Cyc is a long-term artificial intelligence project that intends to construct a complete ontology and knowledge base that encompasses the fundamental concepts and principles governing how the world functions.

Cyc employs implicit knowledge other AI systems may overlook to capture common sense understanding. It is in contrast to information one could access on the Internet or acquire using a search engine like Wikipedia. As a result, Cyc enables semantic reasoners to reason like humans and is less "brittle" in unfamiliar settings.

Overview

Doug Lenat started the groundbreaking but divisive Cyc research project in 1984 to record human consensus knowledge. The background information a person is thought to have to comprehend things like newspaper stories or encyclopaedia entries is called "human consensus knowledge." The strategy used is explicitly entering this knowledge into a sizable knowledge base. Since the project's inception, a group of knowledge workers has been actively involved in this procedure. This two-person century-long, multi-million dollar effort has lasted for ten years. Cyc's engineering objective is to make standard application programmes less fragile by allowing them to rely on prior information. Building a system with human-level common sense and understanding is the scientific aim of Cyc.

Objective

The goal of the Cyc project was to turn the millions of pieces of knowledge that make up human common sense into a code that machines could understand. Along the way, this meant: 

(1) making a representation language, CycL, that was good enough, 

(2) making an ontology that included all human concepts down to a good level of detail, 

(3) making a knowledge base on that ontology framework that included all human knowledge about those concepts down to a good level of detail, and 

(4) making an inference engine that was a lot faster than the ones used in conventional expert systems at the time to be able to use the Cyc language.

Outside of math, most of Cyc's knowledge is only true by default. Cyc is aware, for instance, that parents love their children by default, that when you are made happy, you smile, that taking your first step is a significant success, and that when someone you care about achieves a major accomplishment, it makes you happy. 

When asked whether the image labelled "Someone seeing his kid take her first step" depicts a smiling adult. 

Cyc can logically deduce that the answer is "yes" and "display its work" by delivering a step-by-step logical argument using these five pieces of knowledge from its knowledge base. These are expressed in the language CycL and have syntax comparable to Lisp.

Current trends

In 2008, the researchers mapped numerous Wikipedia pages to Cyc resources. Currently, Cyc is linked to Wikidata. Plans for the future may involve connecting Cyc to DBpedia and Freebase. The majority of Cyc's current effort consists of knowledge engineering, the manual representation of world facts, and the implementation of efficient inference methods based on this knowledge. Increasingly, however, work at Cycorp entails providing the Cyc system with the capacity to connect with end users in natural language and assisting with the ongoing knowledge-building process through machine learning and natural-language comprehension. Another significant initiative at Cycorp is building a set of Cyc-powered ontological engineering tools to lower the barrier of entry for users to contribute to, edit, explore, and query Cyc.

Furthermore, Cycorp has ambitions to use Cyc's natural-language processing to analyse the entire internet. Only it can use the Cyc system to judge ambiguity, metaphor, and ellipsis. Unfortunately, they are scarce if any systematic benchmark studies of Cyc's performance exist.