engleski

Why Not Fuzzy Logic?

Fuzzy logic is an approach to AI which focuses on the mechanization of natural language. It has long been proposed by Zadeh, its originator, as another paradigm for AI and the correct way to achieve "human level machine intelligence". To present day, this approach hasn't prevailed, but in the light of some recent tendencies in AI development it can gain traction. The "black-box property" of the currently predominant method — deep learning — has recently sparked a movement called "explainable artificial intelligence", a quest for AI that can explain its decisions in a way understandable and acceptable to humans. As it has been recognized, a natural way to provide explanations to users is to use natural language, embedded in the fuzzy logic paradigm. However, to model natural language fuzzy logic uses the notion of "partial truth", which has brought some philosophical concerns. The very core tenets of fuzzy logic have often been described as counterintuitive. In this text, we provide philosophical support for fuzzy logic by providing possible answers to the two most common critiques raised about it, as well as by offering independent philosophical motivation for endorsing it.

eXplainable Artificial Intelligence (XAI) ; Natural language ; Fuzzy logic ; Philosophy of vagueness ; Higher-order vagueness ; Contradictions

nije evidentirano