Artificial Intelligence is the natural advance to control theory. One of the main figures that developed the control theory was Norbert Wiener, a German mathematician who worked with Bertrand Russell before. He and his colleagues modelled a regulatory mechanism trying to minimize the error, the difference between the goal state and the current state. More modern approaches try to design regulators as devices that maximize an objective function over time. Artificial Intelligence was considered later as a way to overcome the limits of the mathematics of control theory to face some problems as language, vision and planning.
Nowadays, artificial intelligence has produced an industry and it is being used successfully in several fields of working. It is used for autonomous planning, game playing, autonomous control, medical diagnosis, logistics, robotics, language understanding, and computer vision. On the other hand, it is considered a new science trying to use mathematics again and the scientific method to produce a better comprehension of the process of intelligence.
An agent is something that can gather information about its environment through sensors and can produce an active response through actuators. Any human being can be considered an intelligent agent following this definition. Artificial agents will be defined by their sensor, their actuators and their control rules that produce a certain output from the inputs. Artificial intelligent agents will be able to produce outputs from more complex rules than those ones based on control theory.
Intelligent agents work under a lot of uncertainty. They work trying to maximize the expected performance because the actual performance is always unknown. An intelligent agent learns from the information gathered in order to improve its performance.
Control lets to reduce the complexity of a system because control devices and algorithms increase the probability to preserve the system in a desired state although the number of possible states of the system increases.
A control agent working through control theory rules will not improve its performance to reduce the complexity of the system; however, an intelligent control agent can improve its performance to reduce it through learning.
The Project Management Institute points out three sources of organizational complexity: based on human behavior, based on system behavior and based on ambiguity. The use of artificial agents in the processes of an organization reduces the complexity based on human behavior, however, it can increase the complexity based on the system behavior, because the dynamics of the system changes and it can become unknown, and they do not fit well the complexity from ambiguity.
Ambiguity means: “not knowing what to expect or how to comprehend a situation”. This kind of complexity is very undesirable. Control agents without intelligence cannot cope well with ambiguity, and intelligent agents require a way to measure the expected performance. An intelligent agent can adjust the dynamics of the system through learning; however, it must have an expected performance. Intelligent agents are not omniscient because they cannot know the actual performance but they require a way to measure the expected performance anyway.