Playing With Cyc

One of the world's oldest AI projects needs your help


For more than 20 years, researchers at Cycorp, in Austin, Texas, have been developing an artificial-intelligence system to apply a modicum of common sense about the world around us. Last year the company made part of the system available in an open-source format, so I downloaded the half-gigabyte-plus compendium of facts, inference methods, documentation, and customized Web server and started poking around.

And you know what? Half a gigabyte isn’t nearly enough to emulate the brain.

Cyc contains masses of information on subjects as varied as cats and Anna Karenina , linked thematically in hierarchical ontologies, formal specifications of how the concepts in a given area relate to one another. For instance, it might specify that a cat’s feet are at the bottom, its head and tail at opposite ends, and so on. Cyc also packs a host of tools for reasoning about the subjects, some formal, others heuristic. OpenCyc is a stripped-down version.

OpenCyc’s knowledge is good on some things but not on others. For instance, it ”knows” about the different kinds of ­tactics terrorists use, that blowing things up and taking hostages are different kinds of events but that both are actions that terrorists perform. On the other hand, although it knows that William Shakespeare was a poet, a dramatist, a human being, and the title character of the comic film Shakespeare in Love , it can’t give you a list of his plays.

However, if you just wanted to add real-world intelligence to your software—say, for a computer game or spam-recognition software—a knowledge base like OpenCyc would be a good place to start.