It is a methodological strategy I take to deal with the complexity of the subject matter. "Intelligence" is a very complicated and subtle concept, and is hard to capture. Therefore it is better to first carefully introduce the concepts to be used to describe and discuss intelligence. We can say that the purpose of this chapter is to prepare the "platform" and "tools" to be used in the whole book.
To fully understand any concept, we not only need to know where it can be applied, but also where it cannot be applied. Since to understand "intelligence" means to draw a line between intelligent systems and non-intelligent systems, we need to first include both of them into a wider concept (their superordinate), to introduce a terminology to describe both types, and then to identify their difference.
In this book, I will use "information system" as the superordinate of "intelligent system". This chapter first introduces the concept of information system, as well as the concepts needed to describe such a system, then in the next chapter we can divide information systems into intelligent ones and non-intelligent ones, according to their differences specified using our terminology.
We usually describe a system by specifying its internal activities and its interactions with external environment. Science has been studying various types of systems: physics on atom and galaxy, chemistry on molecule, biology on living organism, and so on. On each case, a system is studied on its inside, by identifying its components, their relations, and how the relations change over time, as well as on its outside, by identifying its environment, the ways they system interact with the environment, and how the interaction extends over time. We will do the same for information system and intelligence system
An information system is a system whose internal activities and interactions with its environment can be described abstractly as state changes, without specifying the concrete entity and process that carries out the activities and interactions.Furthermore, when a system is seen as an information system, its internal activity is referred to as information processing, which abstractly describes how the components of the system response to each other; its interaction with its environment is referred to as information transferring, which abstractly describes how the state of the system changes when the state of the environment changes. Here informationn is an abstract description of the state or state change in a system or its environment.
According to the above working definition, what makes information system different from the other systems is that it can be described abstractly, without going into the details of the concrete mechanism and process that carry out the activities.
Comparing the above working definition with the other definitions of "information system" (such as those from Wikipedia and Britannica), we can see that what is stressed here is that "Information system" and the related concepts belong to a methodology, not to an ontology. According to this opinion, the correct question is whether a concrete system can be seen as an information system, not whether it is an information system. When a system is referred to as an information system, it does not mean that it is not a physical system, a chemical system, or a biological system (it can be any of them, and many more), but that it can be meaningfully described in an abstract language, without mentioning its physical, chemical, or biological details.
Whether a system can be seen as an information system not only depends on the features of the system, but also depends on the purpose of the observer who use the "information system" methodology. For example, it is reasonable for a linguist to treat a human subject as an information system during a translation experiment, but it is not a good idea for a surgeon to treat the patient in this way during an operation. A computer is an information system to a programmer, but a piece of furniture to a janitor. In both examples, what matters in the former case is the system's activity when described abstractly, while in the latter case, it is the physical or biological details of the system. Similarly, it is natural to describe a chess game in the information system language, since the physical properties of the items involved (such as the materials, size, and weight of the pieces and the board) usually does not matter to the result of the game. On the contrary, it is not the case for a billiards game, where if some physical properties of the items involved get changed, it may change the result.
Therefore, when talking about information system, we always assume the existence of an observer with a specify purpose. In this sense, it is not an "objective nature" of the system. Even so, it does have some common usages. Among naturally formed systems, human beings are often referred to as information systems, and sometimes the term is also applied to some animals. Among artifacts, typical information systems include computers and various control systems. Please note that even though information system is an abstraction, it does not include the "abstract systems" that only exist in concepts, not in reality, such as "Turing Machine".
Even when a system is not treated as an information system, it often can still be "modeled" or "simulated" in an information system. It means that the system can be described at an abstract level, and another system is built that have the same high-level description, though these two systems are completely different at a lower level, which contains essential features of the system to be modeled. For example, a hurricane can be simulated by a computer. However, if they system being modeled is such a system that all of its major properties are shown at the "information-system" level, then we no longer call the above process "modeling" or "simulating", but call it "replicating", "reproducing", or "implementing". For example, when playing chess with a computer, you don't call it "simulated chess", because "chess" is practically defined by its features in an information system, rather than by its physical features. On the contrary, a simulated hurricane is not a hurricane, because it does not have its defining features.
In contrary to people's naive impression, science does not describe its research object as it is, or always includes as much details as possible in the description of the objects. Instead, many general patterns become recognizable only when some details are filtered out as irrelevant. By choosing a vocabulary with a certain granularity, each theory selectively focuses on certain aspects of objects and events by describing them at a certain level of abstraction.
This is especially true for an interdisciplinary theory, which, as the name suggests, studies objects in multiple disciplines with completely different details.
A good example is Cybernetics, inverted by Norbert Wiener (1948). Behind the various differences between animals and machines, Wiener identified important commonality in their control and communication processes. In cybernetics, these processes are described in an abstract language with terms like "control", "feedback", and "communication". Similar examples include the General System Theory of Ludwig von Bertalanffy (1968), the Information Theory of Claude Shannon (1948), and the theory of computational machinery of Alan Turing (1937).
To establish a general theory for intelligence is a problem that is similar to the above ones. Since intelligence may happen in systems with very different details, it has to be described at a level of abstraction, and so does its superordinate, information system.
If an abstract description of information system is necessary, what basic notions should be included in this theory?
Let us start with the notion of state. Described abstractly, an information system has internal states, which changes over time, partially in corresponding to the external states of the environment.
Though every theory can talk about the state of the object under description, the states in an information system is different, in that it is an abstraction. Every state in the human mind can be described in biological terms, and every state in a computer can be describe in physical terms, but as long as they are treated as information systems, those "low-level details" are excluded from the description, and the only basic property of states are that they can be distinguished from each other, by the observer who is taking the system as an information system.
Very often, a state of a system or its environment needs to be analyzed further, but it should still be described using the vocabulary of information system, rather than in that of physics or biology. For example, the state of the whole system can be reduced into a combination of the states of its components, which are also described abstractly.
For states in isolation, there is not too much to be said. For our purpose, the important things are how the states are related to each other, either within or across the system boundary, and how they and their relation change over time. We will address these issues in the following.