The view of a typical AI researcher on the Cyc project seems to be
``Cyc is generally viewed as a failed project. The basic idea of typing in a lot of knowledge is interesting but their knowledge representation technology seems poor.''
R. V. Guha, who was the co-leader of the project with Lenat for years and the co-author of the book Building large knowledge based systems [Lenat and Guha, 1990] left the team in 1994. He is quoted saying in an interview [Stipp, 1995]
``We were killing ourselves trying to create a pale shadow of what had been promised. Cyc may prove useful in commercial applications like data mining, but the goal of creating a system that would exhibit real common sense failed.''
Lenat, who recently founded the privately-owned company Cycorp to continue the development of the Cyc project, still seems optimistic. His view is [Lenat, 1996]
``The development of Cyc was a very long-term, high-risk gamble that has begun to pay off. Begun as a research project in 1984, Cyc is now a working technology with applications to many real-world business problems. Cyc's vast knowledge base enables it to perform well at tasks that are beyond the capabilities of other software technologies. The applications currently available, or in development include: natural-language processing, integration of heterogeneous databases, knowledge-enhanced retrieval of captioned information, distributed ai, www information retrieval...''
In this review I will try to discuss whether or not Cyc actually failed, and whether its basic approach has any hope of reaching human level intelligence. The disturbing thing about reviews such as this one, or in general papers about competing approaches to AI, is that the arguments never seem sound and conclusive. The reader should expect this as a natural outcome, though. The reason why there exists competing approaches in our field is that they cannot be dismissed by simple deductive reasoning. Each counter-example introduced for a methodology, can typically be met by a small patch. After any number of patches, another counter-example can be found. Such as the nature of things, how can one argue objectively for or against an approach? Below are some possible types of arguments: