Search-Related Techniques in AI

Abstract

Almost all artificial intelligence (AI) programs can be said to be doing some form of problem-solving whether it be interpreting a visual scene, parsing a sentence or planning a sequence of robot actions. Search is one of the central issues in problem-solving systems. It becomes so whenever the system, through lack of knowledge, is faced with a choice from a number of alternatives, where each choice leads to the need to make further choices, and so on until the problem is solved. Playing chess is a classic example of this situation. Other examples include the attempt to diagnose a malfunction in some complex piece of machinery or determining how best to cut material to make an item of clothing with the minimum of waste. Thus interpreting a visual scene and parsing a sentence can be regarded as searches for a plausible interpretation of possibly ambiguous visual or aural data, and making a plan can be regarded as a search in a space of plans to find one that is internally coherent and achieves the given goals.