A computer program capable of acting intelligently in the world must have a general representation of the world in terms of which its inputs are interpreted. Designing such a program requires commitments about what knowledge

Present day systems, intelligent or otherwise, are limited by the conceptualizations of the world given to them by their designers. In this paper, we propose a novel, first-principles approach to performing incremental reformulations for computational efficiency. First, we define a reformulation to be a shift in conceptualization: a change in the basic objects, functions, and relations assumed in a formulation. We then analyze the requirements for automating reformulation and show the need for justifying shifts in conceptualization. Inefficient formulations make irrelevant distinctions. A new class of meta-theoretical justifications for a reformulation called irrelevance explanations, is presented. A logical irrelevance explanation demonstrates that certain distinctions made in the formulation are not necessary for the computation of a given class of problems. A computational irrelevance explanation shows that some distinctions are not useful with respect to a given problem solver fo...

The performance of all problem-solving systems depends crucially on problem representation. The same problem may be easy or difficult to solve depending on the way we describe it. Researchers have designed a variety of learning algorithms that deduce important information from the description of the problem domain and use the deduced information to improve the representation. Examples of these representation improvements include generating abstraction hierarchies, replacing operators with macros, decomposing a problem into subproblems, and selecting primary effects of operators. There has, however, been little research on the common principles underlying the representation-improving algorithms and the notion of useful representation changes has remained at an informal level. We present preliminary results on a systematic approach to the design of algorithms for automatically improving representations. We identify the main desirable properties of such algorithms, present a framework for...

At the present stage of research in artificial intelligence, machines are still remote from achieving a level of intelligence comparable in complexity to human thought. As computer applications become more sophisticated, however, and thus more influential in human affairs, it becomes increasingly important to understand both the capabilities and limitations of machine Intelligence and its potential impact on society. To this end, the artificial intelligence field was examined in a systematic manner. The study was divided into two parts: (1) Delineation of areas of artificial intelli gence, and postulatio " of hypothetical prod ucts resulting from progress in the field, and (2) A judgmental portion, which involved applications and implications of the products to society. For the latter purpose, a Delphi study was conducted among experts in the artificial intelligence field to solicit their opinion concerning prototype and commercial dates for the products, and the possibility and desirability of their applications and implications. 1.

The Pigeonhole Principle (PHP) has been one of the most appealing methods of solving combinatorial optimization problems. Variations of the Pigeonhole Principle, sometimes called the "Hidden" Pigeonhole Principle (HPHP), are even more powerful and often produce the most elegant solutions to nontrivial problems. However, some Operations Research approaches, such as the Linear Programming Relaxation (LPR), are strong competitors to PHP and HPHP. They can also be applied to combinatorial optimization problems to derive upper bounds. It has been an open question whether PHP or LPR establish tighter upper bounds and how efficiently, when applied to the same problem. Challenged by this open question, we identify that the main reason for the lack of ability to compare the efficiency of PHP and LPR is the fact that different problem representations are required by the two methods. We introduce a problem representation change into an Integer Programming form which allows for an alternative way of solving combinatorial problems. We also introduce several combinatorial optimization problems, and show how to perform representation changes to convert the original problems into the Integer Programming form. Using the new problem model, we re-define the Pigeonhole Principle as a method of solving Integer Programming problems, determine the difference between PHP and HPHP, prove that PHP has the same bounding power as LPR, and demonstrate that HPHP and Integer cuts are actually similar representation changes of the problem domains.

The paper introduces a general approach for solving engineering problems by transforming them to problems in other engineering fields, through general discrete mathematical models called graph representations. The idea of the method is first to raise the problem to an abstract mathematical level of graph representations. At that abstract level, either the solution is found through the tools of graph theory, such as the graph duality principle, or the problem is transformed further to another engineering domain, where the problem is associated with a known solution. The paper demonstrates a number of applications of the approach, among them: deriving new concepts in engineering; establishing new engineering designs; analyzing complicated engineering systems, and a generic treatment of both analysis and design. The paper draws the correlation of the suggested approach to known AI topics: representation change and classification problem solving method.