The annual conference of the Association for the Advancement of Artificial Intelligence (AAAI) is the premier U.S. gathering for artificial intelligence (AI) theoreticians and practitioners. On the commercial side, AAAI is the only event with a comprehensive exhibition that includes most significant U.S. vendors of AI products and services. In 1986 some 5100 people attended AAAI- a very good showing considering that the 1987 International Joint Conference on Artificial Intelligence (IJCAI) drew about the same number of people even with its substantial international support. The commercial exhibits at AAAI-86 (110 exhibitors; 80,000 square feet) gave us opportunity to take a snapshot of an industry in transition. What I saw was a dramatic increase in the commercialization of AI technology and a decrease in the mystique, smoke, and hype. A preliminary tour of the AAAI-86 exhibits indicated that participants could expect substantial changes from the situation at IJCAI-85.

In 1940, a 20-year-old science fiction fan from Brooklyn found that he was growing tired of stories that endlessly repeated the myths of Frankenstein and Faust: Robots were created and destroyed their creator; robots were created and destroyed their creator; robots were created and destroyed their creator-ad nauseum. So he began writing robot stories of his own. "[They were] robot stories of a new variety," he recalls. "Never, never was one of my robots to turn stupidly on his creator for no purpose but to demonstrate, for one more weary time, the crime and punishment of Faust. My robots were machines designed by engineers, not pseudo-men created by blasphemers. My robots reacted along the rational lines that existed in their'brains' from the moment of construction. " In particular, he imagined that each robot's artificial brain would be imprinted with three engineering safeguards, three Laws of Robotics: 1. A robot may not injure a human being or, through inaction, allow a human being to come to harm. 2. A robot must obey the orders given it by human beings except where such orders would conflict with the first law. The young writer's name, of course, was Isaac Asimov (1964), and the robot stories he began writing that year have become classics of science fiction, the standards by which others are judged. Indeed, because of Asimov one almost never reads about robots turning mindlessly on their masters anymore. But the legends of Frankenstein and Faust are subtle ones, and as the world knows too well, engineering rationality is not always the same thing as wisdom. M Mitchell Waldrop is a reporter for Science Magazine, 1333 H Street N.W., Washington D C. 2COO5. Reprinted by permission of the publisher.

Stochastic simulation is a method of computing probabilities by recording the fraction of time that events occur in a random series of scenarios generated from some causal model. This paper presents an efficient, concurrent method of conducting the simulation which guarantees that all generated scenarios will be consistent with the observed data. It is shown that the simulation can be performed by purely local computations, involving products of parameters given with the initial specification of the model. Thus, the method proposed renders stochastic simulation a powerful technique of coherent inferencing, especially suited for tasks involving complex, nondecomposable models where “ballpark” estimates of probabilities will suffice.

Only the section is currently online. Wiley-Interscience

This report contains the reading list for the Qualifying Examination in Artificial Intelligence. Areas covered include search, representation, reasoning, planning and problem solving, learning, expert systems, vision, robotics, natural language, perspectives and AI programming. An extensive bibliography is also provided.

Complex Systems, 1 (5), 995-1019.

"The ultimate goal of work in cognitive architecture is to provide the foundation for a system capable of general intelligent behavior. That is, the goal is to provide the underlying structure that would enable a system to perform the full range of cognitive tasks, employ the full range of problem solving methods and representations appropriate for the tasks, and learn about all aspects of the tasks and its performance on them. In this article we present SOAR, an implemented proposal for such an architecture. We describe its organizational principles, the system as currently implemented, and demonstrations of its capabilities." Artificial Intelligence, 33(1):1-64.

A computer vision system has been implemented that can recognize three-dimensional objects from unknown viewpoints in single gray-scale images. Unlike most other approaches, the recognition is accomplished without any attempt to reconstruct depth information bottom-up from the visual input. Instead, three other mechanisms are used that can bridge the gap between the two-dimensional image and knowledge of three-dimensional objects. First, a process of perceptual organization is used to form groupings and structures in the image that are likely to be invariant over a wide range of viewpoints. Second, a probabilistic ranking method is used to reduce the size of the search space during model-based matching.

The long-term goal of our field is the creation and understanding of intelligence. Productive research in AI, both practical and theoretical, benefits from a notion of intelligence that is precise enough to allow the cumulative development of robust systems and general results. The concept of rational agency has long been considered a leading candidate to fulfill this role. This paper outlines a gradual evolution in the formal conception of rationality that brings it closer to our informal conception of intelligence and simultaneously reduces the gap between theory and practice. Some directions for future research are indicated.

Many AI tasks can be formulated as constraint-satisfaction problems (CSP), i.e., the assignment of values to variables subject to a set of constraints. While some CSPs are hard, those that are easy can often be mapped into sparse networks of constraints which, in the extreme case, are trees. This paper identifies classes of problems that lend themselves to easy solutions, and develops algorithms that solve these problems optimally. The paper then presents a method of generating heuristic advice to guide the order of value assignments based on both the sparseness found in the constraint network and the simplicity of tree-structured CSPs. The advice is generated by simplifying the pending subproblems into trees, counting the number of consistent solutions in each simplified subproblem, and comparing these counts to decide among the choices pending in the original problem.