We are victims of one common superstitionthe superstition that we understand the changes that are daily taking place in the world because we read about them and know what they are. The anthropological stories and the concept of memes were brought to my attention several years ago by Lynn Conway Much of the vision and some of the material was drawn from a paper that we worked on together but never published. The important distinction between process and product, was made crisp for me by John Seely Brown, who also has encouraged and made possible projects like Trillium, which I watched with interest, and like Colab, in which I participated. Joshua Lederberg kindled my interest in biological issues and a respect for knowledge processes and their partial automation that has not faded Dan Bobrow listened to my ramblings on several runs, agonized over my confusions, helped to get the kinks out of the arguments, and suggested the title for the article Sanjay Mittal and I have spent many hours speculating together on the issues in building community knowledge bases and knowledge servers and in understanding the principles of knowledge competitions Austin Henderson helped me to understand the Trillium story and to report it accurately. Austin and Sanjay hounded me to say, more precisely, what a knowledge medium is Agustin Araya and Mark Miller participated in a Colab session in which we tried to jointly lay out these ideas, and together asked me to make the prescriptions clearer Ed Feigenbaum persuaded me to be more precise in the discussion of the limits of today's expert systems technology Thanks to Agustin Araya, Dan Bobrow, John Seely Brown, Lynn Conway, Bob Engelmore, Ed Feigenbaum, Felix Frayman, Gregg Foster, Austin Henderson, Ken Kahn, Mark Miller, Sanjay Mittal, Julian Orr, Allen Sears, Lucy Suchman, and Paul Wallich for reading early drafts of this paper and for helping to clarify the ideas and improve the article's readability Stephen Cross triggered the writing of this article when he invited me to give the keynote address at the Aerospace Applications of Artificial Intelligence Conference in Dayton, Ohio, in September 1985.

We use our experience with the Dipmeter Advisor system for well-log interpretation as a case study to examine the development of commercial expert systems. We discuss the nature of these systems as we see them in the coming decade, characteristics of the evolution process, development methods, and skills required in the development team. We argue that the tools and ideas of rapid prototyping and successive refinement accelerate the development process. We note that different types of people are required at different stages of expert system development: Those who are primarily knowledgeable in the domain, but who can use the framework to expand the domain knowledge; and those who can actually design and build expert system tools and components We also note that traditional programming skills continue to be required in the development of commercial expert systems Finally, we discuss the problem of technology transfer and compare our experience with some of the traditional wisdom of expert system development. We have observed during this effort that the development of a commercial expert system imposes a substantially different set of constraints and requirements in terms of characteristics and methods of development than those seen in the research environment.

Inherent batch-to-batch variability, aging, and contamination are major factors contributing to variability in oilfield cement-slurry performance. Of particular concern are problems encountered when a slurry is formulated with one cement sample and used with a batch having different properties. Such variability imposes a heavy burden on performance testing and is often a major factor in operational failure. We describe methods that allow the identification, characterization, and prediction of the variability of oilfield cements. Our approach involves predicting cement compositions, particlesize distributions, and thickening-time curves from the diffuse reflectance infrared Fourier transform spectrum of neat cement powders.

Early this year fifty people took an experimental course at Xerox PARC on knowledge programming in Loops During the course, they extended and debugged small knowledge systems in a simulated economics domain called Truckin Everyone learned how to use the Loops environment, formulated the knowledge for their own program, and represented it in Loops At the end of the course a knowledge competition was run so that the strategies used in the different systems could be compared The punchline to this story is that almost everyone learned enough about Loops to complete a small knowledge system in only three days. Although one must exercise caution in extrapolating from small experiments, the results suggest that there is substantial power in integrating multiple programming paradigms. We extend our special thanks to the course participants from Applied Expert Systems, Daisy Systems, ESL, Fairchild AI Lab, Lawrence-Livermore Laboratories, Schlumberger-Doll Research Laboratory, SRI International, Stanford University, Teknowledge, and Xerox Corporation Their participation and feedback are vital to the ongoing experimental process for simplifying the techniques of knowledge programming We enjoyed and will long remember their spirited involvement. As in many situations in life, pat solutions and simple mathematical models just aren't good enough. To cope with messiness, AI researchers have found that large amounts of problem-specific knowledge are usually needed.

Preventive-maintenance schedules occurring in industry are often suboptimal with regard to maintenance coallocation, loss-of-production costs, and availability. We describe the implementation and deployment of a software decision support tool for the maintenance planning of gas turbines, with the goal of reducing the direct maintenance costs and the often costly production losses during maintenance down time. The optimization problem is formally defined, and we argue that the feasibility version is NPcomplete. We outline a heuristic algorithm that can quickly solve the problem for practical purposes and validate the approach on a real-world scenario based on an oil production facility. We also compare the performance of our algorithm with results from using integer programming and discuss the deployment of the application.

Moreover, the system was designed from the beginning to be maintained on an ongoing basis without the involvement of senior knowledge engineers. In the manufacture of paper, wood is first pulped to separate its fibers. One of the predominant pulp processes is done in a kraft pulp mill and consists of cooking wood chips at elevated temperature and pressure in the presence of certain chemicals (alkali and sulfide), washing the resultant brown pulp, bleaching to make the pulp white, and drying the pulp for shipment to a paper mill. Pitch, or wood resin, is the material in wood that is insoluble in water but soluble in organic solvents. It usually makes up 14 percent of the weight of wood after the bark is removed and is often a sticky material.

Enabling organizations to capture, share, and apply the collective experience and know-how of their people is seen as fundamental to competing in the knowledge economy. As a result, there has been a wave of enthusiasm and activity centered on knowledge management. To make progress in this area, issues of technology, process, people, and content must be addressed. In this article, we develop a road map for knowledge management. It begins with an assessment of the current state of the practice, using examples drawn from our experience at Schlumberger.

In this issue of AI Magazine, we continue our presentation of extended versions of papers presented at IAAI-12 (held in Toronto, Ontario, Canada) that were selected for their description of AI technologies that are in practical use. Our selections for this issue describe deployed applications. They explain the context, requirements, and constraints of the application, how the technology was adapted to satisfy those factors, and the impact that this innovation brought to the operation in terms of cost and performance. The articles also supply useful insights into use cases that we hope can also be translated to other work that the AI community is engaged in. In the first of these deployed application articles, eBird: A Human/Computer Learning Network to Improve Biodiversity Conservation and Research by Steve Kelling, Carl Lagoze, Weng-Keen Wong, Jun Yu, Theodoros Damoulas, Jeff Gerbracht, Daniel Fink, and Carla Gomes, the authors describe an intriguing application that successfully combines the best in human and artificial computing capabilities with an active feedback loop between people and machines.

AI activities are also being pursued at other Schlumberger locations, often jointly with SDR The locations related to logging and interpretation include: Schlumberger-Doll Research, Ridgefield, Connecticut (Contact: Peter Wu'l); Schlumberger Well Services, Austin, Texas (Contact: Scott Gut/my); Schlumberger Well Services, Houston, Texas (Contact: Scott Ma&s); Nippon Schlumberger, K K, Tokyo, Japan (Contact: Dennzs O'NezU); I&ude et Production Schlumbcraer. Other Schlumberger companies involied in Ai research include! Expert Systems Current work in expert, systems is concerned with developing techniques for building more robust and versatile log interpretation systems. One shortcoming of "first generation" expert systems, such as the Dipmeter Advisor, is their inability to reason about the task that they attempt to perform. Any description of the overall task is usually procedurally encoded and unavailable for examination.

Shell U.K. Exploration and Production (Aberdeen, U.K.) has implemented an advanced forecasting system for predicting oil field production. The expert system helped Shell achieve over $1.6 million in cost savings for its Brent Field site within 2 months of implementation. The National Research Council has awarded Nestor (Providence, R.I.) a grant to develop a neural network-based video sensor system, crossingguard Arvin Industries (Columbus, Ind.) is working with the U.S. Air Force to develop a neural network system that can determine the quality of noise in such vehicles as automobiles and aircraft. The neural network will help determine what exactly an annoying sound is and how it can be fixed. Using virtual reality hardware and software, Parke-Davis (Morris Plains, N.J.) has been able to improve the molecular modeling research techniques it uses to develop new pharmaceutical products.