Journal Pre-proof

Study on Artificial Intelligence: The State of the Art and Future

Prospects

Caiming Zhang , Yang Lu

PII: S2452-414X(21)00024-8
DOI: https://doi.org/10.1016/j.jii.2021.100224
Reference: JII 100224

To appear in: Journal of Industrial Information Integration

Received date: 10 March 2021

Revised date: 27 April 2021
Accepted date: 1 May 2021

Please cite this article as: Caiming Zhang , Yang Lu , Study on Artificial Intelligence: The
State of the Art and Future Prospects, Journal of Industrial Information Integration (2021), doi:
https://doi.org/10.1016/j.jii.2021.100224

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2021 Published by Elsevier Inc.

 Study on Artificial Intelligence: The State of the
Art and Future Prospects
this study is that it clarifies the state of the art of AI for
future study.
Caiming Zhang1, #, Yang Lu2 Keywords—Artificial intelligence (AI), machine learning,
natural language processing (NLP).

1. China University of Labor Relations I. INTRODUCTION

Beijing, 100048 I N1956, at a conference at Dartmouth University, scholars

formally proposed the term "artificial intelligence." That
moment was the first step in anew topic of studying how
machines simulate human intelligent activities. In early 2016,
China AlphaGo defeated the world chess champion. This event
immediately aroused global interest in artificial intelligence
(AI)(Xu, 2013). The development of artificial intelligence has
brought huge economic benefits to mankind and has benefited
2. University of Central Oklahoma all aspects of life, even as it has greatly promoted social
development and brought social development into a new
era(Luand Xu. 2018). Many scholars started AI-relevant
Edmond, OK 73034 USA research since the end of 20th century (Xu, 1999).
AI is the general term for the science of artificial intelligence.
It uses computers to simulate human intelligent behaviors and it
trains computers to learn human behaviors such as learning,
# {[email protected] } judgment, and decision-making(Xu, et al., 2021). AI is a
knowledge project that takes knowledge as the object, acquires
knowledge, analyzes and studies the expression methods of
knowledge, and employs these approaches to achieve the effect
Abstract—In the world, the technological and industrial of simulating human intellectual activities(Duan and Xu,
revolution is accelerating by the widespread application of 2012). AI is a compilation of computer science, logic, biology,
new generation information and communication psychology, philosophy, and many other disciplines, and it has
technologies, such as AI, IoT (the Internet of Things), and achieved remarkable results in applications such as speech
blockchain technology. Artificial intelligence has attracted recognition, image processing, natural language processing, the
much attention from government, industry, and academia. proving of automatic theorems, and intelligent robots(Duan, et
In this study, popular articles published in recent years that al., 2009). AI plays an indispensable role in social
relate to artificial intelligence are selected and explored. development, and it has brought revolutionary results in
This study aims to provide a review of artificial intelligence improving labor efficiency, reducing labor costs, optimizing
based on industry information integration. It presents an the structure of human resources, and creating new job
overview of the scope of artificial intelligence using demands(Duan, et al., 2019).
background, drivers, technologies, and applications, as well The rest of the paper is structured as follows (Figure 1): the
as logical opinions regarding the development of artificial origin and development of AI is addressed in Section II. Section
intelligence. This paper may play a role in AI-related III describes enabling drivers and technologies of AI. The
research and should provide important insights for research fields are depicted in Section IV. Section V illustrates
practitioners in the real world.The main contribution of the application scenarios of AI. Three logical views of AI are
explained in Section VI, and Section VII summarizes the paper.
1
China University of Labor Relations, Beijing 100048 China
2
University of Central Oklahoma, Edmond, OK 73034 USA
*
Corresponding Author: [email protected]
 S e c tio n I In tr o d u c tio n millions of dollars. After that, people began to try to study
general artificial intelligence programs, but they encountered
F o u n d a tio n o f A I A I a n d D iffe r e n t D isc ip lin e s R o a d m a p o f th e A r tic le serious obstacles and reached a deadlock. Artificial intelligence
is at a trough again. In 1997, the success of "Deep Blue" put the
S e c tio n II T h e O r ig in a n d D e v e lo p m e n t o f A I development of AI on the agenda. With the improvement of
computing power, the bottleneck of AI was broken, and the
F o u n d in g P e r io d T h e F ir st G o ld e n A g e T h e S e c o n d G o ld e n A g e T h e T h ir d G o ld e n A g e
development of deep learning and enhanced learning based on
big data continued. With the continuous development of GPUs,
S e c tio n III D r iv e r s a n d T e c h n o lo g ie s o f A I
the successful development of custom processors has also
N a tio n a l L a n g u a g e
B ig D a ta A lg o r ith m s M a c h in e L e a r n in g P r o c e ssin g H a r d w a r e C o m p u tin g V isio n continuously improved computing power; this has laid the
foundation for the explosive development of AI(Zhang, 2019).
Artificial intelligence has experienced a long development
S e c tio n IV T h e R e sa r c h F ie ld s o f A I
process, with a history of more than 70 years. Its development
In te llig e n t D e c isio n S u p p o r t process can be divided into several stages: in 1943, the artificial
E x p e r t S y ste m s M a c h in e L e a r n in g R o b o tic s S y ste m P a tte r n R e c o g n itio n
neuron model was proposed, and this opened the era of artificial
neural network research(Zeng, et al., 2012). In 1956, the
S e c tio n V A p p lic a tio n S c e n a r io s o f A I
Dartmouth Conference was held, and the concept of artificial
A u to m o tiv e F in a n c ia l M a r k e t H e a lth c a r e R e ta ilin g M e d ia
intelligence was put forward; this marked the birth of artificial
intelligence. During this period, the trend of artificial
intelligence research by the international academic community
S e c tio n V I V ie w p o in ts o f A I
was on the rise, and academic exchanges were frequent. In the
S c e n a r io s 1960s, the main types of connectionism and submissive fell
S y m b o lism C o n n e c tio n ism B e h a v io r ism
into disuse, and smart technology took a downturn in
development. Research on the backpropagation algorithm
S e c tio n V II C o n c lu sio n
began in the 1970s, and the cost and computing power of
F ig u r e 1 . R o a d m a p o f T h e S tu d y computers gradually increased, which made the research and
application of expert systems difficult. Moving forward
II. THE ORIGIN AND DEVELOPMENT OF ARTIFICIAL became difficult, but artificial intelligence was
INTELLIGENCE graduallymaking breakthroughs. In the 1980s, backpropagation
neural networks were widely recognized, algorithm research
Artificial intelligence is the study of how to make computers
based on artificial neural networks was rapidly developed,
perform intelligent tasks that, in the past, could only be
computer hardware functions had rapidly improved, and the
performed by humans(Huang, et al., 2019a). In recent years, AI
development of the Internet reduced the development of
has developed rapidly, and it has changed people's
artificial intelligence. In the first decade of the 21st century, the
lifestyles(Huang, et al., 2019b). The development of AI has
development of mobile Internet brought more artificial
become an important development strategy for countries
intelligence application scenarios(Zhangand Chen, 2020.). In
around the world, enhancing national competitiveness and
2012, deep learning was proposed, and artificial intelligence
maintaining security(Rajkomar, et al., 2018). Many countries
achieved breakthrough development; the algorithm has made
have introduced preferential policies and have strengthened the
breakthroughs in speech and visual recognition. The following
deployment of key technologies and talents in order to take the
table (Table 1) addresses the events of the AI timeline.
lead in a new round of international competition(Xu, et al.,
Table I
2008.). AI has become a research hotspot in science and
The Development of AI
technology; major companies such as Google, Microsoft, and
Period Events &Details
IBM are committed to AI and are applying AI to more and more
fields(Shi, et al., 2007). Founding In the summer of 1956, at an academic
AI is a multidisciplinary technology, one with the capability period conference held at Dartmouth University,
of integrating cognition, machine learning, emotion several distinguished scientists discussed
recognition, human-computer interaction, data storage, and how to make machines simulate
decision-making(Lu, 2019). It was first proposed by John intelligent. McCarthy first proposed the
McCarthy at the Dartmouth Conference in the mid-20th term artificial intelligence, which gave
century. Since 1993, AI has achieved some milestone results. birth to artificial intelligence. This
Due to the wide application of the BP algorithm, the neural emerging discipline.
network has developed rapidly. In a large-scale environment, The first Since the Dartmouth Conference, the
the extensive use of expert systems has saved the industry a lot golden age emergence of expert systems, such as the
of cost and has improved industry efficiency(Zhang and Chu, Bender Chemical Mass Spectrometry
2020). For example, the PROSPECTOR expert system System, MTCIN Disease Diagnosis and
successfully analyzed mineral deposits worth hundreds of Treatment System, and Hearsay-11
Language Understanding System, laid
 the foundation for the practicality of this summarized laws and methods. However, this abstraction
manual. operating. intelligence. In the method has great limitations and low accuracy. Researchers
early days of artificial intelligence have been inspired by babies; no one teaches babies how to
research, the Massachusetts Institute of recognize objects, but they learn. Based on this, people have
Technology, Carnegie Mellon proposed a machine learning method that summarizes the laws
University, Stanford University, and and methods of identifying objects (Li, 1999a & b; Yang, et al.,
other universities established artificial 2001a;). For example, if you input a lot of photos of dogs into a
intelligence laboratories and received computer, the computer can learn the characteristics of dogs
R&D funding from government through training models (such as neural networks) and can
agencies. In the late 1970s, Feigenbaum accurately recognize dogs in other photos based on these
proposed the concept of knowledge characteristics(Zhangand Fu, 2021).
engineering. Expert systems have Machine learning enables computers to automatically learn
developed rapidly, and their applications and analyze big data, and then to make decisions and
have also produced huge benefits. predictions about events in the real world. Algorithms make AI
However, a series of problems such as possible (Chi-Hsienand Nagasawa, 2019). And beyond the
the difficulty of obtaining knowledge application of AI-relevant algorithms in the field of pattern
from expert systems gradually emerged, recognition, satisfactory results have also been achieved in
and artificial intelligence entered the first other fields, among them speech recognition, search engines,
trough. semantic analysis, and recommendation systems(Zhang, et al.,
The second The Hopfield neural network and BT 2019a.). Each of these has made considerable progress under
golden period training algorithm proposed in 1982, set the impetus of AI algorithms.
off a boom in the development of
C. Machine Learning
artificial intelligence such as speech
translation and speech recognition. The basic idea of machine learning is the use of an algorithm
However, in the late 1990s, people that improves its performance by learning from data(Nilsson,
thought that artificial intelligence was 2014.). The four most important types of problems that need to
still far from our social life. Therefore, be solved via machine learning are prediction, clustering,
around 2000, artificial intelligence classification, and dimensionality reduction(Erhan, et al.,
entered a trough again. 2010). Considering the classification of learning methods,
The third From 2006 to now is a period of rapid machine learning can be divided into four categories:
golden period development of artificial intelligence. supervised learning, unsupervised learning, semi-supervised
The rapid development is mainly due to learning, and reinforcement learning(Bose, 2017).
the widespread popularity of GPUs, and Supervised learning refers to the use of labeled data to train,
parallel processing can be faster and in order to predict the type or value of new data. According to
more powerful. Another reason is the different prediction results, this can be divided into two
unlimited expansion of storage capacity, categories: classification and regression. The typical methods
which allows large-scale access to data of supervised learning are SVM (Super Vector Machine) and
such as maps, pictures, text and data linear discrimination(Najafabadi, et al., 2015, Neyshabur, et al.,
information. 2017). The regression problem refers to the prediction of the
output of continuous values, one might analyze housing price
data, fit it according to sample data input, and then obtain a
III. ENABLING DRIVERS AND TECHNOLOGIES OF ARTIFICIAL
INTELLIGENCE continuous curve to predict housing prices. The classification
problem refers to predicting the output of discrete values, such
A. Big Data as judging whether the current photo is a dog, or a cat based on
Big data is a prerequisite for AI, and it is a core factor that a series of features; the output value is 1 or
promotes AI to improve recognition rate and accuracy. With 0(Morocho-Cayamcela, et al., 2019).
the development and the wide application of the Internet of When the data has no labels, unsupervised learning is data
Things, the amount of data generated has increased mining. Unsupervised learning is mainly reflected in clustering.
exponentially, with a great increased annual growth rate. In In short, data can be classified according to different
addition to increasing the number, the dimensionality of the characteristics without tags. Typical methods of unsupervised
data has also been expanded(Zhang, 2020). These large learning include k-clustering and principal component analysis.
amounts of high-dimensional data make the data more The important premise of k-clustering is that the difference
comprehensive and more sufficient to support the development between data can be measured by Euclidean distance. If it
of AI(Chenand Lin, 2014). cannot be measured, it needs to be converted to a usable
Euclidean distance. Principal component analysis is a statistical
B. Algorithms
method. By using orthogonal transformation, the relevant
In traditional pattern recognition, the researchers variables are transformed into uncorrelated variables; the
transformed variables are called principal components. The these two processes, the machine can hear and understand. The
basic idea is to replace the original related indicators with a set machine equipped with speech recognition and semantic
of independent comprehensive indicators(Baryannis, et al., understanding technology optimizes the algorithm in
2019.). continuous learning, so that the machine can not only listen, but
In its most literal meaning, semi-supervised learning can be also understand – and even understand emotions(Bostromand
understood to be a mixture of supervised learning and Yudkowsky, 2014).
unsupervised learning. In fact, labeled data and unlabeled data
E. Hardware
are mixed in the learning process. Under normal circumstances,
the amount of unmarked data is much larger than the amount of In machine learning, some "deep" neural network models are
marked data. The idea of semi-supervised learning is ideal, but used to solve complex problems. Machine learning is a way to
it is not much used in practical applications. Commonly used realize AI, and deep learning is a kind of machine
semi-supervised learning algorithms include self-training, learning(Makkar, et al., 2020). The hardware platform that runs
graph-based semi-supervised learning, and semi-supervised deep learning is mainly a GPU produced by NVIDIA. GPU
support vector machines (S3VM). acceleration is a new computing model that uses massively
Reinforcement learning is a method of obtaining rewards by parallel processors to accelerate applications with parallel
interacting with the environment, judging the quality of actions functions. In the past, it took a month for the CPU to obtain the
by reward levels, and then training the model. The importance training results. Now, the GPU can get results in one day. The
of exploration and development in reinforcement learning is a powerful parallel computing functions of the GPU alleviate the
thorny issue: to obtain better rewards, people must choose the training bottleneck period for deep learning algorithms, thereby
action that may get the highest reward, but people must also releasing the potential of artificial intelligence(Brynjolfssonand
find unknown actions(Chollet, 2017).The foundation of Mcafee2017).
reinforcement learning comes from behavioral psychology. In F. Computing Vision
1911, Thorndike proposed an effective rule: in an environment The goal of computer vision is to enable computers to
that makes people or animals feel comfortable, people or recognize and to understand the world through vision, as
animals will continue to strengthen this action. Conversely, if a humans do. It mainly uses algorithms to identify and analyze
person or an animal feels uncomfortable, the person or animal images. The most widely used computer visions are facial
will reduce its movement(Collinsand Moons, 2019). In other recognition and image recognition(Huand Hanzo, 2019).
words, reinforcement learning can enhance reward behavior Since 2015, image processing methods that use deep learning
and weaken punishment behavior. One can train the model for classification have been widely used. The neural network is
through the trial-and-error mechanism to find the best operation built into the network through neurons, and the model has a
and behavior to get the greatest return. This imitates the nonlinear fitting function through the activation
learning model of humans or animals and does not need toguide function(Dhingra, et al., 2016). Once the need to design input
agents to learn in a certain direction(Li, et al., 2007; Qi, et al., and output for the model is met, the model can automatically
2007; Wang, et al., 2013). learn the process of feature extraction and training
D. National Language Processing (NLP) classification. The use of deep learning simplifies the most
Natural language processing (NLP) refers to the ability of time-consuming and laborious image classification process and
computers to recognize and understand human text language, improves both the effect and the efficiency of image
which is an interdisciplinary subject between computer science classification. The VGG conventional neural network, ResNet
and human linguistics. The biggest difference in natural (Residual Neural Network) and Startup are the most used
language is that human thinking is based on language, so structures in engineering. The model used in the project must
natural language processing also represents a goal of AI. consider both efficiency and effectiveness; that is, it must
Natural language processing is divided into seven directions: ensure accuracy while ensuring speed. After training the model,
grammatical and semantic analysis, information extraction, text the model will be fine-tuned and reduced. Faster R-CNN,
mining, information retrieval, machine translation, the question Mask-RCNN, and YOLO (You Only Look Once) are now
answering system, and the dialogue system(Zhang, et al., commonly used network models. The common characteristics
2019b). of these models are high accuracy and fast speed. In the field of
Natural language understanding (NLP) is a technology that facial recognition, these models can detect and obtain results in
uses natural language to communicate with computers. The key real time(Davenportand Ronanki, 2018).
to processing natural language is to allow computers to Computer vision divides the picture into pixels and then
"understand" natural language, so it is also called processes the pixels. The meaning of semantic segmentation is
computational linguistics, which sits at the intersection of to understand the meaning of pixels after segmentation; this
language information processing and artificial intelligence. As allows for the recognition, for example of people, motorcycles,
a machine, it first collects sound signals. Natural language cars, and streetlights in a picture. It needs to distinguish dense
processing technology converts sound signals into text signals pixels. Convolutional neural networks promote the
and the meaning of the text. Then, machine converts sounds development of semantic segmentation algorithms(Zhao, et al.,
into words and converts words into meanings. After completing 2019a). The most basic method in semantic segmentation is to
perform classification prediction through a sliding window. the system structure and provide a
The emergence of the Fully Convolutional Network (FCN) material basis for the operation of
replaced the fully connected layer of the network. Based on computer software. The hardware
FCN, the Encoder-Decoder architecture was developed. Since platform running deep learning is mainly
the encoder was failure for smaller space size operations, the GPU.
decoder is used to restore the space size and detailed Computing Computer vision is a science that studies
information. After that, the pooling operation was replaced by Vision how to make machines "see". In addition,
Dialated / Atrous. The advantage of hole convolution is that the it refers to the use of cameras and
spatial resolution can be maintained. In addition to the previous computers instead of human eyes to
method, there is a method called Conditional Random Field identify, track and measure objects. This
(CRF) that can improve the segmentation effect(Zhao, et al., processing becomes an image more
2019b). suitable for human observation or
The following table (Table 2) address key drivers and transmission to an inspection instrument.
technologies of AI.
Table II IV. THE RESEARCH FIELDS OF AI
Drivers and Technologies of AI
A. The Expert System
Driver or Explanation
Technology An expert system is a knowledge system based on the
Big data refers to a collection of data that existing knowledge of human experts. The expert system was
Big Data
cannot be captured, managed, and the earliest field AI research. It is widely used in medical
processed using conventional software diagnosis, in geological surveying, and in the petrochemical
tools within a specific time frame. Big industry. Expert systems usually refer to various knowledge
data requires new processing models to systems(Miotto, et al., 2018). This is an intelligent computer
have stronger decision-making program, based on knowledge, that uses professional
capabilities, insight and discovery knowledge provided by human experts to simulate the thinking
capabilities, and process optimization process of human experts and uses knowledge and reasoning to
capabilities and diversified information solve complex problems that only domain experts can solve.
assets. The expert system has a large amount of information and
There are certain steps to perform any reasoning process in a specific field, including a large amount
Algorithms
operation. The methods and steps taken of professional knowledge and experience, and it can store,
to solve the problem are called reason, and judge at the same time (Tung, 2019). The core is the
algorithms. knowledge base and the reasoning engine(Cai, et al., 2018).
Machine learning is the science of The application method of the expert system is first to store
Machine
artificial intelligence. The main research the knowledge, experience, and research information of experts
Learning
object in this field is artificial in a specific field in the database and the knowledge base, and
intelligence, especially how to improve then to call them by the interpreter and reasoning engine and
the performance of specific algorithms in provide them to the experts as needed. Users do this through a
empirical learning. Machine learning is human-computer interaction interface(Zappone, et al., 2019).
the study of computer algorithms, which The application advantage of the expert system in teaching does
can be improved automatically through not depend on the limitation of time and space, nor on the
experience. Machine learning uses data limitation of environment and emotional influence. Expert
or experience to optimize the systems should be used in education. They are, in fact, widely
performance standards of computer used, and their advantages to distance education are well
programs. known(Pouyanfar, et al., 2018).
National Natural Language Processing (NLP) is B. Machine Learning
Language the field of computer science, artificial For a computer to possess knowledge, the knowledge must
Processing intelligence and linguistics. It focuses on be expressed as input into the computer in a manner that is
the interaction between computers and acceptable to the computer, or the computer itself has the ability
human (natural) language. The earliest to acquire knowledge, and to continuously summarize and
research work on natural language improve knowledge in practice(Young, et al., 2018). This
processing is machine translation. method is called machine learning(Liu, et al., 2020). The
Hardware Hardware refers to the general term for research of machine learning is mainly carried out (1) to study
various physical devices composed of the human learning mechanism and human brain thinking
electronic, mechanical, and process, (2) to study the learning mechanisms of people, and (3)
optoelectronic components in a computer to study machine learning methods and to establish a learning
system. These physical devices constitute system for specific tasks(Liu, et al., 2015). The research of
annuity according to the requirements of
machine learning is based on a variety of disciplines, such as al., 2018, Reichstein, et al., 2019).
information science, brain science, neuropsychology, logic, and
E. The Pattern Recognition
fuzzy mathematics(Hu, et al., 2021).
The concept of deep learning comes from artificial neural Pattern recognition is the study of how to make machines
networks. Common deep learning algorithms include the that have perception capabilities(Tan, et al., 2010). It mainly
Restricted Boltzman Machine (RBN), the deep belief network studies the recognition of visual and auditory patterns, i.e.,
(DBN), the convolutional neural network (CNN), and the identifying objects, terrain, images, fonts, etc. It has a wide
stacked auto-encoder(Litjens, et al., 2017.). In the early range of uses in daily life and for the military. In recent years,
artificial neural network, after the network layer exceeded four the level of application of fuzzy mathematical models and
layers, there was a problem. The traditional backpropagation artificial neural network models has developed rapidly,
algorithm is trained Convergence, a multilayer perceptron with gradually replacing both traditional statistical models and
multiple hidden layers is a deep learning structure(Huang, et al., structural pattern recognition methods(Yang, et al., 2019, Chen,
2019). Deep learning discovers the distribution characteristics et al., 2020a).
of data by combining low-level features to form high-level Table III
attribute categories or features(Shokri and Shmatikov, 2015, Research Fields of AI
Zhang, et al., 2019c). Important artificial neural network Fields Explanation
algorithms include the perception neural network (PNN), back Expert System Expert system is a computer intelligent
propagation (BP), the self-organizing network (SON), the program system with professional
self-organizing map (SOM), and learning vector quantization knowledge and experience. By modeling
(LVQ)(Yuan, et al., 2010, Jamshidi, et al., 2020) the ability of human experts to solve
problems, it uses knowledge
C. The Robotics representation and knowledge reasoning
A robot is a machine that can simulate human behavior. The techniques in artificial intelligence to
research into robots has experienced three generations of simulate complex problems that are
development(Liu, et al., 2018). usually solved by experts. It can reach the
The first generation program controls the robot. This kind of same level of problem-solving skills as
robot can be programmed by the designer, and then is stored in an expert can.
the robot and works under the control of the program. Before Machine Machine learning is a multi-disciplinary
the robot performs a task for the first time, the technician will Learning interdisciplinary, involving probability
instruct the robot to perform the operation, and the robot will theory, statistics, approximation theory,
perform the entire operation step by step(Li and Li, 1999; convex analysis, algorithm complexity
Yang, et al., 2014; Lu, et al., 2018). Each operation recorded on theory and other disciplines. Machine
the ground is represented as an instruction. The second learning specializes in the study of how
generation is adaptive robots (Zhang, et al., 2019d). This kind computers simulate or realize human
of robot is equipped with corresponding sensory sensors (such learning behaviors in order to acquire
as vision, hearing, and tactile sensors), which can acquire new knowledge or skills and reorganizes
simple information (such as about the working environment the existing knowledge structure to
and operating objects). The robot is processed by a computer to continuously improve its own
control the operation activities. The third generation is performance.
intelligent robots(Shone, et al., 2018). The intelligent robot has Robotics Robotics is the science related to the
human-like intelligence and is equipped with high-sensitivity design, manufacture and application of
sensors(Sodhro, et al., 2019). Its sensory ability exceeds that of robots. It mainly studies the relationship
ordinary people. The robot can analyze the information that it between robot control and processing
perceives, control its behavior, respond to changes in the objects.
environment, and complete complex tasks(Sünderhauf, et al., Decision Decision support system belongs to the
2018;Tambe, et al., 2019). Support category of management science and has
D. The Intelligent Decision Support System System an extremely close relationship with
"knowledge-intelligence".
The decision support system belongs to the category of
Pattern Pattern recognition is the study of how to
management science and has an extremely close relationship
Recognition make machines have perception
with "knowledge-intelligence." In the 1980s, the expert system
capabilities. It mainly studies the
was successful in many ways(Abduljabbar, et al., 2019). AI,
recognition of visual and auditory
especially the application of intelligence and knowledge
patterns.
processing technology in decision support systems, has
expanded the application scope of decision support systems and
has improved the system’s ability to solve problems. It V. APPLICATION SCENARIOS OF AI
becomes an intelligent decision support system(Cayamcela, et Based on the relatively mature development of technical
conditions such as data, algorithms, and computing
capabilities, AI has begun to truly solve problems and to management manpower and material resources, establish a
effectively create economic benefits(Zhang, et al., 2020). From scientific and reasonable risk management system, and lay the
an application perspective, industries with a good data foundation for the company's development(Duan, et al., 2019).
foundation (such as finance, healthcare, automotive, and retail)
C. The Health Industry
have relatively mature AI application scenarios.
In the medical field, AI-related algorithms are used to
A. The Automotive Industry provide medical assistance, to detect cancer, and to develop
In the automotive industry, let’s consider autonomous new drugs(Ravì, et al., 2016). Extensive promotion of medical
driving as an example. Autonomous driving is a product of the information is of great significance to the development of
in-depth integration of new generation information medical undertakings on a global scale(Hametand Tremblay,
technologies such as the automotive industry, AI, and Internet 2017). One of the most famous is undoubtedly Watson, IBM's
of Things(Liand Xu,2001). This is an important direction for intelligent robot. The IBM technical team first input a large
current global transportation and for travel information and amount of data and information into Watson. This huge
contacts. Autonomous driving uses sensors like lidar, as well as database includes medical information and reports, clinical
other sensors, to collect road conditions and pedestrian guidelines, drug use reports, and thousands of patient medical
information and combines this information with advanced AI records. Since then, AI algorithms have been used for analysis
algorithms to continuously optimize and, ultimately, to offer and processing, in order to provide stakeholders with medical
the best route and control plan for vehicles on the road(Ding, et assistance and to perform medical diagnosis more effectively
al., 2020). and precisely(Esteva, et al., 2019).
China is closing the gap between Europe and the United
D. The Retailing Industry
States in the field of driverless cars, and it is even achieving
simultaneous development(Tong, et al., 2019). On December In the retail industry, offline physical retail stores use AI to
21, 2014, Google announced the first driverless prototype car. achieve true unmanned retail, thereby reducing costs and
Switzerland and France jointly manufactured driverless buses greatly improving efficiency. AmazonGo, a smart physical
in 2015 and planned to conduct a two-year road test. In 2017, retail store established by e-commerce giant Amazon, has
German Audi released its new "Audi AI" trademark and added firepower to smart retail in a very short period of
applied AI to the Olympic Games. The fixed-speed cruise time(Lu, et al., 2013). A technology in AmazonGo called "Just
automatic parking and other auxiliary drive devices installed on Walk Out" combines machine learning, computer vision, and
the car have realized automatic driving to a certain extent, and sensors. By distributing sensors, cameras, and signal receivers
the driverless car has completely liberated human behavior in the store, it can monitor the placement and the removal of
when driving(Došilović, et al., 2018, Wang, et al., 2018). goods on shelves, as well as monitor the goods in the virtual
shopping cart. The application of AI in the recommendation
B. The Financial Markets system will increase online sales, achieve more accurate market
AI has been successfully applied in the financial markets; forecasts, and reduce inventory costs. The recommendation
example uses include intelligent risk control, intelligent system builds an online product recommendation model based
consulting, market forecasting, and credit rating, among on the user's potential preferences, and it has been applied on
others(Wu, et al., 2020). This has brought finance into a new many e-commerce websites(Erokhin, 2019).
era of innovation through AI embeddedness. Some
E. The Media Industry
cutting-edge Internet companies in Silicon Valley are trying to
use AI algorithms to lower the threshold for users to accept In the media industry, content communication robots and
financial products. Models are trained based on the knowledge brand communication robots offer one-click generation of
and experience of financial analysts and are applied to track content that users want, and up to 10,000 articles can be
customer needs and to minimize costs(Feng, Li, and Chen, published in one minute(Nunavathand Goodwin, 2018). This
2001; Yang, et al., 2001b; Li, Li, and Wang, 2003;). A Japanese intelligent media platform based on artificial intelligence can
startup company (Alpaca) uses deep learning to analyze and combine current hot events, public opinion, and public relations
recognize images, aiming to help users quickly find foreign marketing content, can study media delivery and delivery rules,
exchange trading charts from a large amount of and can automatically generate content that users want to
information(Khalaf, et al., 2019). read(Misra, et al., 2020). It can intelligently connect to
In the financial sector, the AI market is growing. Through the mainstream media platforms and automatically send
technical means of machine learning, it can predict both risks information synchronously, so as to achieve effective
and the direction of the stock market. Financial institutions dissemination. In terms of brand promotion, the intelligent
employ machine learning methods to manage financial risks, to platform estimates and matches media products and media
integrate multiple data sources, and to provide people with channels based on brand content, promotion budgets, and
real-time risk warning information(Chen, et al., 2020b). Also, promotion effects, so as to bring the company the greatest
financial institutions adopt big data to analyze the communication value (Haenleinand Kaplan, 2019).
corresponding financial risks, provide real-time risk warning of F. Smart Payment Systems
corresponding financial assets, save investment and financial Nowadays, it has become a habit of many people to shop
without cash. A mobile phone can easily complete the payment. are the primitive aspect of human cognition, and the process of
If customers have not used the scan code to pay for items, human cognition is the process of symbolic calculation and
another new payment method is voiceprint payment and face reasoning. Symbolism first expresses people's cognitive objects
scanning. Shoppers don’t need to bring their wallets when they in the form of symbols through mathematical logic, and then
go out, don’t need to scan a QR code, and don’t even need to uses the computer's own symbol processing capabilities to
enter a password (Farivar, et al., 2019). They can pay through simulate human cognitive processes. The supporting symbol
via their own voice and face. Among them, voiceprint principle is mainly the assumption of the physical symbol
recognition is a new technology developed based on biological system and the principle of limited rationality (Jhaand Topol,
characteristics (KullayaSwamyand Sarojamma, 2020). For 2016).
example, Alipay Pay is an AI-based payment system that The main research content of symbolism is knowledge
applies in KFC restaurants. Two identical twins can be representation and reasoning technology, based on logic. Its
distinguished quickly within a few seconds, without mistake. main representatives are Newell, Simon, McCarthy, Robinson,
The technique applied is the eye brushing technique. Each and Shot field (Wang, et al., 2007). The main research results
person’s eye pattern is different. Users brush their faces for include analytical reasoning methods, heuristic algorithms,
eating, brush their faces for driving, brush their faces for expert systems, knowledge engineering theory, and technology,
lodging; the application range of brushing one’s face is getting among others. Symbolism has achieved great success in solving
wider and wider. logical problems. For example, artificial intelligence solves the
"four-color conjecture" problem that has not been proved by
G. Smart Home
hand, as well as the successful development and application of
Smart homes use advanced technology to integrate facilities expert systems in the 1970s (Jin, et al., 2014).
related to daily life to create efficient residential facilities and
daily family affairs management systems, in order to make life B. Connectionism
more comfortable (Finogeev, et al., 2019.). A smart home Also known as the Bionics, connectionism’s main principle
involves many aspects of household products; TV, bathroom, is that human intelligence depends on the physiological
refrigerator, air conditioner, door lock and other products are structure and the working methods of the human brain.
all-encompassing, serving users through smart Connectionism believes that the basic unit of human cognition
intercommunication (Qelaand Mouftah, 2012). A complete is the neurons of the human brain and that the cognitive process
smart home system is not only a device – it is also a is the process of information processing by the human brain
combination of many home products with different functions. (Fu, et al., 2018, Wang, et al., 2020). Therefore, connectionism
The user in a family is not one person, but multiple users. The advocates imitating the human brain from the structure and
goal of smart home systems is to efficiently and intelligently working mode, to truly realize the simulation of human
coordinate household products and people into a unified system intelligence on the machine. The main content of
that is able to learn, connect, and adapt to itself (Ghahramani, connectionism research is neural networks. Its main
2015, Yigitcanlar, et al., 2020.). representatives include McCleland and Rumelhart. The main
Compared with the interactive form of buttons and touch research results are brain model research and the back
screens, voice assistant hardware is more convenient, and propagation (B-P) algorithm in the multi-layer network
currently, voice control has become an important entry point (Dwivedi, et al., 2019).
for smart homes (Sun, et al., 2014, Sun, et al., 2017). Large
C. Behaviorism
internet companies and technology companies rely on their
own advantages to enter the smart speaker market to seize Behaviorism is also called the evolution of cybernetics. The
market dividends. Smart speakers of various brands in the main principle is that intelligence depends on perception and
market are considered to be the gateway to the control of smart action (Shi, et al., 2020). It does not require knowledge,
homes in the field of AI (Srivastava, et al., 2017). The representation, or reasoning. Behaviorism believes that the
important function of the smart speaker is to interact with the basic human ability is the ability to act, perceive, and sustain
human voice. People interact with smart speakers through life and self-replication. Intelligent behavior is embodied
voice, so smart speakers can understand people's needs and through interaction with the real-world environment; AI should
perform services that people need (Liand Du, 2007). be realized through gradual evolution, like human intelligence
(Haydariand Yilmaz, 2020). It has nothing to do with the
VI. THREE VIEWPOINTS OF AI representation of and the reasoning of knowledge. The focus of
behaviorism is to simulate various human control behaviors. Its
Symbolism, connectionism, and behaviorism represent the
main representative is Brooke. The main development result is
three main viewpoints of concepts in the field of AI research.
the realization of intelligent control and intelligent robot
They are the most important theoretical theorems in the
systems. Behaviorism has not yet formed a complete theoretical
development of AI disciplines, and they are the theoretical
system but, because it is completely different from the
foundation for the development of AI disciplines.
traditional AI viewpoint, it has attracted a lot of attention of the
A. Symbolism AI community (Tokui, et al., 2015, Wanigasekara, et al., 2021).
Symbolism is also called logicism. It believes that symbols The above three viewpoints have critical significance and
far-reaching influence on the development of artificial Chollet, F. 2017. ―Xception: Deep learning with depthwise separable
convolutions.‖ In Proceedings of the IEEE conference on computer vision and
intelligence. It can be said that it is under the guidance of these pattern recognition, 1251-1258.
three viewpoints that various specific artificial intelligence Collins, G.S. and Moons, K.G. 2019. ―Reporting of artificial intelligence
analysis models, implementation methods, and algorithms have prediction models.‖ The Lancet, 393 (10181): 1577-1579.
Davenport, T.H. and Ronanki, R. 2018. ―Artificial intelligence for the real
been born (Zhang, et al., 2019, Zhangand Tao, 2020). Even the
world.‖ Harvard business review, 96(1): 108-116.
latest data mining, knowledge discovery, and intelligent agent Dhingra, M., Jain, M., and Jadon, R.S. 2016. ―Role of artificial intelligence in
technologies are deeply influenced by the above viewpoints. enterprise information security: A review.‖ In 2016 Fourth International
Conference on Parallel, Distributed and Grid Computing (PDGC), 188-191.
IEEE.
VII. CONCLUSION Ding,R., Palomares, I., Wang,X., Yang,G., Liu, B., Dong,Y., Enrique H., and
The study presents a systematic overview of AI that focuses Herrera, F. 2020.―Large-Scale decision-making: Characterization, taxonomy,
challenges and future directions from an Artificial Intelligence and applications
on prospects and development, core techniques, applicable perspective.‖ Information Fusion, 59: 84-102.
scenarios, and challenges. This paper conducts a Došilović, F.K., Brčić, M. and Hlupić, N. 2018. ―Explainable artificial
state-of-the-art review of the extant and upcoming research on intelligence: A survey.‖ In 2018 41st International convention on information
AI. and communication technology, electronics and microelectronics (MIPRO),
0210-0215. IEEE.
Artificial intelligence is an interdisciplinary subject that Duan, N., Liu, L. Z., Yu, X. J., Li, Q., and Yeh, S. C. 2019. ―Classification of
involves information, logic, cognition, thinking, systems, and multichannel surface-electromyography signals based on convolutional neural
biology. It has been used for knowledge processing, pattern networks.‖ Journal of Industrial Information Integration, 15: 201-206.
recognition, machine learning, and natural language Duan, L., Xu, L. 2012. ―Business Intelligencefor Enterprise Systems: a
Survey.‖ IEEE Transactions on Industrial Informatics, 8(3): 679-687.
processing. Applications have been achieved across many Duan, L., Xu, L., Liu, Y., and Lee, J. 2009. ―Cluster-based Outlier Detection.‖
fields, such as automatic programming, expert systems, Annals of Operations Research, 168 (1): 151-168.
knowledge system, and intelligent robots. AI not only requires Duan, Y., Edwards, J.S. and Dwivedi, Y.K. 2019. ―Artificial intelligence for
logical thinking and imitation, but emotion is also an decision making in the era of Big Data–evolution, challenges and research
agenda.‖ International Journal of Information Management, 48: 63-71.
indispensable part of it. The next breakthrough in the field of AI Dwivedi, Y.K.,Hughes, L., Ismagilova, E., Aarts, G., Coombs, C., Crick, T.,
can not only give computers more logical reasoning capabilities Duan, Y., Dwivedi, R., Edwards, J., Eirug, A., Galanos, V., Ilavarasan, P.V.,
but can give them emotional capabilities. The intelligence of Janssen, M., Jones, P., Kar, A.K., Kizgin, H., Kronemann, B., Lal, B., Lucini,
the machine may soon surpass that of human. B., Medaglia, R., Le Meunier-FitzHugh, K., Le Meunier-FitzHugh, L.C.,
Misra, S., Mogaji, E., Sharma, S.K., Singh, J.B., Raghavan, V., Raman, R.,
Rana, N.P., Samothrakis, S., Spencer, J., Tamilmani, K., Tubadji, A., Walton,
P., and Williams, M. D. 2019.―Artificial Intelligence (AI): Multidisciplinary
Acknowledgement: Funding of China National Social perspectives on emerging challenges, opportunities, and agenda for research,
Science (No. 15BJY022), Funding of Researching of China practice and policy.‖ International Journal of Information Management, 2019:
101994.
University of Labor Relations (No. JG2106, No. ZYTS201806) Erhan, D., Courville, A., Bengio, Y. and Vincent, P. 2010. ―Why does
unsupervised pre-training help deep learning?‖ In Proceedings of the thirteenth
REFERENCES international conference on artificial intelligence and statistics. JMLR
Workshop and Conference Proceedings, 201-208.
Abduljabbar, R., Dia, H., Liyanage, S., and Bagloee, S.2019. ―Applications of Erokhin, S.D. 2019. ―A review of scientific research on artificial intelligence.‖
Artificial Intelligence in Transport: An Overview.‖ Sustainability, 11(1): In 2019 Systems of Signals Generating and Processing in the Field of on Board
189.Adadi, A. and Berrada, M. 2018. ―Peeking inside the black-box: a survey Communications, 1-4. IEEE.
on explainable artificial intelligence (XAI).‖ IEEE access, 6: 52138-52160. Esteva, A., Robicquet, A., Ramsundar, B., Kuleshov, V., DePristo, M., Chou,
Baryannis, G., Validi, S., Dani, S. and Antoniou, G. 2019. ―Supply chain risk K., Cui, C., Corrado, G., Thrun, S. and Dean, J. 2019. ―A guide to deep learning
management and artificial intelligence: state of the art and future research in healthcare.‖ Nature medicine, 25 (1): 24-29.
directions.‖ International Journal of Production Research, 57(7): 2179-2202. Farivar, F., Haghighi, M.S., Jolfaei, A. and Alazab, M. 2019. ―Artificial
Bose, B.K. 2017. ―Artificial intelligence techniques in smart grid and intelligence for detection, estimation, and compensation of malicious attacks in
renewable energy systems—Some example applications.‖ Proceedings of the nonlinear cyber-physical systems and industrial IoT.‖ IEEE transactions on
IEEE, 105 (11): 2262-2273. industrial informatics, 16(4): 2716-2725.
Bostrom, N. and Yudkowsky, E. 2014. ―The ethics of artificial intelligence.‖ Feng, S., Li, L. X., and Cen, L. 2001. ―An object-oriented intelligent design tool
The Cambridge handbook of artificial intelligence, 1: 316-334. to aid the design of manufacturing systems.‖ Knowledge-based systems,
Brynjolfsson, E. and Mcafee, A. 2017. ―The business of artificial intelligence.‖ 14(5-6), 225-232.
Harvard Business Review, 7: 3-11. Finogeev, A., Finogeev, А., Fionova, L., Lyapin, A., &Lychagin, K. A. 2019.
Cai, H., Xu, L., Xu, B., Zhang, P., Guo, J., and Zhang, Y.2018. ―A service ―Intelligent monitoring system for smart road environment.‖ Journal of
governance mechanism based on process miningfor cloud-based applications.‖ Industrial Information Integration, 15: 15-20.
Enterprise Information Systems, 12(10): 1239-1256. Fu, Y., Wang, S., Wang, C.X., Hong, X. and McLaughlin, S. 2018. ―Artificial
Cayamcela, M.E.M. and Lim, W. 2018. ―Artificial intelligence in 5G intelligence to manage network traffic of 5G wireless networks.‖ IEEE
technology: A survey.‖ In 2018 International Conference on Information and Network, 32(6): pp.58-64.
Communication Technology Convergence (ICTC), 860-865). IEEE. Ghahramani, Z. 2015. ―Probabilistic machine learning and artificial
Chi-Hsien, K.and Nagasawa S. 2019. ―Applying machine learning to market intelligence.‖ Nature, 521 (7553): 452-459.
analysis: Knowing your luxury consumer.‖ Journal of Management Analytics, Haenlein, M. and Kaplan, A. 2019. ―A brief history of artificial intelligence: On
6(4): 404-419. the past, present, and future of artificial intelligence.‖ California management
Chen, H., Li, L., and Chen, Y. 2020. ―Explore success factors that impact review, 61 (4): 5-14.
artificial intelligence adoption on telecom industry in China.‖ Journal of Hamet, P. and Tremblay, J. 2017. ―Artificial intelligence in medicine.‖
Management Analytics, 1-33. Metabolism, 69: S36-S40.
Chen, L., Chen, P. and Lin, Z. 2020. “Artificial intelligence in education: a Haydari, A. and Yilmaz, Y. 2020. ―Deep reinforcement learning for intelligent
review.” IEEE Access, 8: 75264-75278. transportation systems: A survey.‖ IEEE Transactions on Intelligent
Chen, X.W. and Lin, X. 2014. ―Big data deep learning: challenges and Transportation Systems. Early Access.
perspectives.‖ IEEE access, 2: 514-525. Hu, R.Q. and Hanzo, L. 2019. ―Twin-timescale artificial intelligence aided
mobility-aware edge caching and computing in vehicular networks.‖ IEEE
Transactions on Vehicular Technology, 68 (4): 3086-3099.
Hu, S., Liang, Y.C., Xiong, Z. and Niyato, D. 2021. ―Blockchain and Artificial Miotto, R., Wang, F., Wang, S., Jiang, X., and Dudley, J.T. 2018. ―Deep
Intelligence for Dynamic Resource Sharing in 6G and Beyond.‖ IEEE Wireless learning for healthcare: review, opportunities and challenges.‖ Briefings in
Communications. Early Access. bioinformatics, 19 (6): 1236-1246.
Huang, B., Huan, Y., Xu, L., Zheng, L., and Zou, Z. 2019. ―Automated trading Morocho-Cayamcela, M.E., Lee, H. and Lim, W. 2019. ―Machine learning for
systems statistical and machine learningmethods and hardware 5G/B5G mobile and wireless communications: Potential, limitations, and
implementation: a survey.‖ Enterprise Information Systems, 13(1): 132-144. future directions.‖ IEEE Access, 7: 137184-137206.
Huang, C., Cai, H., Xu, L., Xu, B., Gu, Y., and Jiang, L. 2019. ―Data-driven Najafabadi, M.M., Villanustre, F., Khoshgoftaar, T.M., Seliya, N., Wald, R.
ontology generation and evolution towards intelligent servicein manufacturing and Muharemagic, E. 2015. ―Deep learning applications and challenges in big
systems.‖ Future Generation Computer Systems, 101: 197-207. data analytics.‖ Journal of big data, 2(1): 1-21.
Jamshidi, M. B., Lalbakhsh, A., Talla, J., Peroutka, Z., Hadjilooei, F., Neyshabur, B., Bhojanapalli, S., McAllester, D. and Srebro, N. 2017.
Lalbakhsh, P, and Mohyuddin, W.2020. ―Artificial Intelligence and ―Exploring generalization in deep learning.‖ arXiv preprint arXiv:1706.08947.
COVID-19: Deep Learning Approaches for Diagnosis and Treatment.‖ IEEE Nilsson, N. J. 2014. ―Principles of artificial intelligence.‖ Morgan Kaufmann.
Access,8, 109581-109595. Publishers, Inc., Palo Alto, California.
Jha, S. and Topol, E.J. 2016. ―Adapting to artificial intelligence: radiologists Nunavath, V. and Goodwin, M., 2018. ―The role of artificial intelligence in
and pathologists as information specialists.‖ Jama, 316 (22): 2353-2354. social media big data analytics for disaster management-initial results of a
Jin, C., Li, F., Wilamowska‐Korsak, M., Li, L., and Fu, L. 2014. ―BSP‐GA: systematic literature review.‖ In 2018 5th International Conference on
A new Genetic Algorithm for System Optimization and Excellent Schema Information and Communication Technologies for Disaster Management
Selection.‖ Systems Research and Behavioral Science, 31(3): 337-352. (ICT-DM) (pp. 1-4). IEEE.
Khalaf, B.A., Mostafa, S.A., Mustapha, A., Mohammed, M.A. and Abduallah, Pouyanfar, S., Sadiq, S., Yan, Y., Tian, H., Tao, Y., Reyes, M.P., Shyu, M.L.,
W.M. 2019. ―Comprehensive review of artificial intelligence and statistical Chen, S.C. and Iyengar, S.S. 2018. ―A survey on deep learning: Algorithms,
approaches in distributed denial of service attack and defense methods.‖ IEEE techniques, and applications.‖ ACM Computing Surveys (CSUR), 51: (5):
Access, 7: 51691-51713. 1-36.
KullayaSwamy, A. and Sarojamma, B. 2020. ―Bank transaction data modeling Qela, B. and Mouftah, H.T. 2012. ―Observe, learn, and adapt (OLA)—An
by optimized hybrid machine learning merged with ARIMA.‖ Journal of algorithm for energy management in smart homes using wireless sensors and
Management Analytics, 7 (4): 624-648. artificial intelligence.‖ IEEE transactions on smart grid, 3(4): 2262-2272.
Li, D. and Du, Y. 2007. ―Artificial intelligence with uncertainty.‖ CRC press. Qi, J., Wu, F., Li, L., and Shu, H. 2007. ―Artificial intelligence applications in
Taylor & Francis Group, New York. the telecommunications industry.‖ Expert Systems, 24(4), 271-291.
Li, H. X., Li, L. X., and Wang, J. Y. 2003. ―Interpolation representation of Rajkomar, A., Oren, E., Chen, K., Dai, A.M., Hajaj, N., Hardt, M., Liu, P.J.,
feedforward neural networks.‖ Mathematical and computer modelling, 37(7-8), Liu, X., Marcus, J., Sun, M. and Sundberg, P. 2018. ―Scalable and accurate
829-847. deep learning with electronic health records.‖ NPJ Digital Medicine, 1(1): 1-10.
Li, H.X. and Xu, L.D. 2001. ―Feature space theory—a mathematical foundation Ravì, D., Wong, C., Deligianni, F., Berthelot, M., Andreu-Perez, J., Lo, B. and
for data mining.‖ Knowledge-based systems, 14(5-6): 253-257. Yang, G.Z. 2016. ―Deep learning for health informatics.‖ IEEE journal of
Li, H. and Li, L. 1999. ―Representing diverse mathematical problems using biomedical and health informatics, 21(1): 4-21.
neural networks in hybrid intelligent systems.‖ Expert Systems, 16(4), Reichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J.
262-272. andCarvalhais, N.2019. ―Deep learning and process understanding for
Li, L. 1999a. ―Proposing an architectural framework of a hybrid knowledge‐ data-driven Earth system science.‖ Nature, 566(7743): 195–204.
based system for production rescheduling.‖ Expert Systems, 16(4), 273-279. Shi, F., Wang, J., Shi, J., Wu, Z., Wang, Q., Tang, Z., He, K., Shi, Y. and Shen,
Li, L. 1999b. ―Knowledge-based problem solving: an approach to health D., 2020. ―Review of artificial intelligence techniques in imaging data
assessment.‖ Expert Systems with Applications, 16(1), 33-42. acquisition, segmentation and diagnosis for covid-19.‖ IEEE reviews in
Li, L., Warfield, J., Guo, S. J., Guo, W. D., and Qi, J. Y. 2007. ―Introduction: biomedical engineering, 4-15, IEEE.
Advances in intelligent information processing.‖ Information Systems, 32(7), Shi, Z., Huang, Y., He, Q., Xu, L., Liu, S., Qin, L., Jia, Z., Li, J., Huang, H. and
941-943. Zhao, L. 2007. ―MSMiner—a developing platform for OLAP.‖ Decision
Litjens, G., Kooi, T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, Support Systems, 42(4): 2016-2028.
M., Van Der Laak, J.A., Van Ginneken, B. and Sánchez, C.I. 2017. ―A survey Shokri, R. and Shmatikov, V. 2015. ―Privacy-preserving deep learning.‖ In
on deep learning in medical image analysis.‖ Medical image analysis, 42: Proceedings of the 22nd ACM SIGSAC conference on computer and
60-88. communications security, 1310-1321.
Liu, J., Kong, X., Xia, F., Bai, X., Wang, L., Qing, Q. and Lee, I. 2018. Shone, N., Ngoc, T.N., Phai, V.D. and Shi, Q. 2018. ―A deep learning approach
―Artificial intelligence in the 21st century.‖ IEEE Access, 6: 34403-34421. to network intrusion detection.‖ IEEE transactions on emerging topics in
Liu, L., Ouyang, W., Wang, X., Fieguth, P., Chen, J., Liu, X. and Pietikäinen, computational intelligence, 2 (1): 41-50.
M. 2020. ―Deep learning for generic object detection: A survey.‖ International Sodhro, A.H., Pirbhulal, S. and De Albuquerque, V.H.C. 2019. ―Artificial
journal of computer vision, 128(2): 261-318. intelligence-driven mechanism for edge computing-based industrial
Liu, Z., Luo, P., Wang, X. and Tang, X. 2015. ―Deep learning face attributes in applications.‖ IEEE Transactions on Industrial Informatics, 15(7): 4235-4243.
the wild.‖ In Proceedings of the IEEE international conference on computer Srivastava, S., Bisht, A. and Narayan, N., 2017. ―Safety and security in smart
vision, 3730-3738. cities using artificial intelligence—A review.‖ In 2017 7th International
Lu, F., Yamamoto, K., Nomura, L.H., Mizuno, S., Lee, Y. and Thawonmas, R. Conference on Cloud Computing, Data Science & Engineering-Confluence,
2013. ―Fighting game artificial intelligence competition platform.‖ In 2013 130-133, IEEE.
IEEE 2nd Global Conference on Consumer Electronics (GCCE): 320-323. Sun, C., Shrivastava, A., Singh, S. and Gupta, A. 2017. ―Revisiting
IEEE. unreasonable effectiveness of data in deep learning era.‖ In Proceedings of the
Lu, L., Xu, L., Xu, B., Li, G., and Cai, H. 2018. ―Fog computing approach for IEEE international conference on computer vision, 843-852.
music cognition system based on machine learning algorithm.‖ IEEE Sun, Y., Wang, X. and Tang, X. 2014. ―Deep learning face representation from
Transactions on Computational Social Systems, 5(4), 1142-1151. predicting 10,000 classes.‖ In Proceedings of the IEEE conference on computer
Lu, Y. 2019. ―Artificial intelligence: a survey on evolution, models, vision and pattern recognition, 1891-1898.
applications and future trends.‖ Journal of Management Analytics, 6(1): 1-29. Sünderhauf, N., Brock, O., Scheirer, W., Hadsell, R., Fox, D., Leitner, J.,
Lu, Y. and Xu, L. D. 2018. ―Internet of Things (IoT) cybersecurity research: A Upcroft, B., Abbeel, P., Burgard, W., Milford, M. and Corke, P. 2018. ―The
review of current research topics.‖ IEEE Internet of Things Journal, 6(2): limits and potentials of deep learning for robotics.‖ The International Journal of
2103-2115. Robotics Research, 37(4-5): 405-420.
Makkar, A., Garg, S., Kumar, N., Hossain, M.S., Ghoneim, A. and Alrashoud, Tambe, P., Cappelli, P. and Yakubovich, V. 2019. ―Artificial intelligence in
M. 2020. ―An Efficient Spam Detection Technique for IoT Devices using human resources management: Challenges and a path forward.‖ California
Machine Learning.‖ IEEE Transactions on Industrial Informatics, 17(2): Management Review, 61 (4):15-42.
903-912. Tan, W., Xu, Y., Xu, W., Xu, L., Zhao, X., Wang, L. 2010. ―A Methodology
Misra, N.N., Dixit, Y., Al-Mallahi, A., Bhullar, M.S., Upadhyay, R. and toward Manufacturing Grid-based Virtual Enterprise Operation Platform.‖
Martynenko, A. 2020. ―IoT, big data and artificial intelligence in agriculture Enterprise Information Systems, 4(3): 283-309.
and food industry.‖ IEEE Internet of Things Journal. Early Access. Tong, W., Hussain, A., Bo, W.X. and Maharjan, S. 2019. ―Artificial
intelligence for vehicle-to-everything: A survey.‖ IEEE Access, 7:
10823-10843.
Tokui, S., Oono, K., Hido, S. and Clayton, J. 2015. ―Chainer: a next-generation Zhang,C. and Fu,W.2021. ―Optimal Model for Patrols of UAVs in Power Grid
open source framework for deep learning.‖ In Proceedings of workshop on under Time Constraints.‖ International Journal of Performability Engineering,
machine learning systems (LearningSys) in the twenty-ninth annual conference 17(1): 103-113.
on neural information processing systems (NIPS), 5: 1-6. Zhang, C., Patras, P. and Haddadi, H. 2019a. ―Deep learning in mobile and
Tung, K. 2019. ―AI, the internet of legal things, and lawyers‖. Journal of wireless networking: A survey.‖ IEEE Communications surveys & tutorials,
Management Analytics, 6(4): 390-403. 21(3): 2224-2287.
Wanigasekara, C., Oromiehie, E., Swain, A., Prusty, B. G., and Nguang, S. Zhang, C., Xu, X., and Chen, H. 2019b. ―Theoretical Foundations and
K.2021. ―Machine learning-based inverse predictive model for AFP based Applications of Cyber-Physical Systems.‖ Journal of Library Hi Tech, 38(1):
thermoplastic composites.‖ Journal of Industrial Information Integration, 22: 95–104.
100197. Zhang, J. and Tao, D., 2020. ―Empowering Things with Intelligence: A Survey
Wang, C., Xu, L., and Peng, W. 2007. ―Conceptual design of remote of the Progress, Challenges, and Opportunities in Artificial Intelligence of
monitoring and fault diagnosis systems.‖ Information Systems, 32(7): Things.‖ IEEE Internet of Things Journal. Early Access.
996-1004. Zhang, L., Liang, Y.C. and Niyato, D. 2019. ―6G Visions: Mobile
Wang, C.X., Di Renzo, M., Stanczak, S., Wang, S., and Larsson, E.G. 2020. ultra-broadband, super internet-of-things, and artificial intelligence.‖ China
―Artificial intelligence enabled wireless networking for 5G and beyond: Recent Communications, 16 (8): 1-14.
advances and future challenges.‖ IEEE Wireless Communications, 27(1): Zhang, S., Yao, L., Sun, A. and Tay, Y. 2019c. ―Deep learning-based
16-23. recommender system: A survey and new perspectives.‖ ACM Computing
Wang, J., Ma, Y., Zhang, L., Gao, R.X. and Wu, D. 2018. ―Deep learning for Surveys (CSUR), 52(1): 1-38.
smart manufacturing: Methods and applications.‖ Journal of Manufacturing Zhang, W. Q., Xiang, Y. D., Liu X. H.,and Zhang P. Z. 2019d. ―Domain
Systems, 48: 144-156. ontology development of knowledge base in cardiovascular personalized health
Wang, L., Zou, H., Su, J., Li, L., and Chaudhry, S. 2013. ―An ARIMA‐ANN management‖.Journal of Management Analytics, 6(4): 420-455.
hybrid model for time series forecasting.‖ Systems Research and Behavioral Zhang, Z., Cui, P. and Zhu, W. 2020. ―Deep learning on graphs: A survey.‖
Science, 30(3), 244-259. IEEE Transactions on Knowledge and Data Engineering. Early Access.
Wu, H., Han, H., Wang, X. and Sun, S. 2020. ―Research on Artificial Zhao, R., Yan, R., Chen, Z., Mao, K., Wang, P., and Gao, R.X. 2019. ―Deep
Intelligence Enhancing Internet of Things Security: A Survey.‖ IEEE Access, learning and its applications to machine health monitoring.‖ Mechanical
8: 153826-153848. Systems and Signal Processing, 115: 213-237.
Xu, L. 1999. ―Preface.‖ Expert Systems with Applications, 16(1): 1-2. Zhao, Z.Q., Zheng, P., Xu, S.T. and Wu, X. 2019. ―Object detection with deep
Xu, L. 2013. ―Introduction: Systems Science in Industrial Sectors.‖ Systems learning: A review.‖ IEEE transactions on neural networks and learning
Research and Behavioral Science, 30(3): 211-213. systems, 30 (11): 3212-3232.
Xu, L. D., Lu, Y. and Li, L. 2021. ―Embedding Blockchain Technology into
IoT for Security: A Survey.‖ IEEE Internet of Things Journal. Early
Access.DOI: 10.1109/JIOT.2021.3060508.
Xu, L., Tan, W., Zhen, H. and Shen, W. 2008. ―An approach to enterprise
process dynamic modeling supporting enterprise process evolution.‖
Information Systems Frontiers, 10(5): 611-624.
Yang, B., Li, L. X., Ji, H., and Xu, J. 2001a. ―An early warning system for loan
risk assessment using artificial neural networks.‖ Knowledge-Based Systems,
14(5-6), 303-306.
Yang, B., Li, L. X., Xie, Q., and Xu, J. 2001b. ―Development of a KBS for
managing bank loan risk.‖ Knowledge-Based Systems, 14(5-6), 299-302.
Yang, G., Xie, L., Mäntysalo, M., Zhou, X., Pang, Z., Xu, L., ... and Zheng, L.
R. 2014. ―A health-IoT platform based on the integration of intelligent
packaging, unobtrusive bio-sensor, and intelligent medicine box.‖ IEEE
Transactions on industrial informatics, 10(4), 2180-2191.
Yang, H., Wen, J., Wu, X., He, L. and Mumtaz, S. 2019. ―An efficient edge
artificial intelligence multipedestrian tracking method with rank constraint.‖
IEEE Transactions on Industrial Informatics, 15(7): 4178-4188.
Yigitcanlar, T., Desouza, K.C., Butler, L. and Roozkhosh, F. 2020.
―Contributions and risks of artificial intelligence (AI) in building smarter cities:
Insights from a systematic review of the literature.‖ Energies, 13(6): 1473.
Young, T., Hazarika, D., Poria, S. and Cambria, E. 2018. ―Recent trends in
deep learning based natural language processing.‖IEEE Computational
Intelligence magazine, 13(3): 55-75.
Yuan, R., Li, Z., Guan, X., and Xu, L. 2010. ―An SVM-based Machine
Learning Method for Accurate Internet Traffic Classification.‖ Information
Systems Frontiers, 12(2): 149-156.
Zappone, A., Di Renzo, M. and Debbah, M. 2019. ―Wireless networks design
in the era of deep learning: Model-based, AI-based, or both?‖ IEEE
Transactions on Communications, 67(10): 7331-7376.
Zeng, L., Li, L., and Duan, L. 2012. ―Business intelligence in enterprise
computing environment.‖ Information Technology and Management, 13(4):
297-310.
Zhang, C. 2020. ―Research on the Economical Influence of the Difference of
Regional Logistics Developing Level in China.‖ Journal of Industrial
Integration and Management, 05(02): 205–223.
Zhang, C. 2019. ―Research on the Fluctuation and Factors of China TFP of IT
Industry.‖ Journal of Industrial Integration and Management, 04(04): 1950013.
Zhang,C. and Chen,Y. 2020. ―A Review of Research Relevant to the Emerging
Industry Trends: Industry 4.0, IoT, Blockchain, and Business Analytics.‖
Journal of Industrial Integration and Management, 01(05): 1950019.
Zhang, C. and Chu, H. 2020. ―Preprocessing Method of Structured Big Data in
Human Resource Archives Database.‖ 2020 IEEE International Conference on
Industrial Application of Artificial Intelligence (IAAI), Harbin, China,
379-384.
 Declaration of interest statement

We declare that we have no financial and

personal relationships with other people or
organizations that can inappropriately
influence our work, there is no professional
or other personal interest of any nature or
kind in any product, service and company
that could be construed as influencing the
position presented in, or the review of the
manuscript entitled A review of Artificial
Intelligence: The State of the Art and Future
Prospects.

Yours Sincerely,

Caiming Zhang