August 1998, Vol. 15, No. 8 Who's in the driver's seat: research or applications? AAAI '98: a report from Madison The 15th National Conference on Artificial Intelligence (AAAI '98), held last month in Madison, Wis., opened with a classic "good news/bad news" scenario. According to NEC Research Institute's David Waltz, current president of the AAAI, the AI field is thriving, but in his view most of the successes are at the surface level. "Very little work is being done on solving the deepest problems of intelligence," he said. With his opening remarks, Waltz neatly summarized the basic dichotomy that has long characterized the AI field: Is research driving applications, or are applications driving the research? Acknowledging the numerous and significant industrial applications currently deployed (see p. 8 of print edition for a review of the Innovative Applications of AI conference), Waltz nonetheless argued that the field needs to see a greater emphasis on science. "We need to be problem-driven, not solution-driven," he said. The fundamental question for AI researchers, Waltz said, should not be "What is intelligence?" but rather, "What is intelligence for?" He chastised the field for ignoring most of the work being done by neuroscientists, and urged that researchers should develop AI systems capable of modeling human reasoning. The AI field being what it is, of course, most applications-oriented developers tended to disagree with Waltz's concerns relative to real-world systems. Following is an overview of some of the other memorable presentations at this year's AAAI show. Prescribing an AI cure It can be shown that from robotics to heuristics, the application of artificial intelligence can reap tremendous rewards. And as the use of AI begins to pile up more and more success stories, other fields such as the entertainment industry (i.e., music and game playing) are beginning to stand up and take notice. The field of medicine, however, is no stranger to AI, and while some may ask, "Why AI?" Russ Altman, a professor at Stanford University, offered proof as to why AI works in the medical profession. According to Altman, many characteristics of the medical field make it a good candidate for using AI techniques: • The medical field is very knowledge-intensive. • In the medical profession, it is very easy to reach "information overload." • The medical field is a "diverse field with diverse opportunities." • The medical field requires many "stereotypical reasoning tasks." Currently, one of the hottest areas of AI application is the Human Genome Project. Begun in 1990, the U.S. Human Genome Project is a 15-year effort coordinated by the U.S. Department of Energy and the National Institutes of Health to both identify all the estimated 60,000 to 80,000 genes in human DNA, and determine the sequences of the 3 billion chemical bases that make up human DNA. In doing so, doctors will be able to better track the spread of infection and "diagnosis of some diseases may only require the determination of a certain sequence." It is predicted that medical practices will be radically altered when new clinical technologies based on DNA diagnostics are combined with information emerging from genome maps. Emphasis will shift from treatment of the sick to a prevention-based approach. Researchers will be able to identify individuals predisposed to particular diseases and will devise therapeutic regimens based on new classes of drugs, avoidance of environmental conditions that may trigger disease, and possible replacement of defective genes through gene therapy. Additional hot spots for AI in medicine include: • telemedicine (delivering care remotely to rural or battlefield areas); • intensive care units; • clinical trials (enrolling patients and following their progress); • information integration and retrieval (point-of-care access and patient education). How smart is your PC? Are computers as smart as we give them credit for? Are they machine or human or a mixture of their creators? These were some of the questions raised by Stanford University's Clifford Nass, co-author of a book entitled The Media Equation: How People Treat Computers, Televisions, and New Media as Real People and Places. According to Nass, computers can fall prey to things like flattery and social rules. By performing human studies on computers, Nass found that computers can be flattered, dominant or submissive, and generalists or specialists. In studies, Nass has found that people are polite to computers, respond to praise from them and view them as teammates. They "like" computers with personalities similar to their own, find masculine-sounding computers extroverted, driven and intelligent, while they judge feminine-sounding computers knowledgeable about love and relationships. Nass has also performed studies using television and what he found was that labels make a difference. By putting the label "Specialist" on one TV and showing nothing but one specific topic, viewers rated that particular television superior and more "intelligent" to another television simply labeled "Generalist." Additional studies that Nass and his team at Stanford have conducted include the following: • In a social setting, when a person is asked a question about themselves from another person, they will respond in a polite manner, as opposed to being less polite (and less consistent) if they were asked the same questions in a written questionnaire. The team at Stanford performed two experiments to determine if the same occurs during human-computer interaction. In the first experiment, subjects performed a task with a "text-based" computer, and were then interviewed about the performance of that computer in one of three ways: 1) the same computer; 2) a paper-and-pencil questionnaire; 3) a different (but identical) text-based computer. In the majority of the cases, those subjects evaluated by the same computer answered more politely and more consistently than they did with the paper-and-pencil questionnaire or different computers. • It has been determined that people with similar personalities tend to be drawn toward one another. However, what about people being drawn to computers with similar personalities? Nass and his staff at Stanford believe that computers can not only have personalities but that people can be drawn to different types of computer "personalities." Using the "If you were stranded on a desert island, which items would you take?" scenario, Nass found that computers can have both dominant and submissive personalities (which can be easily created using a "minimal set of cues"), and that humans — depending on their personality — would be drawn to the computer with similar "personality" traits. AI is stranger than (science) fiction "The irony is that throughout history, humans have designed machines to behave like slaves. The more we build our machines like slaves, the more they are going to choose their codes of conduct," said Joan Slonczewski during a panel discussion of science fiction authors at the AAAI '98 show. Science fiction offers readers a chance to escape this world and enter an entirely different one. This world can open doors we never considered and excite us in ways we never thought possible. At the AAAI '98 show, three science fiction writers opened their doors of speculation of the future of AI and how our world will be different based on this ever-changing field. Sarah Zettel is the author of Reclamation and a new book, Playing God, which will hit bookstores in November. At the conference, Zettel read a chapter from her book, Fool's War, in which humanity depends largely on AI to maintain communities on several planets. In this particular chapter, one of the main characters deals with what can happen when technology literally rules the world. Joan Slonczewski, both a professor of molecular biology at Kenyon College and a science fiction author, read from her book, Daughter of Elysium. In this book, all human needs are served by robots, and the conflict of the story begins when one of the robots (whose "purpose" in life is to carry the train worn by its human master) realizes it has another purpose in life and tries to escape to another planet. The third participant in the panel discussion — James Hogan — has been a writer for more than 20 years, having begun his career working for Digital Equipment Corp. He's written several books, including Inherit the Stars, Mind Matters and Bug Park. All three writers talked about the future of AI and where they predict this field — as well as science fiction — will be hundreds of years from now. "Science fiction will continue to wander along like it always has, but I think the media will change the nature of storytelling," said Zettel, adding that this method of writing gives people a chance to predict the future, to paint a picture of a world to come. In addition to predicting the future, Hogan believes that writers and readers have always had a fascination with "what-if questions" and "creating other worlds for these what-if questions." His predictions for the future (and more specifically, the future of AI) is one of a virtual world: "AI could converge with virtual reality where you're more of a participant than a spectator," Hogan said. While Slonczewski agrees that science fiction enables people to attempt to predict the future, it does pose a dilemma: How do you create machines that will do what they should do but will not do as her character in Daughter of Elysium, i.e., develop a mind of its own? "I suppose that is an age-old question," she said. "I think the moral dilemma is, do we want to do that?" Vendor round-up One promising sign at this year's show was a slight but discernible upgrading in the status of the exhibition. In recent years the trade show aspect of the AAAI shows has been almost invisible, but this year more than just a small handful of companies chose to display their wares. It was particularly good to see some new faces at the show, such as Numan Intelligence (a vendor of neural network and genetic algorithm-based tools), Register Machine Learning Technologies (developer of the AIM Learning Technology, a genetic programming system), and InQuizit Technologies (developer of intelligent software that interprets the meaning and concepts of plain English in context). Long-time AAAI exhibitors Brightware, Franz and Harlequin again had booths at the show, as did a number of research labs and book publishers. Along with the increased scope of the exhibition came an upswing in overall attendance at AAAI '98: 1,300 attendees, up from the 1,000 who were at the 1997 show in Providence. Given that next year's show will be held at the most popular vacation site in the world — Orlando, Florida — it's not unreasonable to expect this upsurge in attendance to continue. Web Site © Copyright 1997, 1998 by Lionheart Publishing, Inc. All rights reserved. Lionheart Publishing, Inc. 2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA Phone: 770-431-0867 | Fax: 770-432-6969 E-mail: lpi@lionhrtpub.com Web: www.lionheartpub.com Web Design by Premier Web Designs E-mail: lionwebmaster@preweb.com