Director of Research Institute, RWCP
|
| Junichi Shimada
Director of Research Institute, RWCP |
Today, I'd like to discuss how the RWC (Real World Computing) Project was originally formed, what its current status is, and how it will be developed.
The RWC Project was initiated as a 10-year scheme in 1992 by the Ministry of International Trade and Industry. The overall purpose of the project is to create new trends in the area of information technology and promote them worldwide.
The RWC Project has two areas for technology development. One is real world intelligence technology based on information integration, and the other is parallel and distributed computing technology based on seamless computing system technology.
The real world intelligence technology was, in some sense, the starting point of the RWC Project. Our living environment contains a lot of information with various but ambiguous meanings. The RWC Project was designed to develop a set of technologies capable of processing such information as is. Before the RWC Project, only logic-oriented programs were available. So, we wanted a new kind of technology to deal with the real world in a "rough" way.
But there was another background. Since Japanese typewriters were not sufficiently practical for general use, we had a long history of handwriting. So, there was strong demand for devices capable of reading handwritten letters and characters. Such machines were one of the main development targets of information processing society in Japan. We therefore invested huge resources in the area of pattern recognition, and as a result Japan was considered to be one of the leaders in this field. And we gradually realized that we were reaching the point of developing new technologies to deal with real world information.
In addition, we thought that conventional computers were not capable of processing various and ambiguous information as is, so we decided to develop a new type of computer. This marked the beginning of our research on parallel and distributed computing systems.
During the first five years, our activities spanned a relatively broad range in order to identify areas to focus on. After the first five years had passed, we discussed which core research field we should invest our resources in, and decided on seamless computing system technology or seamless system technology, for short.
"Seamless" is a very common word and can be found in any dictionary. I don't think it's necessary to explain the meaning. I believe, however, the concept "seamless system" was clearly defined for the first time by the RWC Project. It was like discovering a tiny spring in the project that's now flowing and growing into a strong current. Our objective in this project is to turn that small flow into a large river and a global ocean current.
A "seamless system" by our definition is a computer system that connects different types of independently-used computers such as personal computers, workstations, and supercomputers via a fast information network and allows them to function as a single parallel computer. Since they are independently operable, "seams" are inevitably generated when connecting the systems. Those "seams" are actually hardware, software, or data. The purpose of the project is to eliminate such "seams" and make the group of various computers seem like a single parallel processing system.
Why did we choose "seamless system" as the core of our project?
First, we intend to cause a qualitative change in society by enabling even small organizations to use a large computing power, because at present supercomputers are restricted to large organizations. History suggests that ordinary laboratories and smaller offices will need ever more computing power.
Small organizations usually cannot afford a supercomputer, but a parallel computer which can be used as an enhanceable supercomputer is affordable. The idea is to initially build a small system and gradually develop it into a larger one, but this causes a problem. When you need to enhance your system, new microchips will probably have been developed and you won't be able to obtain older types of chips, so you have to replace your parallel computer entirely. Very few organizations will be able to bear the high financial cost, and that's why you need a parallel computer that can use any types of chips.
Second, we intend to help maintain industrial structure diversity. If parallel computers that can use only a certain type of chip are produced, that chip will gradually become dominant in the market and reduce competition, which is the backbone of the free economy. We believe that a parallel computer which can accept different chips could prevent any type of chip from dominating the market and thus help maintain the diversity of industry.
As a milestone in the development of seamless system technology, we have created a system called "PC Cluster." One of the characteristics of our PC Cluster is its high-speed communication capability, and it also offers a multi-user environment. The PC Cluster is already in practical use throughout the world; the Los Alamos National Laboratory in the United States was the first to employ the technology, but it's widely used in Japan and will soon spread all over the world through many routes.
I've been discussing seamless computing technology--the core technology of parallel and distributed computing system technology. We have also made remarkable achievements in the other RWC technology area, real world intelligence.
One of those achievements is what we call "VideoSpotter," a system for learning and retrieving video data. It stores video data such as TV programs on a magnetic disc as a database with a learning function. It can accept as a query a motion image which has been separately prepared and retrieves related motion images in a real-time manner. Suppose you have a database of all games that the Giants, a Japanese professional baseball team, played in the previous year and you input as a query a live TV program of the game that the Giants is playing now. You can get all scenes in which Hideki Matsui, a team slugger, hit a homerun during the previous season if he scores one in the present match. When Matsui starts running to the first base a moment later after the homerun, you can also view all his running scenes last season after hitting a homerun. A Japanese company is now working to introduce this VideoSpotter to the market, and you'll soon be able to get it as package software.
This can be used not only as a single package software but also as part of a DVD (digital video disc) system. Once integrated, VideoSpotter will take full advantage of the random access function of the DVD system, so VideoSpotter should be a key player in the DVD market that is estimated to be worth hundreds of billions of yen.
Also, we've been working on a text data retrieval system. It's called "Galaxy" and is now in the final stage of development. Unlike conventional search programs that need a key word in an exact form, Galaxy can find various related information with a key word in a vague form. We decided to use the software for retrieving information from a huge amount of patent data and all associated agents are now starting to use it.
So far, I've emphasized the application of technology, for the reason that information technology has a peculiar characteristic that materials or devices do not have. That is, it is somewhat similar to art. An absolute evaluation is quite difficult, so the key is to assess the extent to which it is actually used and how the user evaluates it.
It is also useless if you can't use technology until the final goal of development has been achieved. Information technology is perishable. If you leave it as it is, IT will not only become obsolete but also rotten. Technology development and its application should go hand in hand. Although any kind of technology is perishable, this is especially true for information technology, so I think "being used" is very important.
Incidentally, the Japanese government provides funds for the development of fundamental technology in most of the national projects including MITI. The private sector then takes over the basic technology, makes modifications to it, and puts it into practical use. This is a typical path of technology development. It is, however, not applicable for information technology. Information technology must be usable even during the development process, and that's why we firmly believe that something has value only when it's actually used.
Toward the completion of the RWC Project, we will use an additional policy to make the results "visible." Of course, "visible" in this case doesn't mean "being able to be seen physically or physiologically", but that the results should be understandable even for those outside the project.
One small tip to make the results "visible" is to give them names. Conversely, you can't call it a "result" if it has no name. Naming makes it easier for people engaged in research and development to focus on a new technology. And christening makes it easier for those outside the project to recognize it.
What will count more is how the results can be utilized. Hopefully, they should be immediately put into practical use or converted into actual products.
We should create an environment where new technology can be self-developed once it enters the public domain. In the case of software, for example, one way to achieve this is to make it available for free, so that all users get the chance to contribute to the development of the software.
If any research subject has not reached a practical level by the end of the project, one possibility is for the members of the group responsible to continue the research after returning to their respective organizations. However, I personally would like the Japanese government to take over a large technology program like this seamless system technology as a national project if it is not complete at the end of its term. I believe that "power comes from continuation."