Koichi Hori (born 1956)
Doctor of engineering
Professor, Department of
Advanced Interdiscipinary
Studies, School of Engineering. The University of
Tokyo
Under the leadership of Dr. Ryuichi Oka, chief of the Multi-Modal Functions and Information-Base Functions TRC Laboratories, the labs are studying systems in which people can manipulate information in the same form as they exist in the real world.
In the June 1999 issue of "Advertising," a monthly magazine published by Dentsu, Professor Koichi Hori of the University of Tokyo and Dr. Oka discussed the current status and objectives of the RWC Project in plain words. The discussion should be helpful for understanding the research activities of RWCP, and so we obtained permission from Professor Hori and Dentsu to reprint the article here.
What benefits will RWCP's research bring? How will our lives be changed by such benefits? And what does "Prefer 3-D Lover to 2-D Lover" mean? Enjoy reading their candid discussion.
The Ministry of International Trade and Industry of Japan started the Real World Computing (RWC) Project, a 10-year plan, in 1992.
The purpose of real world computing is to "develop a revolutionary information-processing system that can handle real-world information." I went to the RWC Tsukuba Research Center to meet Dr. Ryuichi Oka, and discussed with him the present status and future direction of the project. Although the project was rather difficult to understand, the discussion was fascinating.
Hori: Thank you very much for the interesting demonstrations. I'm raring to ask a lot of technical questions (laugh). First of all, would you please give a brief history of RWC Project in plain language that general readers can understand.
Oka: MITI started this kind of large-scale project some 27 years ago. During the first ten years, we worked on pattern information. In the next decade, it was the Fifth Generation Computer Project. And now, we are working on the Real World Computing Project. I believe this is MITI's third largest project. Probably because of current trends, the project is focusing on how human-related information-or raw data in scientific meaning-can be transferred into other forms that are easier for people to deal with.
Hori: What surprised me today was the fact that such a large program could run on a notebook PC. I wonder how much technology will change the world.
Oka: Our project is geared to creating fundamental methods, and later we'll work on implementing them. No matter how complicated a research theme seems to be, the fundamental method can be efficiently implemented on a notebook PC if it has really good potential. It's actually one criterion for evaluating technology.
Hori: I see. So when notebook PCs become able to recognize human gestures and speech, computers will be used very differently from the current method with icons and the mouse.
Oka: In terms of technology, it's a long road from success in the laboratory to practical use by housewives and children at home. If a man-machine interface works well at the general household level, then it really is easy to use. I believe that interfaces which can easily accept gestures and speech will be used more extensively.
Hori: I heard one of the purposes of your project is to promote creativity. How is this related with the development of multimodal interfaces?
Oka: If the Internet continues to grow, we must be able to handle an enormous volume of information. We'll also have to manipulate a greater variety of data types such as motion images, speech, still images and so forth, in addition to text. The critical issue will be whether we can effectively use them or not. As the old saying goes, "More than enough is too much." Too much of something could be nothing. The effective use of information is one way to avoid drowning in the sea of information. So we need to identify what to do, and that's one of the problems we are tackling in technology development.
A good interface between people and a massive volume of information should have real-time capabilities to handle text, speech, motion image, and still image data, and allow that data to be freely accessed, even by ambiguous requests. Without such an interface, we can't use information no matter how much Internet traffic increases, in which case the Internet would lose its raison d'etre. We have to develop new technologies for the effective use of information on the Internet. It's important that predictions for the future are consistent with our technology development efforts.
Hori: Do you mean that effectively using a mass of information will need creativity?
Koichi Hori (born 1956)
Doctor of engineering
Professor, Department of
Advanced Interdiscipinary
Studies, School of Engineering. The University of
Tokyo
Oka: Exactly. Mr. Takashi Tachibana* wrote an article entitled "Chi no Bakuhatsu" (Explosion of Intelligence) in a recent issue of Bungei Shunju**. He said that "concentration of information" or "concentrated information" is necessary for people to effectively use an enormous amount of information. Let me interpret his theory in my own words. To effectively use information means to enter a reproduction cycle of information. The reproduction cycle of information can be driven by intellectual creativity. If we only possess and consume information, in other words if we don't use information for the reproduc tion purpose or if we don't use it as materials for creativity, then information will be dead or meaningless.
Hori: In your demonstration, I saw a screen displaying all the text data of newspapers for one year in a space. Such a display could give you new ideas or help you produce new information.
Oka: Yes. And that is one of the purposes of "visualization." An important role of computers is to process in advance what people cannot handle, and display it for them. Human creativity excels in areas that can be intuitively understood. But the number of such fields is somewhat limited.
Merely visualizing things isn't what we call "visualization." The primary purpose of "visualization" is to visualize invisible things by constructing a virtual world or a vast pool of information and creating a new visual world. That is the difference from the visualization of phenomena in the common scientific world.
Hori: I agree. By arranging all newspaper texts systematically in a space as you did in your demonstration, we can see what the text is missing.
Oka: But of course people should make the final judgment. Humans are capable of making correct decisions but cannot control or manipulate huge amounts of information. So, we leave the control of information to computers and computers provide the service to people. The point is to find how we can provide such an information space. That's the purpose of technology development.
Hori: And you're developing basic technology as an intermediary means.
Hori: Your demonstration today looked like it could soon be commercialized.
Oka: We have received offers of commercializing parts of our technology and preparations are underway. This kind of technology, however, may find new applications after entering practical use. Researchers cannot predict what will happen once a new technology is released, and experiences out there in the market may breathe life into it. I think that's the flow of technology development. What we offer is just a "seed." People outside laboratories certainly have more ideas about how to use such seeds in better ways.
Hori: Very interesting. What kind of flower do you expect will bloom from that seed?
Oka: In the demonstration, you saw a search of past videos from live videos. Let me explain. When you watch a baseball game on TV, for example, you are anticipating what will happen based on a mixture of your own memories and hopes. How would it be if a computer made such predictions based on a vast volume of past data? In other words, your computer could tell you that Hideki Matsui*** will swing his bat in such a way and hit a home run when he stands like that.
Suppose a computer momentarily indicates in advance what is very likely to happen. Of course, the reality may differ from the computer's prediction. Like weather forecasts, however, we enjoy such differences as a kind of intellectual entertainment. Or suppose when a person repeats the same comment, your computer immediately retrieves data from the database and tells you about a trouble that happened when the person said it in the past. What's more, the computer could pick up certain images and speech from ordinary TV and retrieve related data from an immense archive of past information. It would be interesting if the computer could provide in real time new information that's beyond our imagination. It's frightening in a sense.
Hori: For example, when a Congress member says something, you'll soon be able to find out what he has said about it in his public commitment. Yes, it must be frightening for them (laugh)!
Oka: Providing data from an archive in many real-time ways will have a huge impact.
Hori: Thinking without receiving any response is qualitatively different from thinking with a response in real time from the outside world. Even when you get a really good idea, you often cannot remember it a few minutes later (laugh).
Ryuichi Oka(born 1945) After completing a master's course of engineering at |
Video Spotter's monitor screen showing a real-time |
Hori: Linked at that moment and place?
Oka: Yes. We'll always be supported by the results output from the entire system. Just making fast calculations is not "real-time" support. The entire system and all functions are concentrated and displayed in each frame. I think such calculation methods will qualitatively change the relations between people and the environment.
Hori: So, it won't be confined to the use of computers but will cause many production methods of many things to change. For example, we would be able to get real-time information during live broadcasting--not in pre-edited programs--of baseball or sumo, and know how a certain bout was exactly the same as one fought by two other sumo wrestlers some 32 years ago.
Oka: Many archives, including electronic libraries, are now being built, and although much data are being stored electronically, it's difficult to reuse the data. It's very hard to manually label every piece of information and standardization of labeling is impossible. Labeling is a complex task requiring a large vocabulary, so the retrieval of approximate raw data could be one way of reusing data in its essential meaning. In that sense, the key is how we can use electronic libraries in the sphere of reproduction of information in the future development of electronic technologies. This is related with the use of approximate raw data.
Hori: To make data available for later use, we had to process the data before storage. Once the technology you've just explained has been created, raw data can be used for retrieval as it is .
Oka: Exactly. And it won't have to undergo labeling like artificial intelligence of the conventional definition does. Retrieval that requires labeling as a prerequisite is a relatively simple technology. Labeling every motion image, audio signal, and still image is theoretically possible, but how would such labeling function? So, we will pick up raw data that are approximate to what we really need, and we'll do that in real time. It's a data retrieval by approximation. That means "phase" plays a significant role in the world of information retrieval.
Hori: I see.
Oka: I think that the "retrieval business", which is quite different from the conventional "recognition business", will become a new trend.
Hori: I hear that very specific subjects like a photo of Kakuei Tanaka* yawning are often in great demand in the TV broadcasting industry. What they do now is to label each photo with descriptions such "yawning," "smoking" and so forth, but you're saying that such individual labeling will become unnecessary.
This chart shows how one of the four data sets - speech |
Oka:At present, it's impossible to distinguish subtle facial expressions. But motion images contain much more information than numerous still images because motion images embody information in a time sequence. Some experiments have proved that motion images can provide relatively high-resolution information, so motion images could be a possible approach for many things, including for improving precision.
Hori: The past directly becomes a database as data can be stored without labeling for later reuse. Since every moment that we pass through becomes the past, not only the business world but also our domestic lives would be greatly changed by this technology.
Oka: As is often said, the purpose is not to use a computer, but to do something by using a computer, so we should put all our energies into that doing "something." Focusing on the operation of computers is not a proper way because the computer is just a means of achieving an end. If everything were to happen in real time, our attention would be directed to the "something" and we'd be able to do what we want without being aware of the computer's presence. However, the computer is still there and helps us to achieve the purpose. The point is how to provide a technological guarantee for the purpose of computer use. If we only think about how to operate computers, in-depth thinking will be impossible.
Hori: In the previous issue of this magazine, Mr. Kengo Kuma, an architect, said that he was concerned about how buildings could be made unnoticeable. The purpose of computer researchers is also to make computers unnoticeable. Another thing I learned today is that very different computer games would be possible, so that children's creative power could be trained.
Oka: Some recent types of computer games require players to use their hands to push buttons and feet to step on panels. This trend is backed by the idea that whole-body exercise is much better than being confined indoors in terms of mental health. If we have to use our entire body in addition to touching the keyboard, it would be no different from sports. And then parents would no longer be against video games (laugh). If it's possible to talk with the computer by whole-body motions and gestures, then yes, I guess an entirely new type of video game is likely to be invented.
Hori: You take a small terminal with you to a park, and when you see a bird, you immediately use the terminal to do a search, which responds with what kind of bird it is. I guess that a variety of new world would be realized.
Oka: I think that whole-body motions and gestures are a natural development of the interface between man and game machines. The problem is whether or not it can be implemented on a normal PC. If it's expensive, it won't be feasible.
Hori: If it's possible, it would help expand children's creativity.
Oka: Such a new video game should not consist of a human player and a machine, and it shouldn't be autistic. The game machine should exist between a group of children and a child, or it should be a "community game" that involves some kind of human relations.
Hori: Not confining children within a virtual world, but forcing them to go out into the real world.
Oka: That's right. I hear some young people can only fall in love with 2-D-world partners (laugh). They love being in a 2-D world because it's interesting. If the same degree of interest can be provided through a channel in this real 3-D world, people will return and obtain real feelings. It shouldn't be done by force but by providing interest. And it is important to create an interface for that purpose. But since 3-D-world partners are not virtual, it may cost young people a lot of blood, sweat and tears (laugh).
Hori: Tell me your dreams for the future.
Oka: After seven years of battling against complications in this 10-year project, we have successfully developed some individual technologies. We want to integrate those individual technologies in many different combinations, implement them, and perform some kind of demonstrations. However, elements for integration are just starting to appear, so we'd like to provide a firm basis for the fundamental architecture of the algorithms underlying those technologies implemented on computers.
Otherwise, our research might be short-lived and vanish into the night. To sum up our research activities, the "integration of symbols and patterns" exceeds the reality right now. Our activities must have new mathematical contents in the truest sense.
Hori: The computer society of the future looks very interesting. Thank you very much for joining us today.
Self-organiztion of data by the Incremental Path Method. TIme-sequence data such as motion image, sppech, and text can be atuomatically transferred into a network databases. This chart shows how TV images will be transferred into a network database.
| After the Interview
| More people than you might expect are concerned that they
won't be able to survive in the cyber-world of tomorrow without computer
literacy and may be forced to use electronic machines against their will.
Some advertising suggests that those who haven't used the Internet are not
modern people. I'm offended and wondering why so many people can be
irresponsible to spread such ridiculous and threatening messages. This
interview with Dr. Ryuichi Oka reveals that computer experts don't agree
with such threatening messages, but are conducting research and
development to make computers invisible in the background.
We cannot clearly imagine now how the world will look if computers become invisible and work hard for people by stealth. Rice cookers and washing machines with built-in computers are perhaps precursors of the cyber-world. But what will happen when such stealthy cyber-computing expands into the sphere of human intellectual activity? Dr. Oka and I are among the many who are trying to build a cyber-community where people can enjoy creative activities. However, such advanced technology could be used by a dictator for other purposes. The primary purpose of engineering is the welfare of human beings. Generally, researchers optimistically work on technology development in the pursuit of human happiness. Dr. Okašs technology has great potential. If used correctly, it should contribute to the well-being of people. I encourage those working in the creative industry to try out many applications of the technology. |