Logic and Reasoning in Uncertain Conditions

At the core of Artificial Intelligence, two major pathways of knowledge extraction and representation have been the cornerstone for many decades: Deductive Learning, based on sets of "rules" from Predicate Calculus and Horn clauses that represent the domain experts' knowledge; and Inductive Learning, based on 'generalization by examples' by more or less 'black box' algorithms.
In this forth lecture, uncertainty is introduced as a core factor in designing robust AI algorithms with real-world applications. Beyond the inherent difficulties of the pure probabilistic theory of the Bayes rule, the Certainty Factors approach is introduced as such an example, coming back from the mid-70s and the MYCIN expert system (LISP). Some elements of Fuzzy Logic are discussed as comparison to more modern approaches, paving the way to the robust data-driven paradigms of Machine Learning of the last three decades.

Keywords: Machine Learning, Data Analytics, AI, Artificial Intelligence, lecture
Video: https://youtu.be/NswyEh4aA_Q