ON THE THRESHOLDS OF KNOWLEDGE

We articulate the three major findings and hypotheses of AI to date: (1) The Knowledge Principle: If a program is to perform a complex task well, it must know a great deal about the world in which it operates. In the absence of knowledge, all you have left is search and reasoning, and that isn't enough. (2) The Breadth Hypothesis: To behave intelligently in unexpected situations, an agent must be capable of falling back on increasingly general knowledge and analogizing to specific but superficially far-flung knowledge. (This is an extension of the preceding principle.) (3) AI as Empirical Inquiry: Premature mathematization, or focusing on toy problems, washes out details from reality that later turn out to be significant. Thus, we must test our ideas experimentally, falsifiably, on large problems. We present evidence for these propositions, contrast them with other strategic approaches to AI, point out their scope and limitations, and discuss the future directions they mandate for the main enterprise of AI research.