Intelligent Systems for Engineers and Scientists

A Practical Guide to Artificial Intelligence

Adrian A. Hopgood

• Publisher: CRC Press
• Year: 2021
• ISBN:
  9781000484120
• Edition: 4
• Topic: Technology & Engineering
• Subtopic: Electrical Engineering & Telecommunications
• Index: Technology & Engineering

The fourth edition of this bestselling textbook explains the principles of artificial intelligence (AI) and its practical applications. Using clear and concise language, it provides a solid grounding across the full spectrum of AI techniques, so that its readers can implement systems in their own domain of interest.

The coverage includes knowledge-based intelligence, computational intelligence (including machine learning), and practical systems that use a combination of techniques. All the key techniques of AI are explained—including rule-based systems, Bayesian updating, certainty theory, fuzzy logic (types 1 and 2), agents, objects, frames, symbolic learning, case-based reasoning, genetic algorithms and other optimization techniques, shallow and deep neural networks, hybrids, and the Lisp, Prolog, and Python programming languages. The book also describes a wide range of practical applications in interpretation and diagnosis, design and selection, planning, and control.

Fully updated and revised, Intelligent Systems for Engineers and Scientists: A Practical Guide to Artificial Intelligence, Fourth Edition features:

• A new chapter on deep neural networks, reflecting the growth of machine learning as a key technique for AI
• A new section on the use of Python, which has become the de facto standard programming language for many aspects of AI

The rule-based and uncertainty-based examples in the book are compatible with the Flex toolkit by Logic Programming Associates (LPA) and its Flint extension for handling uncertainty and fuzzy logic. Readers of the book can download this commercial software for use free of charge. This resource and many others are available at the author's website: adrianhopgood.com.

Whether you are building your own intelligent systems, or you simply want to know more about them, this practical AI textbook provides you with detailed and up-to-date guidance.

Chapter 1 Introduction

DOI: 10.1201/9781003226277-1

1.1 Artificial Intelligence and Intelligent Systems

Over many centuries, tools of increasing sophistication have been developed to serve the human race. Physical tools such as chisels, hammers, spears, arrows, guns, carts, cars, and aircraft all have their place in the history of civilization. Humanity has also developed tools of communication: spoken language, written language, and the language of mathematics. These tools have not only enabled the exchange and storage of information, but they have also allowed the expression of concepts that simply could not exist outside of language.

The last century saw the arrival of a new tool—the digital computer. Computers are able to perform the same sort of numerical and symbolic manipulations that an ordinary person can, but faster and more reliably. They have therefore been able to remove the tedium from many tasks that were previously performed manually and have allowed the achievement of new feats. Such feats range from huge scientific models to the more familiar online banking facilities.

Although these developments are impressive, the computer is actually only performing quite simple operations, albeit rapidly. In such applications, the computer is still only a complex calculating machine. The intriguing idea now is whether we can build a computer (or a computer program) that can think (Turing 1950). As Penrose (1999) has pointed out, most of us are quite happy with machines that enable us to do physical things more easily or more quickly, such as digging a hole or traveling along a freeway. We are also happy to use machines that enable us to do physical things that would otherwise be impossible, such as flying. However, the idea of a machine that can think for us is a huge leap forward in our ambitions, and one that raises many ethical and philosophical questions.

Research in artificial intelligence (or simply AI) is directed toward building such a machine and improving our understanding of intelligence. Here is a simple definition, adapted from Hopgood (2005) by the insertion of the phrase in parenthesis:

Artificial intelligence is the science of mimicking (or exceeding) human mental faculties in a computer.

The ultimate achievement in this field would be to construct a machine that can mimic or exceed all human mental capabilities, including reasoning, understanding, imagination, recognition, creativity, and emotions. We are a long way from achieving that ambition, but some successes have been achieved in mimicking specific areas of human mental activity. For instance, machines are now able to play chess at the highest level, to interpret spoken sentences, to translate foreign languages, and to diagnose medical complaints. An objection to these claimed successes might be that the machine does not tackle these problems in the same way that a human would. A further objection might be that the field of AI should not be limited to human intelligence, but it should include other animal species too. These objections will not be addressed in this book, which is intended as a guide to practical systems and not a philosophical thesis.

In some specialized fields, like interpreting astronomical images, AI can already exceed human capabilities (Hausen and Robertson 2020). In some other fields, the aim is not to produce the best possible form of AI, but rather to mimic human behavior or capabilities. Some examples include an AI opponent in video games, or a model of human responses in cybersecurity development.

In achieving its successes, research into AI, together with other branches of computer science, has resulted in the development of several useful computing tools that form the basis of this book. These tools have a range of potential applications, but this book emphasizes their use in engineering and science. The tools of particular interest can be roughly divided between knowledge-based systems (KBSs), computational intelligence (CI), and hybrid systems. KBSs include expert and rule-based systems, frame-based systems, intelligent agents, and case-based reasoning. CI includes neural networks, genetic algorithms, and other optimization algorithms. Techniques for handling uncertainty, such as fuzzy logic, fit into both categories.

Machine learning describes intelligent systems that improve their performance through experience. Typically, this phrase refers to CI and specifically to complex neural networks that learn to recognize patterns from enormous volumes of data (so-called big data), as described in Chapter 9: Deep Neural Networks. However, there are many other forms of data-driven machine learning, described in Chapters 6–8. Furthermore, some KBSs also have the ability to grow their knowledge, and these techniques are covered in Chapter 5: Symbolic Learning.

Intelligent systems is a broad term, covering a range of AI techniques and the ways in which they can be put into practice. They include applied KBSs, CI, machine learning, and their hybrids. Intelligent systems have not solved the problem of building an artificial mind. Indeed, some would argue that they show little, if any, real intelligence. Nevertheless, they have enabled a range of problems to be tackled that were previously considered too difficult, and they have enabled a large number of other problems to be tackled more effectively. From a pragmatic point of view, these successes alone make intelligent systems interesting and useful.

1.2 A Spectrum of Intelligent Behavior

The definition of AI presented earlier leaves the notion of intelligence rather vague. To explore this further, a spectrum of intelligent behaviors can be drawn based on the level of understanding involved (Hopgood 2003, 2005), as shown in Figure 1.1. The lowest-level behaviors include instinctive reactions, such as withdrawing a hand from a hot object or dodging a projectile. The mid-level behaviors include vision and perception, language and interaction, and common-sense responses to unexpected events. These are capabilities that come naturally to humans, with little conscious thought. High-level behaviors demand specialist expertise such as in the legal requirements of company takeovers or the interpretation of radiological images. Such a spectrum of intelligent behaviors is useful for charting the progress of AI, although it has been criticized for oversimplifying the many dimensions of intelligence (Holmes 2003).

Figure 1.1 A spectrum of intelligent behavior.

Conventional computing techniques have been developed to handle the low-level decision-making and control needed at the low end of the spectrum. Extremely effective computer systems have been developed for monitoring and controlling a variety of equipment. An example of the close regulation and coordination that is possible is demonstrated by various humanoid robots that show human-like mobility (Sakagami et al. 2002; Firth 2007; Gouaillier et al. 2009; Niemüller et al. 2010). As their capabilities for autonomous thought and understanding are improved through technological development, their behaviors can be expected to expand upward from the lower end of the spectrum.

Early AI research, in contrast, began with problems at the high-level end of the spectrum. Two early applications, for example, concerned the specialist areas of mass spectrometry (Buchanan et al. 1969) and bacterial blood infections (Shortliffe 1976). These early triumphs generated great optimism. If a computer could deal with difficult problems that are beyond the capabilities of most ordinary people, it was assumed that more modest human reasoning would be straightforward. Unfortunately, this assumption is false.

The behaviors in the middle of the spectrum, that humans perform with barely a conscious thought, have proven to be the most difficult to emulate in a computer. Consider the photograph in Figure 1.2. Although most of us can recognize the three rabbits in the picture (one of which is a stone statue), the perception involved is an extremely complex behavior. First, recognizing the boundary between objects is difficult. Once an object has been delineated, recognition is far from straightforward. For instance, rabbits come in different shapes, sizes, and colors. They can assume different postures, and they may be partially occluded, like the one in the cage in Figure 1.2. Yet a fully sighted human can perform this perception in an instant, without considering it a particular mark of intelligence. The task’s astonishing complexity is revealed by attempting to perform it with a computer (Hopgood 2003, 2005).

Figure 1.2 The challenges of image recognition and interpretation.

1.3 Knowledge-Based Systems (KBSs)

The principal difference between a KBS and a conventional program lies in its structure. In a conventional program, domain knowledge is intimately intertwined with software for controlling the application of that knowledge. In a KBS, the two roles are explicitly separated. In the simplest case, there are two modules: the knowledge module is called the knowledge base, and the control module is called the inference engine (Figure 1.3). In more complex systems, the inference engine itself may be a KBS containing metaknowledge, that is, knowledge of how to apply the domain knowledge.

Figure 1.3 The main components of a knowledge-based system.

The explicit separation of knowledge from control of its application makes it easier to add new knowledge, either during program development or in the light of experience during the program’s lifetime. There is an analogy with the brain, the control processes of which are approximately unchanging in their nature, like the inference engine. On the other hand, individual behavior is continually modified by new knowledge and experience, l...

Table of contents

• Cover
• Half Title Page
• Title Page
• Copyright Page
• Dedication
• Contents
• Preface
• Acknowledgements
• Author
• Chapter 1 Introduction
• Chapter 2 Rule-Based Systems
• Chapter 3 Handling Uncertainty: Probability and Fuzzy Logic
• Chapter 4 Agents, Objects, and Frames
• Chapter 5 Symbolic Learning
• Chapter 6 Single-Candidate Optimization Algorithms
• Chapter 7 Genetic Algorithms for Optimization
• Chapter 8 Shallow Neural Networks
• Chapter 9 Deep Neural Networks
• Chapter 10 Hybrid Systems
• Chapter 11 AI Programming Languages and Tools
• Chapter 12 Systems for Interpretation and Diagnosis
• Chapter 13 Systems for Design and Selection
• Chapter 14 Systems for Planning
• Chapter 15 Systems for Control
• Chapter 16 The Future of Intelligent Systems
• References
• Index