Probably the best talk I saw at AAAI this summer was by Edward Feigenbaum. It was great because he didn't use powerpoint slides (which was quite refreshing...), but most of all because he's an interesting guy who knows a lot about AI.
I'm not yet ready with my summary of his talk, but I did come across the following set of remarks by him that covers a little bit of what he talked about at AAAI. It's about becoming an expert, and the nature of expertise... http://www-cs.stanford.edu/News/2004_commencement_remarks.php ======================================================================= Commencement Remarks June 13, 2004 Edward Feigenbaum Kumagai Professor of Computer Science Emeritus To our graduates, from the faculty and staff of the CSD, heartfelt congratulations. We are proud that you are our graduates. There probably is no better college and graduate education than the one you received here. You are at the cutting edge. You might even be an expert at some of the things you have learned. I'll revisit expertise in a few minutes, after telling some stories. To parents, sisters and brothers, wives, children, significant others, and all the relatives and friends who have come here to celebrate with you, thank you. You made this possible. And I don't mean just the money you spent for this expensive schooling. Your love and friendship is more important to the lives of our students and graduates than all that they have learned here. Whether it takes them a few months, a few years, or a few decades to realize this, they will surely come to understand. And my thanks to our Department Chairman, Professor Garcia-Molina, for inviting me to give this commencement talk. I haven't done this in the last quarter century. Though wiser now, I have lost whatever specialized expertise I had in giving commencement talks. The good news for you, sitting in the hot sun, is that I get to do it for only another nine minutes, by order of the Chairman. The scientific research I've done for almost 50 years has been in Artificial Intelligence or AI. In 1963, I edited a book whose title, Computers and Thought, sums up the AI idea nicely, as does the phrase Thinking Machines. . Psychologists whose research impinges upon this field call their work Cognitive Science. For me, nothing could have been more fun and more rewarding than AI. It is an immense challenge, and one that I personally regard as the "manifest destiny" of computer science. The working strategy of scientists facing very big problems is called divide-and-conquer. Break the problem into manageable parts and study just a part of the big problem. The part that I chose to study was expertise, the thinking of experts in various fields of professional specialization, for example, internal medicine; chemical analysis; detection of submarines from sonar signals; or detection of oil and gas in wells. My group wrote software that modeled the thinking of these and many other kinds of experts in their domains of specialization. We called such software Expert Systems. Some in Cognitive Science did experiments with human experts to understand what made them experts, what made them different from the novices in their field. Were they using their reasoning skills better, or did they just know more, or perhaps both? How much of thinking is knowing a great deal, versus thinking one's way through new situations/decisions/problems? What does a master learn in the ten years or so that it takes to become a master of a field? The answer, for both people and Expert Systems (or indeed any AI software that behaves at near-human quality level) is surprising. A rated chess master can recognize within a second or less about fifty thousand distinct and important chess positions; a novice about fifty or maybe a hundred. Yet both analyze about the same number of paths in the game tree. The master has learned to see the important things about chess. The novice has to think them through, and the thinking doesn't get him or her to expert level within the time frame of a move in the game. In Chinese and Japanese, a Kanji character represents a concept, an idea. More complex thoughts are built up from these Kanji. To graduate from high school, a Japanese youth must know 1800 Kanji. The average Japanese women's magazine, covering a wide range of subjects, uses about 2400 Kanji. The average undergraduate leaves the university knowing about 3200 Kanji. But professors in literature, law, and linguistics, among those professions that are deeply grounded in language--i.e. language experts-- know about fifty thousand Kanji. Let me try to make the point in a more direct way. I hope no one in this audience will get ill today (especially not ill from what I am saying). But if one of you were to become ill, we could quickly have you attended by one of our CS faculty, who have won innumerable awards and medals, and are truly world-class thinkers. Or we could rush you across campus to the Stanford Hospital to be seen by a medical doctor, who may not have won any awards for his or her thinking but does know fifty thousand things about medicine, disease, and the body. Which would you choose? A few minutes ago, I called the empirical results about expertise "surprising." Why "surprising?" Because they seem to contradict guidance we have been given by our mentors and by the professionals in education, from our high school days through college. I heard this advice most recently from a CS professor at a major research university. I asked him: if he were I, what would he say to the Stanford CS graduates? Here is my paraphrase of his reply: "Tell them that it's not the specifics of what they have learned that will be important to them in the future, but the general methods of analysis and reasoning they have learned." This reminded me of the advice given to me by a high school guidance counselor 55 years ago, when Latin was still being taught. She advised me to take Latin as my high school language because even though it was a dead language, taking Latin would help to train my mind. I didn't believe it then, and I don't believe it now, especially with my experience in modeling expert behavior. If you want to become an expert at something, you have to be prepared to learn thousands of things about the something. Your Stanford education has given you a great start. If you aspire eventually to become a world-class expert in that something, then you'll probably have to learn more-or-less fifty thousand things about that something. It will take several years of study, learning, and hard work. So, what's a young graduate to do with this insight? If you're a graduate with a new Ph.D., you've done much or most of it already. Your first two or three years of research and teaching will do the rest for you. And this is also what post-doctoral fellowships are all about--to complete the job of making you a world-class expert. Masters Degree graduates, take heart! You've done what you should have done, and are well launched. Now pick the domain of specialization that is the most fun, the most motivating, for you and go after it with hard work. Get absorbed in your work and learn, learn, learn. If your first job doesn't match this pattern, then change jobs or change domains. Bachelor Degree graduates, if you are heading for a Masters program in the fall, you've already done the right thing. If instead you are heading out of university life toward a career, use that opportunity and your time wisely, scanning widely to try to find out what area of specialization will inspire you into your 30s and 40s. In what would you like to become a world-class expert? Then go for it. You probably should return to some university for a Masters Degree in that area whether that area is CS, another technical field, Business, Public Administration, or anything. You have a great education. Now you need more knowledge. So always learn, learn, learn. There are, as I have said, literally tens of thousands of things to know about your area. Finally, for all of us, students, graduates, and faculty alike: in a field like ours, the fifty thousand things we know or aspire to know, keep changing, and the decay rate is so fast that we have to run hard in place just to maintain our level of expertise. In our field, the best of us can be a has-been within a span of five to ten years. Knowing fifty thousand relevant and important things about your area is hard work, and every day is Learning Day. I wish all of you good luck, and a great life! -- Ted Pedersen http://www.d.umn.edu/~tpederse ------------------------ Yahoo! Groups Sponsor --------------------~--> Make a clean sweep of pop-up ads. Yahoo! Companion Toolbar. Now with Pop-Up Blocker. Get it for free! http://us.click.yahoo.com/L5YrjA/eSIIAA/yQLSAA/x3XolB/TM --------------------------------------------------------------------~-> Yahoo! Groups Links <*> To visit your group on the web, go to: http://groups.yahoo.com/group/nlpatumd/ <*> To unsubscribe from this group, send an email to: [EMAIL PROTECTED] <*> Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
