Conference Report: The Ninth Conference on Artificial Intelligence for Applications (CAIA '93) Orlando, Florida, 2/28-3/5, 1993 Peter G. Selfridge, AT&T Bell Laboratories (Note: this report will appear in IEEE Expert Magazine, June, 1993. Currently, copyright for this report is held by AT&T Bell Laboratories.) The Ninth Conference on Artificial Intelligence for Applications (CAIA '93) was held at The Disney Yacht Club Resort, Disney World, Orlando, Florida, February 28 through March 5, 1993. The setting was terrific: not only is the Yacht Club an excellent hotel and conference facility, but the weather cooperated and Disney World location gave attendees a plethora of activies after (and sometimes during!) the conference. This year's call-for-papers attracted 215 submission (a big jump over last year's 145), and 80 papers were accepted: 61 (or 28 percent of those submitted) as full papers and 19 (or 9 percent) as poster session papers. This year's conference had strong international representation. Accepted papers and posters come from 15 countries: nine from Canada, six from Germany, three from Japan (one coauthored with a US collaborator), three from France, two each from Scotland and Italy, and one each from England, India, Portugal, China, Australia, Austria, Singapore, and Mexico. The rest (47) came from the United States. The conference started with two days of Tutorials and Workshops. There were eight tutorials, ranging in topic from AI and Business to Qualitative Reasoning to Intelligent User Interfaces. 102 people took these tutorials and they were well received. The Workshop on Validation and Verification of Intelligent Systems featured presentations by speakers including Robert Plant, Alun Preece and Daniel O'Leary. These presentations focused on specific issues in V&V, including life cycle impact on V&V, methodologies for system development including V&V, the use of metaknowledge in V&V, systematic determination of knowledge base anomolies and the statistical analysis of expert systems. A round table discussion of different applications followed the presentations. The V&V of a number of applications, including the "Pilot's Associate" and some of its derivatives, were discussed. The second workshop was a meeting of the IEEE P1252 Standards Group, whose charter is to propose a Frame Based Knowledge Representation standard. Topics discussed included the semantics of various aspects of the proposed standard (such as system defined frames, slot attachment, and enumeration functions), validation of knowledge bases, detecting inheritance conflicts, and what aspects of a knowledge base are modifiable after the initial definition. This is an ongoing activity. The technical program started with an invited plenary address by Wendy Lehnert of the Computer Science Department, University of Massachusetts, Amherst, Mass. The title of the presentation was "Portability and Scalability for Information Extraction Systems." She described the evolution of Natural Language research from basic research and demonstration systems to problem-oriented research with an emphasis on evaluation and portability. This change in emphasis has been accomplished in part by priorities of funding agencies and the availability of common problem sets with agreed upon input and output. This allows different groups of researchers to work on the same problems and compare results. Dr. Lehnert concluded her talk by listing five new directions of natural language research: (1) applications must stimulate basic research, (2) domain portability through training, (3) automated knowledge acquisition, (4) put a human in the loop, and (5) hybrid systems. In response to a audience question, Dr. Lehnert described the "weakest link" in this kind of work as that of discourse analysis, or the analysis of multiple sentences. At the Wednesday banquet (surprisingly good food and very well attended) there were two special events. The first was the acceptance of the Emanual R. Piore Award, an IEEE award for long-standing applications-oriented research, to Dr. Makoto Nagao, Kyoto University, Kyoto, Japan. This award comes with a handsome plaque and a generous honorarium, and Dr. Nagao chose the Conference on AI for Applications as the IEEE conference for accepting this award. Dr. Nagao discussed his research into global approaches to Natural Language and Image analysis on Thursday. Second, Oliver Selfridge, of GTE Research Laboratories, Waltham, Massachusetts, gave the banquet speech entitled "AI and the Future of Software". His controversial message is that computer science and artificial intelligence should take a new approach to software. Instead of trying to do everything at the level of specifications, and then transform these specifications into code, the community ought to acknowledge that specifications will always be ambiguous, incomplete, and, most important, will change. This implies that software should be written with change in mind, and that the types of change software undergoes should be extensively and empirically studied. If we do that, Dr. Selfridge argued, we can begin to design software to change and can begin to build evaluation into the programs themselves. Then we can begin to understand and build truly adaptive, intelligent software systems. Two other invited talks were part of CAIA '93. Patrick Winston, of the MIT AI Laboratory, gave a talk on his experiences in business. One message of his talk was that in a commercial product, the amount of actual AI technology tends to shrink far beyond expectations, and in some cases, vanishes. This mirrors other people's experiences in the amount of work necessary to turn a demonstration system into something closer to a product - that amount of work is much more than it took to build the prototype! Dr. Frank Mayadas, of the Alfred T. Sloan Foundation, gave an assessment of computer technology. In his view, the extraordinary pace of improvements in processing power, memory capacity, and the like, are beginning to slow. The biggest challenge by far, he claimed, is now to explore and develop new application areas for our technology. This, of course, was an appropriate message for the Conference on AI for Applications! The bulk of the conference was, of course, paper sessions and panels. The papers covered a large number of topic areas, from parallel processing to case-based reasoning, constraint satisfaction, and genetic algorithms. Two papers received Best Paper awards: Amir Hekmatpour, of IBM and also a student at UCSD, and Charles Elkan, a professor at UCSD, for their paper "Categorization-Based Diagnostic Problem Solving in the VLSI Design Domain", and Leonard A. Hermens, a student at Washington State University, and Jeffrey C. Schlimmer, a professor at WSU, for their paper "A Machine-Learning Apprentice for the Completion of Repetitive Forms". Panel presentations were very popular, and included a panel on statistical approaches to natural language, evaluation AI applications, AI on Wall Street, AI for Manufacturing, Organizational Memory, Applications of Case-Based Reasoning, and Massifely Parallel AI. The success of CAIA-93 was the result of many people's efforts. Jan Aikins, of Trinzic Corporation, provided terrific leadership and planning for the conference. Dave Waltz, as Program Chair, did a great job, especially during the final program committee meeting. His colleauge at Thinking Machines, Craig Stanfill, did a fantastic job standing in for Dave during an illness. Peter Selfridge, AT&T Bell Laboratories and Don McKay, of Paramax, worked hard to put together a successful tutorial and workshop program. Curt Hall of Intelligent Software Strategies and Doug Dankel of the University of Florida did excellent jobs at Publicity and Local Arrangements, respectively. People at the IEEE office, especially Nancy Wise, Janet Harward, Phyllis Walker, and John Mee, worked hard at many logistic details crucial to the success of CAIA-93. Dari Whitehouse of Thinking Machines also handled many administrative details. The program committee members reviewed all the papers and made many difficult decisions. Finally, the success of a conference like this ultimately rests on those people submitting papers and agreeing to be outside reviewers, invited speakers, and panel members. Thank you all! This brings us to next year's conference, which will be held at the Marriot Riverwalk in San Antonio, Texas. Dan O'Leary, of the Business School at the University of Southern California will be the Conference Chair, and Peter Selfridge of AT&T Bell Laboratories, will be the Program Chair. Dan and I are contemplating a number of changes in the Conference on AI for Applications. We want to ensure that this Conference remains attractive to a wide and diverse audience, from government to acedemia to industry. To do this, we are formulating an integrative theme for the conference that will emphasize that while AI is but one component of a complete application, it can be the most important technical component and the true "market differentiator". We hope the theme will both widen the scope of the conference and help us in the AI and applications community sharpen our focus on the exact role of AI in applications, and the role of AI research in applications endeavors. Dan and I are also committed to a number of structural changes in the conference to make it more efficient. While these changes are still being decided, it is probable that the conference will be shortened from 5 days to 4, and we are exploring various options including integrating tutorials into the technical program, having evening sessions, including demonstration tracks, and trying to facilitate more direct and active participation among the audience and speakers. Of course, a successful conference of this kind must also include outstanding plenary addresses and panels and, last but not least, attractive social events. If you have any ideas or suggestions for events, panels, workshops, tutorials, or are interested in submitting a paper, the due date is August 31, 1993. For more information, contact CAIA '94, IEEE Computer Society, 1730 Massachusetts Avenue NW, Washington, DC 20036, or call 202-371-1013. Alternatively, feel free to contact me, Peter Selfridge, at AT&T Bell Laboratories, Room 2B- 425, 600 Mountain Avenue, Murray Hill, NJ, 07974, phone 908-582-6801, email pgs@research.att.com. .