4 learning, and making inferences (or decisions) from incomplete information.
From historical times, humans have been successful in developing systems, which have supported our existence, advancement, and the quality of life. A variety of engineering, scientific, medical, legal, business, agricultural, and educational systems have been designed, developed, and implemented for our benefit. A feature that is indispensable in these systems is the generation of outputs, based on some inputs and the nature of the system itself. The inputs to a system may include information as well as tangible items, and the out-puts may include decisions as well as physical products.
Example 1.1
A typical input variable is identified for each of the following examples of dynamic systems:
(a) Human body: neuroelectric pulses (b) Company: information
(c) Power plant: fuel rate
(d) Automobile: steering wheel movement (e) Robot: voltage to joint motor.
Possible output variables for each of these systems are:
(a) Muscle contraction, body movements (b) Decisions, finished products
(c) Electric power, pollution rate
(d) Front wheel turn, direction of heading (e) Joint motions, effector motion.
Even when the outputs of a system are not decisions themselves, quite commonly, decision-making is involved in generating those outputs. The process of decision-making may range from numerical computations (e.g., in generating control actions in a low-level control algorithm) to complex, knowledge-based decision-making using natural (human) and artificial (machine) intelligence, as in modern decision support systems. This book prim-arily concerns the latter. Engineering systems such as household appliances and consumer electronics, transportation systems, manufacturing operations, factories, chemical plants, power generating systems, food processing facil-ities, and heating, ventilating, and air conditioning (HVAC) systems are increasingly becoming complex ill-defined. Explicit modeling of these would be a considerably difficult task. In some situations, even experimental model-ing (or model identification) is not feasible because the required inputs and outputs of the system are not accessible or measurable. Even when model-ing is possible, the models could be so complex that accurate algorithms for
1.2 Intel lig ent sy s tems
5 decision-making (e.g., generation of control actions) based on these models could considerably increase the costs of computing and hardware, and seri-ously reduce the operating speed, thereby slowing the process and introducing unacceptable time lags and possibly instabilities. Knowledge-based systems with “intelligent” decision-making capabilities are tremendously beneficial in these systems.Future generations of industrial machinery, plants, and decision support systems may be expected to carry out round-the-clock operation, with min-imal human intervention, in manufacturing products or providing services. It will be necessary that these systems maintain consistency and repeatability of operation and cope with disturbances and unexpected variations within the system, its operating environment, and performance objectives. In essence, these systems should have the capability to accommodate rapid reconfigura-tion and adaptareconfigura-tion. For example, a producreconfigura-tion machine should be able to quickly cope with variations ranging from design changes for an existing product to the introduction of an entirely new product line. This will call for tremendous flexibility and some level of autonomous operation in automated machines, which translates into a need for a higher degree of intelligence in the supporting devices. Smart systems will exhibit an increased presence and significance in a wide variety of applications. Products with a “brain” are found, for example, in household appliances, consumer electronics, transportation systems, industrial processes, manufacturing systems, and services. There is clearly a need to incorporate a greater degree of intelligence and a higher level of autonomy into automated machines. This will require the appropriate integration of such devices as sensors, actuators, and controllers, which them-selves may have to be “intelligent” and, furthermore, appropriately distributed throughout the system. Design, development, production, and operation of intelligent machines have been made possible today through ongoing research and development in the field of intelligent systems and control.
1.2.1 Machine intelligence
In the context of machine intelligence, the term machine is normally used to denote a computer or a computational machine. In this sense machine intel-ligence and computational intelintel-ligence are synonymous. Historically machine intelligence, computer intelligence and artificial intelligence (AI) also have come to mean the same thing. Artificial intelligence can be defined as the science of making machines do things that would require intelligence if done by humans. Conventional AI has relied heavily on symbolic manipulation for the processing of descriptive information and “knowledge” in realizing a degree of intelligent behavior. The knowledge itself may be represented in a special high-level language. The decisions that are made through processing of such “artificial” knowledge, perhaps in response to data such as sensory signals, should possess characteristics of intelligent decisions of a human.
Knowledge-based systems and related expert systems are an outcome of the efforts made by the AI community in their pursuit of intelligent computers and intelligent machines. The field of soft computing is important here,
1 Introduction to intel lig ent sy s tems and sof t c omputing
6 developments of which have taken a somewhat different path from tradi-tional AI, yet they have contributed to the general goal of realizing intelligent machines, thereby broadening the meaning of machine intelligence. This topic, which forms the foundation of the present book, will be treated in detail in subsequent chapters.
The first digital computer may be traced back to the development of universal Turing machines in the mid-1930s, but it was in the 1960s that significant activity in the field of artificial intelligence began with the object-ive of developing computers that could think like humans. Just like neurons in the brain, the hardware and software of a computer are themselves not intelligent, yet it has been demonstrated that a computer may be programmed to demonstrate some intelligent characteristics of a human. According to Marvin Minsky of the Massachusetts Institute of Technology, “artificial intelligence is the science of making machines do things that would require intelligence if done by men.” In this context it is the outward characteristics (outputs) of a computer that may be termed “intelligent” rather than any similarity of the computer programs to how a human processes information.
In particular, it is not the physical makeup that is considered in ascertaining whether the machine is intelligent. Specifically, an analogy may be made here with intelligent behavior of a human whose brain, made up of neurons, is a physical entity that assists in realizing that behavior. A considerable effort has gone into the development of machines that somewhat mimic humans in their actions. Because it is the thought process that leads to intelligent actions, substantial effort in AI has been directed at the development of artificial means to mimic the human thought process. This effort is somewhat related to cognitive science. The field of study that deals with analyzing and modeling of the information processing capabilities of a human is known as cognitive science and is important in AI. Nevertheless, in developing an intelligent machine it is not essential to master cognitive science.
1.2.2 Meaning of intelligence
It is useful to first explore the meaning of the term itself. A complete yet simple definition of intelligence is not available. At the basic or atomic level of implementation, what a computer does is quite procedural and cannot be considered intelligent. Similarly, one may argue that the 20 billion neurons in a human brain perform operations that are hardly intelligent when taken individually. It is these same neurons, however, that govern the intelligent behavior of a human. In particular, the attributes of knowledge acquisition – making logical inferences, learning, and dealing with incomplete or qualit-ative information and uncertainty – are all associated with human intelli-gence. Since intelligence is a soft and complex concept, it is rather unrealistic to expect a precise definition for the term. Narrow definitions can result in misleading interpretations, similar to how a group of blind people defined a proverbial elephant.
A definition for intelligence, then, has to be closely related to the char-acteristics or outward “appearance” of an intelligent behavior. This is called
1.2 Intel lig ent sy s tems
7 description through characterization. There exist many characteristics that can be termed intelligent, and partly for this reason it has not been possible to give a precise definition for intelligence. It is the outward characteristics of a system that qualify it to be classified as being intelligent – for example, possession of a memory, ability to learn and thereby gain knowledge and expertise, ability to satisfactorily deal with unexpected and unfamiliar situ-ations, and ability to reason, deduce, and infer from incomplete information.In particular, pattern recognition and classification play an important role in intelligent processing of information. For example, an intelligent system may be able to recognize and acquire useful information from an object that is aged or distorted, having been previously familiar with the original, undis-torted object. Somewhat related are the concepts of approximation, which may lead one to treat the ability to approximate as also a characteristic of intelligence. It is commonly accepted that an intelligent system possesses one or more of the following characteristics and capabilities:
n Sensory perception
n Pattern recognition
n Learning and knowledge acquisition
n Inference from incomplete information
n Inference from qualitative or approximate information
n Ability to deal with unfamiliar situations
n Adaptability to new yet related situations (through expectational knowledge)
n Inductive reasoning
n Common sense
n Display of emotions
n Inventiveness.
Even though significant advances have been made, particularly in the first five capabilities listed above, the current generation of intelligent machines does not claim to have all these capabilities, particularly the last three in the list.
Clearly this list of items does not represent a formal definition of an intelli-gent system. It is known, however, that humans possess these characteristics and capabilities and that humans are intelligent beings. A handwriting recog-nition system is a practical example of an intelligent system. The underlying problem cannot be solved through simple template matching, which does not require intelligence. Because handwriting can vary temporally for the same person, due to various practical shortcomings such as missing characters, errors, nonrepeatability, sensory restrictions, and noise, a handwriting re-cognition system has to deal with incomplete information and unfamiliar characters and should possess capabilities of learning, pattern recognition, and approximate reasoning, which will assist in carrying out intelligent func-tions of the system.