intelligent machine[in′tel·ə·jənt mə′shēn]
Any machine that can accomplish its specific task in the presence of uncertainty and variability in its environment. The machine's ability to monitor its environment, allowing it to adjust its actions based on what it has sensed, is a prerequisite for intelligence. The term intelligent machine is an anthropomorphism in that intelligence is defined by the criterion that the actions would appear intelligent if a person were to do it. A precise, unambiguous, and commonly held definition of intelligence does not exist.
Examples of intelligent machines include industrial robots equipped with sensors, computers equipped with speech recognition and voice synthesis, self-guided vehicles relying on vision rather than on marked roadways, and so-called smart weapons, which are capable of target identification. These varied systems include three major subsystems: sensors, actuators, and control. The class of computer programs known as expert systems is included with intelligent machines, even though the sensory input and output functions are simply character-oriented communications. The complexity of control and the mimicking of human deductive and logic skills makes expert systems central in the realm of intelligent machines. See Computer vision, Expert systems, Guidance systems, Robotics
Since the physical embodiment of the machine or the particular task performed by the machine does not mark it as intelligent, the appearance of intelligence must come from the nature of the control or decision-making process that the machine performs. Given the centrality of control to any form of intelligent machine, intelligent control is the essence of an intelligent machine. The control function accepts several kinds of data, including the specification for the task to be performed and the current state of the task from the sensors. The control function then computes the signals needed to accomplish the task. When the task is completed, this also must be recognized and the controller must signal the supervisor that it is ready for the next assignment (see illustration). See Adaptive control, Control systems
Automatic, feedback, or regulatory systems such as thermostats, automobile cruise controls, and photoelectric door openers are not considered intelligent machines. Several important concepts separate these simple feedback and control systems from intelligent control. While examples could be derived from any of the classes of intelligent machines, robots will be used here to illustrate five concepts that are typical of intelligent control. (1) An intelligent control system typically deals with many sources of information about its state and the state of its environment. (2) An intelligent control system can accommodate incomplete or inconsistent information. (3) Intelligent control is characterized by the use of heuristic methods in addition to algorithmic control methods. (A heuristic is a rule of thumb, a particular solution or strategy to be used for solving a problem that can be used for only very limited ranges of the input parameters.) (4) An intelligent machine has a builtin knowledge base that it can use to deal with infrequent or unplanned events. (5) An algorithmic control approach assumes that all relevant data for making decisions is available.