NOVEMBER/DECEMBER 2008 (Vol. 23, No. 6) pp. 2-3
1541-1672/08/$31.00 © 2008 IEEE
Published by the IEEE Computer Society
Published by the IEEE Computer Society
Why Do We Need Intelligent Systems?
PDFs Require Adobe Acrobat
This is my last editorial before moving on to other things. The IEEE Computer Society has a policy, which I strongly applaud, of term limits for editors in chief, and I've reached the end of mine. Four years has been long enough to put my stamp on the magazine's direction, but not so long as to cast it permanently into concrete. I have great faith that the magazine will be in good hands and that the incoming editor in chief, Fei-Yue Wang, will help it grow in new and important ways.
Before I leave, however, I want to reflect on an issue important to the future of this magazine and the field it represents. It's an issue that all Computer Society editors in chief face: should their magazine continue to be published?
One obvious way to interpret this question is with respect to the overall issue of whether any magazine should continue to be published in print, especially one in a scientific field. This magazine, and many others published by the IEEE, must eventually face that issue, but it's not something that necessarily will affect the individual publications' existence. The IEEE, and many other publishers, will either find a way to keep functioning in print and the magazines will continue, or they won't and a new model will be found. Either way, that doesn't really get at the core of the question: why should our magazine, IEEE Intelligent Systems, continue to be published in whatever mode eventually wins?
When the IEEE asks this question, it's usually thinking about the magazine's profitability and the staff effort to keep a title alive when there are competing interests. A magazine such as this one, which has been around a long time, competes with new and emerging areas, and the resources are limited. I'm happy to say that we tend to do well on the various measures that the IEEE uses to evaluate these things (subscription rates, digital-library downloads, impact factors, and so on), so I haven't had to work hard to justify continued publishing. In that respect we're doing okay.
However, a different issue does impact the magazine and the whole field of AI: the tendency to focus on the US and Europe. It's increasingly obvious that a huge amount of the world's engineering expertise resides in emerging markets, especially in India and China. Those of us in the traditional research powers tend to dismiss this work as primarily the application of existing technology. That attitude has some basis in truth; these economies' first needs are often for the application of existing approaches. Also, we tend to notice the large number of students from these countries coming to our laboratories, but we forget the large number of researchers and scientists practicing outside our countries. As these countries' economies grow, so does their ability to fund research. Is it a coincidence that India and China have recently launched successful space missions? Furthermore, as the second and third generation of students and researchers return to reinforce the laboratory staffs and research ranks, capability grows.
This means that, as a magazine, we must look at how we reach out to these countries to be able to publish their best research. We also must look for subscriptions (or downloads) from these markets. Growing subscriptions in our traditional markets is difficult, so growing them in these emerging areas is crucial.
It was a fortuitous coincidence that at the time this need was becoming clear, the best candidate for my successor as editor in chief was Fei-Yue Wang. Fei-Yue is a well-known researcher who has a strong background in applied AI. A Fellow of the IEEE and the American Association for the Advancement of Science, and a Distinguished Scientist of the ACM, he's the director of the Laboratory of Complex Systems and Intelligence Science at the Chinese Academy of Sciences' Institute of Automation and is affiliated with the University of Arizona and several of China's top universities. As you can see from this issue, for which he was a guest editor, exciting things are happening in China, and he's deeply involved with them. His column in a Chinese magazine is widely read, and he has many exciting ideas for how we can better interact with scientists outside the US and UK. I'll let him tell you about those when he takes over this column, but for now, let me welcome him warmly both for his knowledge of our field and his insights on this important issue.
Science and Technology Concerns
Although the publishing considerations are important to the magazine, they aren't the concern I have when thinking about this publication's future. This magazine, originally named IEEE Expert, was created when a fairly serious split existed between those researchers focusing on the application of AI technologies and those focusing on the field's theoretical foundations: the famous "scruffies" versus "neats" debate. That was back in the mid-1980s; after 20 years, the field is less split in that dimension. To take one obvious, but important, example, consider the machine-learning community, where applications often provide the data sets for the theoretical work, and the theoretical work inspires the approaches used, for example, to scale the applications. So, given the healthier interaction we see nowadays between the approaches, at least within subfields, do we still need this magazine?
My answer, which won't surprise anyone who has been reading my editorials the past few years, is that this publication plays an important role in our field. The reason rests in the caveat in the previous paragraph's last sentence: "within subfields." Much has been written about the splintering of AI into many subdisciplines. An analogy I've heard often is that AI is like computer science itself; that is, nowadays CS consists of a number of subdisciplines—such as AI, for example. People making the analogy generally go on to argue "and isn't that just a sign of a maturing field?"
But that argument is flawed. The more I work across different areas of AI, the clearer I see that the lack of communication between these fields hinders the development of larger scale, "AI complete" applications. Different subfields are making different assumptions about representation, granularity of action, mathematical frameworks, architectural underpinnings, and many other things. So, when you try to integrate these subfields' products, the jigsaw pieces don't fit together.
That's the reason for this magazine's existence. As we build applications of AI in the real world, we more and more need to integrate pieces that fit together less and less well. More and more, the hard problems we're being asked to help tackle, whether they involve solving an integration issue in an enterprise or attacking one of the UN's millennium challenges, can't be solved by any one AI technique. I don't mean that there aren't problems that can be solved by some particular approach. Rather, I mean that many of our remaining challenges fall outside the province of any one AI approach.
In my mind, IEEE Intelligent Systems has the role of keeping AI practitioners, whether they're researchers or developers, informed about what's happening in other parts of the field. I hope you've noticed this goal in the issues we've published throughout my years as editor in chief. If you're a machine learning researcher, Intelligent Systems doesn't replace your subscription to Machine Learning; it makes you aware of what the rest of the field is doing. If you're doing planning (or knowledge representation, or agent-based modeling, and so on) we don't replace going to the International Conference on AI Planning Systems (or whatever you favorite conference is); we enhance your understanding of what's going on at those other conferences.
Goodbye and Thanks
So, do we need Intelligent Systems? The answer is a resounding, emphatic "Yes!"
That's why, for the past four years, it has been my privilege to head the team of volunteers and staff who make this publication possible. I particularly thank Crystal Shif, Hilda Carmen, and Dennis Taylor, all of whom I've had the pleasure of working with over the past four years. The IEEE has realized how good these people are at what they do, and first Crystal, then Hilda, and now Dennis are being pulled away to positions with greater authority and impact. The staff replacing them are also a talented bunch, and I've appreciated their efforts as well (especially those of Alkenia Winston, who has been working hard to learn our somewhat idiosyncratic manuscript management practices). I also want to thank our advisory board, our hardworking editorial board, and the many guest editors over the past few years. It has been great working with so many of my friends and colleagues in such an important pursuit.
Yours as ever,