
of all experiences, events, skills, words, rules, emotions,
and judgments that have been attained from sensory and
short-term memory.
In terms of the types of information stored long-term
memory can be classified into declarative and non-de-
clarative memories. Declarative memory can be further
divided into episodic and semantic memory, while
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
11
non-declarative memory consists of procedural memory
and conditioning memory.
Understanding ho w memories are stored in the brain is
an essential step toward understanding ourselves. Since
the 1970s, work on isolated chunks of nervous-system
tissue has identified a host of molecular players in mem-
ory formation. Many of the same molecules have been
implicated in both declarative and non-declarative mem-
ory. A key insight from this work has been that
short-term memory involves chemical modifications that
strengthen existing connections, called synapses, be-
tween neurons, whereas long-term memory requires pro-
tein synthesis and probably the construction of new syn-
apses [18].
A brain has distributed memory system, that is, each
part of brain has several types of memories that work in
somewhat different ways, to suit particular purposes.
According to the stored time of contents memory can be
divided into long term memory, short term memory and
working memory. Research topics in memory exist cod-
ing, extract and retrieval of information. Current working
memory attracts more researchers to involve.
Working memory will provides temporal space and
enough information for complex tasks, such as under-
standing speech, learning, reasoning and attention. There
are memory and reasoning functions in the working
memory. It consists of three components: that is, central
nervous performance system, video space primary proc-
essing and phonetic circuit [19].
Memory phenomena have also been categorized as
explicit or implicit. Explicit memories involve the hip-
pocampus-medial temporal lobe system. The most com-
mon current view of the memorial functions of the hip-
pocampal system is the declarative memory. There are a
lot of research issues that are waiting for us to resolve.
What is the readout system from the hippocampal system
to behavioral expression of learning in declarative mem-
ory? Where are the long-term declarative memories
stored after the hippocampal system? What are the
mechanisms of time-limited memory storage in hippo-
campus and storage of permanent memories in extra-
hippocampal structures?
Implicit memory involves the cerebellum, amygdale,
and other systems [20]. The cerebellum is necessary for
classical conditioning of discrete behavioral responses
under all condition. It is learning to make specific be-
havioral responses. The amygdales system is learning
fear and associated autonomic responses to deal with the
situation.
3.4. What Is the Neural Basis of Language?
Language is fundamentally a means for social commu-
nication. Language is also often held to be the mirror of
the mind. Chomsky developed transformational grammar
that cognitivism replaced behaviorism in linguistics [21].
Through language we organize our sensory exp erience
and express our thoughts, feelings, and expectations.
Language is particular interesting from cognitive infor-
matics point of view because its specific and localized
organization can explore the functional architecture of
the dominant hemisphere of the brain.
Recent studies of human brain show that the written
word is transferred from the retina to the lateral genicu-
late nucleus, and from there to the primary visual cortex.
The information then travels to a higher-order center,
where it is conveyed first to the angular gyrus of the pa-
rietal-temporal-occipital association cortex, and then to
Wernicke’s area, where the visual information is trans-
formed into a phonetic representation of the word. For
spoken word the auditory information is processed by
primary auditory cortex. Then the information input to
higher-order auditory cortex, before it is conveyed to a
specific region of the parietal-temporal-occipital associa-
tion cortex, the angular gyrus, which is concerned with
the association of incoming auditory, visual, and tactile
information. From here the information is projected to
Wernicke’s area and Broca’s area. In Broca’s area the
perception of language is translated into the grammatical
structure of a phrase and the memory for word articula-
tion is stored [22].
3.5. How Does the Brain Learn?
Learning is the basic cognitive activity and accumulation
procedure of experience and knowledge. Through learn-
ing the system performance will be improved. Percep tual
learning, cognitive learning, implicit learning are active
research topics in the learning area.
Perceptual learning should be considered as an active
process that embeds particular abstraction, reformulation
and approximation within the Abstraction framework.
The active process refers to the fact that the search for a
correct data representation is performed through several
steps. A key point is that perceptual learning focuses on
low-level abstraction mechanism instead of trying to rely
on more complex algorithm. In fact, from the machine
learning point off view, perceptual learning can be seen
as a particular abstraction that may help to simplify
complex problem thanks to a computable representation.
Indeed, the baseline of Abstraction, i.e. choosing the
relevant data to ease the learning task, is that many
problems in machine learning cannot be solve because of
the complexity of th e representation and is no t related to
the learning algorithm, which is referred to as the phase
transition problem. Within the abstraction framework,
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
12
we use the term perceptual learning to refer to specific
learning task that rely on iterative represen tation changes
and that deals with real-world data which human can
perceive.
In contrast with perceptual learning cognitive leaning
is a leap in the process of cogn ition and generate knowl-
edge through clustering, classification, conceptualization
and so on. In general, there are inductive learning, ana-
logical learning, case-based learning, explanation learn-
ing, evolutional learning connectionist learning.
The core issue of cognitive learning is self-organizing
principles. Kohonen has proposed a self-organizing maps
which is a famous neural network model. Babloyantz
applied chaotic dynamics to study brain activity. Haken
has proposed a synergetic approach to brain activity,
behavior and cognition.
Introspective learning is an inside learning of brain,
which means without input information from outside
environment. We have proposed a model for introspect-
tive learning which employs case-based reasoning and
ontology- based knowled ge [23].
The term implicit learning was coined by Reber to re-
fer to the way people could learn structure in a domain
without being able to say what they had learnt [24]. Re-
ber first proposed artificial grammars to study implicit
learning for unconscious knowledge acquisition. It will
help us to understand the learning mechanism without
consciousness. Since middle of 1980’s implicit learning
become an active research area in psychology.
3.6. How to Think in Human Brain?
Thought is a reflection of essen tial attributes and internal
laws of objective reality in conscious, indirect and gen-
eralization by human brain with consciousness [25]. In
recent years, there has been a noteworthy shift of interest
in cognitive science. Cognitive pro cess rises man’s sense
perceptions and impressions to logical knowledge. Ac-
cording to abstraction degree of cognitive process, hu-
man thought can be divided into three levels: perception
thought, image thought and abstraction thought. A hier-
archical model of thought which illustrates the charac-
teristics and correlations of thought levels has been pro-
posed and shown in Figure 1 [10,26].
Perception thought is the lowest level of thought. Be-
havior is the objective of research in perception thought.
Reflection is a function of stimulus. Perception thought
emphasizes stimulus-reflection schema or perception-
action schema. The thought of animal and infant usually
belong to perception thought because they can not intro-
spect, and also can not declare empirical consciousness.
In perception thought, intelligent behavior takes place
without representation and reasoning.
Attention
Audio Vision Motor
Image 1 Image 2 Image 3
Abstraction Abstraction
Thought
Image
Thought
Perceptual
Thought
Figure 1. Hierarchical thought model of brain.
Behavior-based artificial intellig ence has produced the
models of intelligence which study intelligence from the
bottom up, concentrating on physical systems, situ ated in
the world, autonomously carrying out tasks of various
sorts. They claim that the simple things to do with per-
ception and mobility in a dynamic environment took
evolution much longer to perfect. Intelligence in human
have been taking place for only a very small fraction of
our evolutionary lineage. Machine intelligence can take
evolution by the dynamics of interaction with the world.
Image thought adopts intuitive imagery as thinking
element. Intuitive imagery is one kind of information
which acquires through processing perceptual represent-
tation, but does not yet generate concepts of language.
Typical image thought is pattern recognition which can
deal with pattern information, such as character, image,
speech, classification and recognition of objects, and so
on [10,25].
Based on perceptual knowledge, the process which re-
flects the common properties and exposes internal rela-
tions of distinct objects through concepts, judgment and
inference is called abstraction thought. Concepts are no
longer the phenomena, the separate aspects and the ex-
ternal relations, while reflect the essences and internal
relations of objects. Judgment represents the certain rela-
tions between conceptions. Inference acquires new
knowledge from existing knowledge. There are existing
deductive reasoning, inductive reasoning, and abductive
reasoning currently. By means of judgment and inference
one is able to draw logical conclusions. Logical knowl-
edge is capable of grasping the development of the sur-
rounding world in its totality, the internal relations of all
its aspects.
Attention focuses consciousness to produce greater
vividness and limits the number of thoughts that can be
entertained at one time. Attention forces human thinking
process from parallel to sequential in terms of leaping
from image thought to abstraction thought.
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
13
3.7. What Is the Procedure of Intelligence
Development?
From pregnant with, born to the adult’s coursing human
cerebral cortex passed perception and effectors with ex-
ternal environment that mutual and plasticity develop-
ment takes place, cause corresponding intelligence and
cognitive ability to reach maturity progressively. The
first theory of intelligence development which is a
comprehensive theory about the nature and development
of human intelligence developed by Jean Piaget. It is
primarily known as a developmental stage theory. Jean
Piaget proposes that there are four distinct, increasingly
sophisticated stages of mental representation that chil-
dren pass through on their way to an adult level of intel-
ligence [27].
1) From birth to 2 years old is the sensorimotor stage.
2) From 2 years old to 7 years old is pre-operation
stage. The Preoperational Stage can be further broken
down into the Pre-conceptual stage and the Intuitive stage.
3) Concrete operational stage is between the ages of 7
and 11 years. This stage is characterized by the appropri-
ate use of internalized, reversible, conservative and logi-
cal actions.
4) Formal operational stage is between the ages of 12
and 15 years. In this stage, individuals move beyond
concrete experiences and begin to think abstractly, rea-
son logically and draw conclusions from the information
available, as well as apply all these processes to hypo-
thetical situations.
Soviet psychologist Vygotsky’s most important con-
tribution concerns the inter-relationship of language de-
velopment and thought. This concept, explored in Vy-
gotsky’s book Thought and Language, establishes the
explicit and profound connection between speech, and
the development of mental concepts and cognitive
awareness [28].
In recent years, biological mechanism, computational
theory and cognitive applications of intelligence devel-
opment have obtained very great development. Studying
on Independent motivation and utility system in cogni-
tive computation, internal representation and develop-
ment mechanism based on the basis of sensory percep-
tion and action effection, getting nonspecific study and
characteristics of popularization and application, and
development structure of cognitive computation scholars
have made the abundant achievements. Through the
study of intelligence development the mankind can ex-
pand physical limit greatly.
3.8. What Is the Nature of Emotion?
The mental perception of some fact excites the mental
affection called the emotion, and that this latter state of
mind gives rise to the bodily expression. Emotion is a
complex psychophysical process that arises spontane-
ously, rather than through conscious effort, and evokes
either a positive or negative psychological response and
physical expressions. Research on emotion at varying
levels of abstraction, using different computational meth-
ods, addressing different emotional phenomena, and
basing their models on different theories of affect.
Since the early 1990s emotional intelligence is sys-
tematically studied [29]. Scien tific articles suggested that
there existed an unrecognized but important human
mental ability to reason about emotions and to use emo-
tions to enhan ce thought. Emotion al intelligen ce refers to
an ability to recogn ize the meanings of emotion an d their
relationships, and to reason and problem solve on the
basis of them. Emotional intelligence is involved in the
capacity to perceive emotion s, assimilate e mo tio n-related
feelings, understand the information of those emotions,
and manage them.
Emotional state refers to one’s internal dynamics when
one has an emotion. Emotional states influence available
information in working memory, subjective utility of
alternative choices and style of processing. There are five
models on emotion have been proposed, that is, Ortony
Clore Collins cogn itive model [30], Roseman’ s cognitive
appraisal model [31], three-layer architecture [32],
six-layer architecture [33], four elicitors for emotion
synthesis [34]. However, till today, th ere is no model that
can completely represent the human emotional system.
One of the key problems is how to map emotional states
to behaviors.
Another problem is the possible emotion al circuitry in
the brain. The process of emotion engages parts of the
cortex, in particular th e frontal cortex. The frontal cortex
communicates with the limbic system and impacts deci-
sion-making.
3.9. What Is the Nature of Consciousness?
The most important scientific discovery of the present
era will come to answer how exactly do neurobiological
processes in the brain cause consciousness? The question
“What is the biological basis of consciousness?” is se-
lected as one of 125 questions, a fitting number for Sci-
ence’s 125th anniversary. Recent scientifically oriented
accounts of consciousness emerging from the properties
and organization of neurons in the brain. Consciousness
is the notions of mind and soul.
The physical basis of consciousness appears to be the
most singular challenge to the scientific, reductionist
world view. Francis Crick’s book ‘The astonishing Hy-
pothesis’ is an effort to chart the way forward in the in-
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
14
vestigation of consciousness [35]. Crick has proposed the
basic ideas of researching consciousness:
1) It seems probable, however, that at any one moment
some active neuronal processes in your head correlate
with consciousness, while others do not. What are the
differe nc es b etween them?
2) All the different aspect of consciousness, for exam-
ple pain and visual awareness, employ a basic common
mechanism or perhaps a few such mechanisms. If we
could understand the mechanisms for one aspect, then we
hope we will have gone most of the way to understand-
ing them all.
Bernard Baars has proposed Global Workspace The-
ory (GWT) which integrates conscious contents with
unconscious distributed expertise in the brain [36]. A
theatre metaphor for GWT is a useful approximation.
Unconscious processors in the theatre audience receive
broadcasts from a conscious bright spot on the stage.
Control of the bright spot corresponds to selective atten-
tion. Backstage, unconscious contextual systems operate
to shape and direct conscious contents.
Chalmers suggests the problem of consciousness can
be broken down into several separate questions. The
major question is the neuronal co rrelate of consciousness
(NCC) which focuses on specific processes that correlate
with the current content of consciousness [37]. The NCC
is the minimal set of neurons, most likely distributed
throughout certain cortical and subcortical areas, whose
firing directly correlates with the perception of the sub-
ject at the time. Discovering the NCC and its properties
will mark a major milestone in any scientific theory of
consciousness. Several other questions need to be an-
swered about the NCC. What type of activity corre-
sponds to the NCC? What causes the NCC to oc cur? And,
finally, what effect does the NCC have on postsynaptic
structures, including motor output.
3.10. How to Build Mind Model?
Mind could be defined as: “That which thinks, reasons,
perceives, wills, and feels. The mind now appears in no
way separate from the brain. In neuroscience, there is no
duality between the mind and body. They are one.” in
Medical Dictionary [38]. A mind model is intended to be
an explanation of how some aspect of cognition is ac-
complished by a set of primitive computational processes.
A model performs a specific cognitive task or class of
tasks and produces behavior th at constitutes a set of pre-
dictions that can be compared to data from human per-
formance. Task domains that have received considerable
attention include problem solving, language comprehen-
sion, memory tasks, and human-device interaction.
Researchers try to construct mind model to illustrate
how brains do. Anderson and colleagues have demon-
strated that a production rule analysis of cognitive skill,
along with the learning mechanisms posited in the
ACT-R model, provide detailed and explanatory ac-
counts of a range of regularities in cognitive skill acqui-
sition in complex domains such as learning to program
Lisp [39]. ACT-R also provides accounts of many phe-
nomena surrounding the recognition and recall of verbal
material, and regularities in problem solving strategies
[40-42].
In the early 1980’s, SOAR was develop ed to be a sys-
tem that could support multiple problem solving methods
for many different problems [43]. In the mid 1980’s,
Newell and many of his students began working on
SOAR as a candidate of unified theories of cognition.
SOAR is a learning architecture that has been applied to
domains ranging from rapid, immediate tasks such as
typing and video game interaction to long stretches of
problem solving behavior [44]. SOAR has also served as
the foundation for a detailed theory of sentence process-
ing, which models both the rapid on-line effects of se-
mantics and context, as well as subtle effects of syntactic
structure on processing difficulty across several typo-
logically distinct languages.
Stan Franklin et al. have proposed a mind model
called LIDA [45,46] which is grounded in the LIDA cog-
nitive cycle. Each cognitive cycle the LIDA agent first
makes sense of its current situation as best as it can. It
then decides what portion of this situation is most in
need of attention. Broadcasting this portion, the current
contents of consciousness, enables the agent to finally
chose an appropriate action and execute it.
A new mind model called Consciousness and Memory
(CAM) is proposed by Intelligence Science Laboratory
of Institute of Computing Technology [47]. Figure 2
shows you the architecture of CAM model which con-
sists of three main parts, which are consciousness, mem-
Figure 2. Architecture of CAM.
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
15
ory and high level cognitive functions. The conscious-
ness possesses a set of planning schemes which arrange
the components of CAM to accomplish different cogni-
tive tasks. The memory part contains three types of
memory which are long term memory, short term mem-
ory and working memory. The high lev el cognitive func-
tion part includes a class of high level cognitive func-
tions such as event detection, action execution etc.
In CAM model, for episodic memory we employ
case-based system to retrieve the episode according to
the cues; while for semantic memory we adopt dynamic
description logic to represent concepts and reasoning.
Also we develop the cognitive cycle consists of aware-
ness, intension, action composition which will be dis-
cussed in another paper.
4. Conclusions and Perspective
Intelligence Science is a new paradigm and interdisci-
plinary subject. Ten basic issues of Intelligence Science
have been explored in the paper. These problems will
constitute foundatio n of Intelligence Science and waiting
for scientist to study.
Intelligence Science will let the human dream be real-
ity to replace human brain work by machine intelligence
The incremental efforts in neuroscience and cognitive
science provide us exciting solid foundation to explore
brain model and intelligen t behavior. We shou ld research
on neocortical column, population coding, mind model,
consciousness etc. for the human-level intelligence [48].
We believe that intelligence scien ce will make great pro-
gress and new breakthroughs in the coming years. It is a
good opportunity to co ntribute our intellect and ab ility to
promote the development of intelligence science and
become a bright spot of human civilization in 21 century.
5. Acknowledgements
The author would like to thank to Prof. J. R. Anderson,
Prof. Loft A. Zadeh, Prof. Nils J. Nilsson, Prof. P. Ro-
senbloom and Prof. Yixin Zhong for their valuable dis-
cussions. I also acknowledge my colleagues at the Intel-
ligence Science Laboratory for their contribu tions.
6. References
[1] J. McCarthy, “The Future of AI-A Ma nifesto,” AI Maga-
zine, Vol. 26, No. 4, 2005, p. 39.
[2] H. Jeff and S. Blakeslee, “On Intelligence, Times Books,”
Henry Holt and Company, New York City, 2004.
[3] R. Granger, “Engines of the Brain,” AI Magazine, Vol. 27,
No. 2, 2006, pp. 15-31.
[4] P. Langley, “Cognitive Architectures and General Intel-
ligent Systems,” AI Magazine, Vol. 27, No. 2, 2006, pp.
33-44,
[5] K. Sycara, “Multi-Agent Systems,” AI Magazine, Vol. 19,
No. 2, 1998, pp. 79-93.
[6] N. Cassimatis, “A Cognitive Substrate for Achieving
Human-Level Intelligence,” AI Magazine, Vol. 27, No. 2,
2006, pp. 45-56.
[7] J. Snaider, R. McCall, S. Franklin, “The LIDA Frame-
work as a General Tool for AGI,” AGI2011. (To be ap-
peared).
[8] Intelligence Science Web
http://www.intsci.ac.cn/.
[9] Z. Z. Shi, “Intelligence Science,” in Chinese, Tsinghua
University Press, Beijing, 2006.
[10] Z. Z. Shi, “Artificial Thought and Intelligent Systems,”
AI Summer School’94, Beijing, 1994.
[11] S. Legg and M. Hutter, “A Collection of Definitions of
Intelligence,” 2006.
www.vetta.org/shane/intelligence.html.
[12] M. F. Bear, B. Connors, M. Paradiso, M. F. Bear, B. W.
Connors and M. A. Neuroscience, “Exploring the Brain,”
2nd Edition, Lippincott Williams & Wilkins, Philadelphia
2002.
[13] H. Z. Shouval, “Models of Synaptic Plasticity,” Schola-
rpedia, Vol. 2, No. 7, 2007, p. 1605.
doi:10.4249/scholarpedia.1605
[14] J. J. Gibson, “The Ecological Approach to Visual Percep-
tion” Houghton Mifflin, Boston, 1979.
[15] D. Chang, K. V. Nesbitt and K. Wilkins, “The Gestalt
Principle of Continuation Applies to Both the Haptic and
Visual Grouping of Elements,” WHC ’07 Proceedings of
the Second Joint EuroHaptics Conference and Sympo-
sium on Haptic Interfaces for Virtual Environment and
Teleoperator Systems, 2007, pp. 15-20.
doi:10.1109/WHC.2007.113
[16] R. Eckhorn, H. J. Reitboeck, M. Arndt and P. W. Dicke,
“Feature Linking via Synchronization among Distributed
Assemblies: Simulation of Results from Cat Cortex,”
Neural Computation, Vol. 2, No. 3, 1990, pp. 293-307.
doi:10.1162/neco.1990.2.3.293
[17] Z. W. Shi, Z. Z. Shi, X. Liu and Z. P. Shi, “A Computa-
tional Model for Feature Binding,” Science in China, Se-
ries C: Life Science, Vol. 51, No. 5, 2009, pp. 470-478.
doi:10.1007/s11427-008-0063-3
[18] G. Miller, “How Are Memories Stored and Retrieved?”
Science, Vol. 309, No. 5731 2005, p. 92.
doi:10.1126/science.309.5731.92
[19] M. J. Dehn, “Working Memory and Academic Learning:
Assessment and Intervention,” Wiley, New York, 2008.
[20] M. Tarsia and E. Sanavio, “Implicit and Explicit Memory
Biases in Mixed Anxiety-Depression,” Journal of Affec-
tive Disorders, Vol. 77, No. 03, 2003, pp. 213-225.
doi:10.1016/S0165-0327(02)00119-2
[21] N. Chomsky, “Syntactic Structures,” Mouton, The Hague,
Z. Z. SHI
Copyright © 2011 SciRes. IJIS
16
1957.
[22] R. Mayeux and E. R. Kandel, “Disorders of Language:
The Aphasias,” In: E. R. Kandel, J. H. Schwarz and T. M.
Jessell Eds., Principles of Neural Science, 3rd Edition,
Elsevier, Amsterdam, 1991, pp. 840-851.
[23] Q. Dong and Z. Z. Shi, “A Research on Introspective
Learning Based on CBR,” International Journal of Ad-
vanced Intelligence, Vol. 3, No. 1, 2011. pp. 147-157.
[24] A. S. Reber, “Implicit Learning and Artificial Grammer,”
Journal of Verbal Learning and Verbal Behavior, Vol. 6,
No. 6, 1967, pp. 855-863.
doi:10.1016/S0022-5371(67)80149-X
[25] Z. Z. Shi, “Cognitive Science,” Press of USTC, Anhui,
2008.
[26] Z. Z. Shi, “Advanced Artificial Intelligence,” Science
Press of China, Beijing, 2006.
[27] J. Piaget, “Piaget’s Theory,” In: P. Mussen, Ed. Handbook
of Child Psychology, 4th Edition, Wiley, New York,
1983.
[28] J. Santrock, “A Topical Approach to Life-Span Develop-
ment. Chapter 6 Cognitive Development Approaches
(200 - 225),” McGraw-Hill, New York, 2004.
[29] D. A. Norman, “Emotion and Design: Attractive Things
Work Better,” Interactions Magazine, Vol. 4, No. 4,
2002.
[30] Ortony, G. L. Clore and A. Collins, “The Cognitive
Structure of Emotions,” Cambridge University Press,
Cambridge, 1988. doi:10.1017/CBO9780511571299
[31] McCarthy, O. L. Mejia, H. T. Liu and A. C. Durham,
“Cognitive Appraisal Theory: A Psychoeducational
Model for Connecting Thoughts and Feelings,” The An-
nual Convention of the American Psychological Associa-
tion, San Francisco, 14-18 August 1998.
[32] A. Sloman, “Review of Affective Computing,” AI Maga-
zine, Vol. 20, No. 1, 1999, pp. 127-133.
[33] M. Minsky, “The Emotion Machine,” Simon and Schus-
ter, New York City, 2006.
[34] J. Velasquez, P. Maes, “Cathexis: A Computational
Model of Emotions,” Proceedings of the First Interna-
tional Conference on Autonomous Agents, Marina del
Rey, 05 - 08 February 1997.
[35] F. Crick, “The Astonishing Hypothesis,” Scribner, Hum-
boldt, 1995.
[36] B. J. Baars and S. Franklin, “An Architectural Model of
Conscious and Unconscious Brain Functions: Global
Workspace Theory and IDA,” Neural Networks, Vol. 20
No. 9, 2007, pp. 955-961.
doi:10.1016/j.neunet.2007.09.013
[37] D. Chalmers, “The Conscious Mind: In Search of a Fun-
damental Theory,” Oxford University Press, Oxford,
1995.
[38] Medical dictionary Website:
http://www.medterms.com/script/main/hp.asp.
[39] J. R. Anderson, “The Adaptive Character of Thought,”
Erlbaum, Hillsdale, 1993.
[40] J. R. Anderson and Y. Qin, “Using Brain Imaging to Ex-
tract the Structure of Complex Events at the Rational
Time Band,” Journal of Cognitive Neuroscience, Vol. 20,
No. 9, 2008, pp. 1624-1636. doi:10.1162/jocn.2008.20108
[41] J. R. Anderson, “Using Neural Imaging to Inform the
Instruction of Mathematics,” N. Zhong, K. C. Li, S. F. Lu
and L. Chen, Eds., Brain Informatics, International Con-
ference: Lecture Notes in Computer Science, Toronto,
28-30 August 2010.
[42] J. R. Anderson, S. A. Betts, J. L. Ferris and J. M. Fin-
cham, “Neural Imaging to Track Mental States While
Using an Intelligent Tutoring System,” Proceedings of
the National Academy of Science, Vol. 107, No. 15, 2010,
pp. 7018-7023. doi:10.1073/pnas.1000942107
[43] J. Laird, A. Newell and P. Rosenbloom, “SOAR: An
Architecture for General Intelligence,” Artificial Intelli-
gence, Vol. 33, No. 1, 1987, pp. 1-64.
doi:10.1016/0004-3702(87)90050-6
[44] A. Newell, “Unified Theories of Cognition,” Harvard
University Press, Cambridge, 1990.
[45] J. Baars and S. Franklin, “How Conscious Experience
and Working Memory Interact,” Trends in Cognitive
Science, Vol. 7, No. 4, 2003, pp. 166-172.
doi:10.1016/S1364-6613(03)00056-1
[46] D. Friedlander and S. Franklin, “LIDA and a Theory of
Mind,” IOS Press, Amsterdam, pp. 137-148.
[47] Z. Z. Shi and X. F. Wang, “A Mind Model CAM in Intel-
ligence Science,” International Journal of Advanced In-
telligence, Vol. 3, No. 1, 2011, pp. 119-129.
[48] Z. Z. Shi, “Research on Brain-Like Computer,” Computer
Science, Vol. 5820, 2009, p. 5.
doi:10.1007/978-3-642-04875-3_5