Nothing Is Absolute

This paper argues that approximate knowledge might be better for an AI system; because it is better in dealing with contradictory bases. This argument is based on the hypothesis that categorization is the basic means that we comprehend the world, and this is also the way we abstract specific instances into general rules. However, during abstracting process we might lack some information, which may cause our theories about the world to be incomplete. In this case, if our theories about the world are too certain, we would be unable to predict facts and relations with lower likelihood. And I will demonstrate that if we admit that our knowledge is approximate, despite the incompleteness we will be able to predict facts and relations that are of lower likelihood through Pattern Matching.