ARTIFICIAL INTELLIGENCE

Artificial Intelligence



Artificial Intelligence

Introduction

Artificial intelligence (AI) was first defined by the American computer scientist McCarthy, who invented the term in 1956, as the “engineering and science of creating intelligent machines.” This definition, in its essence, has held, despite considerable shifts in technological paradigms, from the earlier emphasis on creation of intelligent computer programs to the current stress on convergence technologies. However, in the absence of an absolute definition of intelligence, only degrees of intelligence can be defined, with human intelligence being the benchmark to which other intelligences are compared. In addition there is also a lack of consensus on the kind of computational procedures that can be termed intelligent.

While computers can carry out some tasks, they cannot carry out all, and they lack the crucial ability to reason. Computer programs may have tremendous amounts of speed and memory but their abilities are circumscribed by the intellectual mechanisms that have been built into the programs. In fact, the ability to substitute large amounts of computing in lieu of understanding is what gives computers their seeming “intelligence.” Thus, for example, the chess player program Deep Blue substitutes millions of computations of possible moves in the place of reason and intuition.

Discussion

The key barrier to the creation of AI remains the failure to duplicate the nebulous quality of human intelligence that has been defined as the computational part of the ability to achieve goals in the world. With the inability of programs to replicate the essential features of human nature, such as common sense or intuition—attempts to create AI usually fail under the heavy load of rules that had to be written to deal with every problem. A few experts believe that human level intelligence can be achieved by amassing data in computers, but the general consensus is that without a fundamental transformation, it cannot be predicted when human level intelligence will be achieved.

Branches of AI

There are many extant branches of AI including logical AI—where a program corresponds what it experiences about the world generally with the facts of the particular situation on which it must act, with goals constituted by sentences of some mathematical logical language; search programs that scrutinize a large number of possibilities, such as moves in a game of chess; pattern recognition programs that compare what is perceived with a databank of stored images. Generally, the more complex the pattern, for example, a natural-language text or a chess position, the more complex the program; representative AI that denotes facts using languages of mathematical logic, inference machines, where facts are derived from other known facts, commonsense AI, which, in an indications of the difficulties in creating AI, is the least developed in spite of enormous research, except in certain areas such as non-monotonic reasoning and theories of action; AI programs that can learn from experience and are based on connectionism and neural nets that specialize or on the learning of laws expressed in logic.

One example of this is Mit97, which is a comprehensive undergraduate text on machine ...