In the early days of computer science, some optimists believed that robots would soon have the prodigious capability and versatility of the Star Wars robot C3PO. Since then we have learned that in order to have “intelligent” behavior on the part of a computer (or robot), we need to provide that computer with a model of the world that is essentially as detailed as that possessed by humans, including not only facts (“Sally’s phone number is 555-1234”), but principles and relationships (“If you drop something, it usually falls downward”). Knowledge We have made much progress on this problem of “knowledge representation.” representation We have devised abstractions that can be used to help build programs that do certain kinds of reasoning.
One example of such an abstraction is the directed graph, in which nodes represent entities (“the species cat” or “Fluffy”) and arrows (called arcs) from one node to another represent relationships (“Fluffy is a cat,” “cats are animals,” “Fluffy owns Fluffy’s milk saucer”); .
Another useful abstraction is formal logic, which allows us to manipulate facts by applying rules of inference, such as “If X is a cat and Y is the mother of X, then Y is a cat.” Nevertheless, progress on modeling, or abstracting, the real world or significant pieces thereof remains a fundamental challenge of computer science, one that is not likely to be solved completely in the near future.