Semi-formal methods

The Cyc project (and its recent open-source spin-off, OpenCyc) stands as one of the most well known and ambitious projects within Artificial Intelligence. It has undergone over 20 years of research and development in its quest to build software with commonsense. Cyc, and projects like it, are inspired by classical symbolic logics but pragmatically embrace inconsistency and computability so as to cost-effectively construct broad (but comparatively shallow) knowledge bases. Cyc is designed to reason in a way that resembles human 'rational’ thought about typical human mental concepts (including their inherent inconsistencies) and uses representations and deductive mechanisms that are tuned for 'typical’ commonsense reasoning problems.

In Cyc (and OpenCyc), symbols are chosen to correspond to the concepts that human beings use in thought and discussion but are carefully engineered so as to fit within a logically consistent meta-physical ontology. Knowledge in Cyc is bundled into 'microtheories’ that allow locally consistent but globally inconsistent knowledge. 'Microtheories’ improve the modularity of the system and avoid catastrophic propagation of inconsistency across an entire knowledge base (as classical logics are prone to suffer), thereby allowing the system to productively reason about Santa Claus despite knowledge of his non-existence.

A range of other, more recent, projects resemble Cyc’s attempt to create large scale ontologies of human knowledge. Project Halo is one such umbrella effort to encourage the development of a "digital Aristotle". Three competing teams attempt to encode standard college-level Advanced Placement (AP) textbooks into an ontological knowledge base and the systems are tested with typical examination questions (that are translated into logic).

The Standard Upper Ontology (SUO) and other projects like it are attempting to build general ontologies of knowledge that admit detailed knowledge from any discipline and are intended to serve as a potential representational core of the Semantic Web. In fact, the long term ambitions of the Semantic Web are for universal access to machine readable data and it is, therefore, conceivable that, when the Semantic Web incorporates vast stores of knowledge and powerful deductive capabilities, it may demonstrate a practical form of intelligence.