What is Cyc

wiseGEEK Writing Contest

Cyc is an attempt to convert a huge number of common sense propositions into formal logic and use them as the basis for artificial intelligent software. Founded by Douglas Lenat in 1984, who had previously created the AI program Eurisko, Cyc is the largest and best funded, and almost the oldest artificial general intelligence project in the world. By 2002 it had already received over $60 Million of funding and was not yet near completion.

Cyc is frequently discussed in the popular scientific press. It is the only prominent ongoing project exemplifying what is sometimes called GOFAI, or 'Good Old Fashioned Artificial Intelligence". What this means is that basically, Cyc takes the huge data-set of common sense propositions in its database and generates further tentative propositions via both inductive and deductive inference.

Obviously, in the latter case, all of its conclusions will be valid so long as its assumptions are valid, e.g. if its programmers are making sound assumptions and inputting them properly. Deductive inference can also be used to identify hidden inconsistencies between assumptions. For instance, if the Cyc database contained the propositions that animals are dumb, the proposition that humans are animals, the proposition that scientists are people, the proposition that scientists develop scientific theories, and the proposition that a person has to be smart to develop scientific theories, Cyc could deductively conclude that there are no scientists, highlighting the flaw in the assumption that animals are dumb, the failure to include an exclusion for at least some humans.

Inductive reasoning can lead to errors even in the absence of false claims so long as the data set used for training is non-representative of the larger world in some respects. For instance, Cyc contains information on many people. At one point Cyc proposed the hypothesis that all people are famous. This was a reasonable inference from the fact that all of the people it knew about were famous, but was actually a sign that Cyc needed information about a wider range of people.

For several years, a limited version of Cyc has been available on-line for public use. This version is called opencyc. People can use it for suggested applications, such as database integration, but more importantly, they can use it for applications that they invent. By working with opencyc they share their insights regarding the potential functionality of the system. Each user also produces and confirms new propositional inferences which are eventually integrated back into the main Cyc database, allowing Cyc to grow more rapidly than it could if its inferences were only being evaluated by its core development team.