This is the next blog in the continuing series of interviews with leading professionals.
In this blog series, we continue our talk with Professor Hideyuki Nakashima: President Future University – Hakodate; the internationally renowned computer scientist, inspirational visionary and top-ranked leader. If you haven’t looked at Dr. Nakashima’s background, go to the first blog in this series.I had the pleasure of meeting Dr. Nakashima at the invitation-only international large society summit where CIPS participated/presented. Hideyuki’s reputation for innovation and leadership is well known worldwide and this led to this blog series. Dr. Nakashima was the youngest professor to ever be offered the presidency of a university in the history of Japan, due to his significant and outstanding achievements. Dr. Nakashima is creating new branches of computer science and merging others that will impact the world into the future. It’s worth your time to follow his work. There is an element of “Star Trek” which I find compelling and in talking with Hideyuki, I can feel his passion…I wanted to share this with you too!
Thank you and Enjoy!
Stephen Ibaraki, FCIPS, I.S.P., MVP______________________________Multi-Agent Simulations, Evolutionary Algorithms; Cognitive Science; Kyocho Architecture, RoboCup; Artificial Intelligence
Stephen: Can you provide examples of multi-agent simulation? How would evolutionary algorithms for searching be better?
Hidey: I talked about the possibility of future car-navigation systems. Its effectiveness was verified on our multi-agent simulation test bed at Cyber Assist Research Center. We also verified the effectiveness of a new transportation system called "bus on call". We found that in large cities like Tokyo or New York, all the buses should be run on-call mode only. The power of simulation comes from the capability of changing parameters in large scale to an extent that cannot be covered by an actual physical test (Bus-on-call in whole New York cannot be put into test unless it is known to succeed). Evolutionary algorithm is effective to find the best parameter setting from a vast search space.
Stephen: Describe your work in knowledge representation and reasoning, and cognitive science.
Hidey: A human being has an impressive ability to adjust her way of thinking according to the situation in which she is located. It is achieved automatically and no one seems to notice this capability of its own. However once we try to construct a program to reason flexibly, this ability to take the situation into account (we call this "situatedness" of thought) is recognized as a big problem. We simply do not know how to do it. In early AI, the problem was called the "Frame Problem" (McCarthy and Hayes1969).
I recently came to realize that the Eastern view of the world could be the key to a new approach. I am still working on this idea with my fellow researchers in cognitive science and architectural design.
Stephen: Can you tell us more about this new approach and some of the ideas you are developing? In what ways and how does the Eastern view of the world make the difference-what are its elements?
Hidey: The answer must be somewhat philosophical. You should not talk about "effectiveness" and "elements" because the concept is Western.
I would need to write a book to answer your question. Any description within a small space like this interview would need too much simplification. However, let me just illustrate the direction of my idea. When we build a new engineering artifact such as a large building or a highway, the traditional way is to design it completely before you begin construction. Even computer software is built in this way. However, it is an illusion that we can design a complex system completely before we activate it. Reality is that we have to compose it, run it, observe it, make adjustments to the specification and modify or reconstruct it. This process must be continued while the system is in operation. There is no timing that a system is completed. The Eastern view of the world takes bugs' eye (fixed to the running system or process) view rather than birds' eye view (separated from the system). Eastern view tells us that we cannot have a complete specification of a complex system.
Stephen: Describe your work with MultiAgent Systems.
Hidey: There is a phrase in Japanese meaning three ordinary people make one wise man. Its English version is "Two heads are better than one" but I don't think they say the same thing. The English version may follow Amdahl's law of parallel computing (efficiency of computing with N machines are always less than N times of the power of a single machine). But the Japanese version says it can be more than N. We wanted a computing architecture that can allow "the whole is more than the sum of parts". We named the project "Kyocho architecture". "Kyocho" do not have a direct translation into English. It roughly means "cooperation" but it's more than that. My definition of Kyocho is conflict + negotiation + coordination.
We tried many things in the project. One of them is used as a platform of RoboCup simulation leagues. Another research, application of multiagent simulation to societal design, was inherited by Cyber Assist project which was launched later.
Stephen: Can you go into more detail about the RoboCup simulation leagues? Moreover, can you overview another example of things you tried with the Kyocho architecture?
Hidey: RoboCup soccer competition was proposed by my colleagues in Japan and is a world-wide event now. It consists of two categories, physical robot leagues and simulation leagues. Physical league had emphasis on smooth agile movements and simulation league had emphasis on tactics. Our Kyocho project provided the simulator. We also programmed a soccer team using Kyocho architecture.
RoboCup was later extended to include RoboCup rescue. This rescue competition also consists of physical and simulation leagues. Rescue itself is a very interesting problem in a sense that a complete specification of rescue task cannot be given beforehand.
Stephen: What is the current state of artificial intelligence? Where do you see it heading? What are the implications to business, and the consumer?
Hidey: Every subfield of IT research alternates in two phases: basic research phase where each component technology is developed, and integration phase where basic technologies are combined into practical applications. Both phases have their own research issues and they need each other. Integration research may find that some of the element technologies are missing and this leads to basic research. Basic research without applications in mind may lead to problems with little significance.
Both AI and IT in general are in the integration phase in this decade. Computers became so powerful that we can actually apply algorithms - which can "theoretically" solve problems but only applicable to toy problems because of resource limitations ten years ago - to practical problems. This is why ubiquitous computing is a hot issue these days and also why we initiated Cyber Assist project.________________________In the next blog, Hidey will talk about:
I also encourage you to share your thoughts here on these interviews or send me an e-mail at firstname.lastname@example.org.________________________