ABSTRACT

There is little doubt that an AI will need to be able to reason logically. An inability to discover, for example, that a system’s conclusions have reached inconsistency is more likely to be debilitating than the discovery of an inconsistency itself. For a long time there has also been widespread recognition that practical AI systems shall have to cope with uncertainty — that is, they shall have to deal with incomplete evidence leading to beliefs that fall short of knowledge, with fallible conclusions and the need to recover from error, called non-monotonic reasoning. Nevertheless, the AI community has been slow to recognize that any serious, general-purpose AI will need to be able to reason probabilistically, what we call here Bayesian reasoning.