After our last article about the history of the Artificial Intelligence, now it’s necessary examine the classification that – starting from the analysis of the human thinking and learning – gave birth to computational models, which marked the study of this discipline. Investigating the evolution and distinctive elements of these models, will be simpler understand AI development and use nowadays.
Weak and Strong AI: features and definition
From our analysis of the human brain functioning, we can point out some actions and/or typical features of the humankind that Artificial Intelligence should be able to do:
- Make one or more actions autonomously, that is, indiscriminately, compared to a human being;
- Think humanely, solve a problem through specific cognitive functions;
- Rationally think through the use of logic elements, as the human being does;
- Act rationally, that is start a process leading to the best result expected on the base of information and knowledge available. This characteristic is sometimes unconscious, but essential for human rationality;
These elements represent the main fundamentals, because allow to identify and classify the two main research lines of the Artificial Intelligence that are, weak AI and strong AI, as defined by the Neuroscience academic John Searle. This distinction comes essentially from a question: is a machine able to match or overcome human thinking or won’t it be never equivalent?
Weak Artificial Intelligence
The weak AI acts if it has a brain. The aim of the weak AI isn’t, in fact, realizing machines with human intelligence, but systems able to successfully act in relation to some complex human functions. On the base of this theory, there’s the idea of building a machine able to simulate the human behavior, never equaling or overcoming it. The weak AI operates examining cases similar to the problem to be solved, comparing them, elaborating a series of solutions and choosing the most rational and appropriate. The weak AI, in fact, doesn’t want to define processes created as mental ones. It verifies empirically the hypothesis and manages efficiently the task assigned. In the weak AI it doesn’t matter about human cognitive processes, but exclusively solve specific problems.
Strong Artificial Intelligence
According to the same definition by Searle, for the strong AI, the machine shouldn’t be considered as an instrument. In fact, if it is opportunely programmed, it can be compared to the human mind, with an indistinguishable cognitive capacity. In order to understand if an artificial system is also intelligent, had been created the famous Touring Test. If a machine overcomes this test can be defined as intelligent. The rules of the test are really simple: a man, closed in a room, makes questions, through a remote keyboard to a calculator. If the man isn’t able to distinguish if the answers coming from the other room are from a human mind or a machine, we’ve an intelligent calculator. But, as Searle underlined: can a machine be considered intelligent – even passing Touring test – simply for its ability of doing an action, on a series of data received, without understanding the real significance of words and answers given by the user?
As evidently appears, both the positions leave open some relevant issues, especially from philosophical and moral point of views, paying several questions with difficult answers.
Now, we can only analyze that in the last years the theories on Artificial Intelligence tried to overcome this dichotomy, coming to the conclusion that a machine can be considered intelligent only when able to reproduce how a human brain works on a cellular level.
In the next articles we’ll try to understand together if and how reach this ambitious aim.