Toward Fundamental Information Technology of the Next Generation

Hidehiko Tanaka Chairperson of the Preparatory Committee of Project for Fundamental Information Technology of the Next Generation, MITI(Professor of the University of Tokyo)

For the five years leading up to this year, research and development activities have been conducted under the RWC title. For this year, prior to starting the second half of the project, the research subjects of the past five years have been reconsidered, mainly by the Ministry of International Trade and Industry. Based on this reconsideration, I would like to discuss the new research subjects planned to be studied during the second half of this project.

To begin, I would like to summarize the current status of information processing, and its relationship with our plan for the second half of this project. First, I will review the roles of information processing technologies. Second, I will discuss the expectations regarding required technologies for the coming years. Third, I will discuss the currently planned project reformation (Fig. 1).

The term "information society" frequently appears in mass-media (Fig. 2). One feature which characterizes the current situation is the progress of globalization, at the core of which exists personal computers and the Internet. Specifically, the popularization of computers and the Internet and their influence have already gone beyond the critical point.

Therefore, now is the time to expect a forceful impact from the information society; where everybody is involved or forced to be involved in this trend.

In viewing information technologies from this perspective, societal activities normally include a combination of human beings, materials, and money - each factor is accompanied with information. Thus, the key is in how such information should be exploited.

However, we are now in a chaotic age, and it is quite difficult to present a clear view on the future. On the other hand, by thinking over the current status of computers and what is to be expected in the coming years, a future view can be clarified in one way or another.

The current status of information processing can be summarized as shown (Fig. 3).

First, to exploit the full benefit of computers, we need an in-depth knowledge of them; computers are only for professionals at present. Since we cannot trust computers fully, users must know as common sense, the need to make backups in case of computer break-downs.

Second, computers can only support simple things. For instance, PCs may be used for document creation, spreadsheet calculation, compilation and storage of information into files, and electronic mail - all of which use computers as a tool. Processing only addresses the support of syntactic processing, and current information processing only deals with symbol processing of symbolic information.

Taking these situations into consideration, let's go on to discuss future expectations for information processing (Fig. 4).

First, computers should be easier to use for "People at Large," basically anybody. The majority of information people encounter now comes through televisions or radios, which is unilateral communication. In the future, a great change will occur where the average person will not only be the recipient but also the sender of information.

In addition, information processing of the future should be transparent. This is similar to seamless computing, as discussed by Dr. Shimada. With current information processing, it is not so easy to integrate multiple computers. Computer integration should be implemented without having to consider the distances between computers and compatibility issues.

Second, secretarial support should be provided. Everyone should have an electronic personal secretary. This wide-range availability would require a high-level of support to enable the adoption of personal preferences. Of course, to achieve this, computers need to be easy to use and adaptive to individual users. For an example of the need for these electronic personal secretaries, please compare my transparencies and Dr. Shimada's. My transparencies are dark with no pictures; his transparencies are very beautiful with many pictures. In preparing his transparencies he may have been supported by his human secretary, while the support I received by computers still remains at a primitive level.

Third, the support of semantic processing is needed. Current processing only deals with syntax, as I said earlier, which is difficult to use unless the semantics of some given information is processed as well. For this, computers need to understand the user's intention and respond to it. Interaction at a deeper level, in addition to responding to user-entered character strings, will lead to easier and more comfortable interaction.

It may also be necessary to develop new technologies in areas such as ontology, and to exploit the capability of handling non-textual information such as images and sensory information.

Then, what is required to implement such processing? (Fig. 5)

When various elements and products are combined to form new systems, the situation becomes more complicated. Within this complicated framework, an essential consistency check would require processing which makes use of knowledge found in each field. It would be desirable to have the ability to perform these consistency checks easily.

The running of simulations will play a major role in the design methodology in the future. Simulations will be used on a large scale. To write simulation programs, it may be possible to constitute simulators easily by directly mapping objects. This capability naturally requires very high performance computers, which will enhance the design methodology's usefulness.

Next, I would like to discuss the integration of information processing and communication functions. Currently, communication functions are used sometimes basically for processing information. In the future, handling information with communication functions combined with information processing will be done more often.

I would also like to comment on globally available data. Although access to a great amount of data over the Internet will be made easier, the amount of information itself does not guarantee value or usefulness. Semantic access is essential. Support for semantic access is required, rather than at a syntactic access level.

Regarding pattern information, patterns such as voice, image, graphics, and information from sensors should be handled easily and efficiently. For easy and efficient handling; technologies for extracting symbols from those patterns, recognizing them, and converting them from one media to another will be fundamental-in addition to those technologies for storing and retrieving them. The development of these technologies will result in the integration of symbol and pattern processing, and will in turn lead to flexible information processing.

High in priority is to have systems that will be equipped with functions to make them adaptive to the environment. To equip systems with these functions it is necessary as a basic technique to be able to automatically acquire various information from surroundings and automatically evolve its own functions.

Taking these conditions into account, there are many different technologies, which may be classified as shown in the figure (Fig. 6).

Fig. 6

Users will use various applications supported by the fundamental technology as seen in the figure. The fundamental technology can be further divided into two kinds of technology, one being a very easy-to-use high performance computer as a platform, and the other being the high-level support technology placed on the computer. By using the fundamental technology, the application technologies are utilized in many ways in unlimited fields.

As I just said, the application technology is characterized by having many fields and various styles and being open-ended (Fig. 7).

High-level technology may include those shown in Fig. 8.

As shown, high-level technology may include semantic processing, technology for handling incomplete information, and technology for overall judgment by combining multimedia, technology for making systems evolve based on specific situations and environments, and technology for tagging a large amounts of data and automatically classify them.

The basic part of the technology (Fig. 9) requires very fast computers capable of handling an immense amount of calculations. This technology will serve as a basis for implementing high-level interfaces, handling image data, and performing large-scale simulations. The basic technology will also establish world-wide high-speed communication and the environment for many computers to seamlessly communicate with one another. Reliability is crucial for users to use them with a full sense of trust. Although it may be difficult to make individual components completely reliable, the key technological goal is to make systems that are trusted by the users. Trust naturally involves the security issue. New device technologies should include more minuscule silicon technology and optical devices.

With the technologies thus divided, it may be very efficient to have the major projects focus on basic technology, and high-level support technology included in fundamental technology (Fig. 10).

On the other hand, the key to application technology is new ideas. Therefore, application technology is a very dynamic field and must be subjected to the market mechanism; some new ideas are abandoned and some are newly created.

In this sense, fundamental technology may involve the cooperation of people with different interests. This is the core of the five-year project which will be prepared mainly by MITI this year or a little longer. This core research will start from the next fiscal year. While research and development themes are not completely fixed, with discussions still ongoing, the general summary of the research and development themes may follow as shown (Fig. 11).

As shown in Dr. Shimada's transparencies, the new project can be divided into two subjects: intelligent technology for the real world, and the parallel and distributed processing technology.

Real world intelligence will involve research and development activities with foci on such key words as; information integration, learning, and self-organization.

Specific themes considered therein are divided into five fields: multimodal system, autonomous learning system, self-organizing information base, theory and algorithm, and new reconfigurable devices (Fig. 12).

Our goal is to refine current information processing technologies by covering them with listed technologies. One component of this goal is information integration. Information integration is a mechanism for handling and judging ambiguous and incomplete information in an integrated manner in the real world. Another component of this goal is self-organization and learning, which refers to technology that enables systems to be adaptive to the real world and evolve themselves through interaction with the real world. Given this technology, humans do not have to learn complicated written manuals; computers will take care of them. The second theme is parallel and distributed processing technology, and our goal is to provide an easy-to-use and integrated environment as a basic platform.

Parallel and distributed processing technology includes those fields represented by seamless computing, and the application programs and their subsequent programming technologies (Fig. 13).

The detail of these will be discussed in the special session to be held in the afternoon on the 31st.

With these themes and goals taken into consideration, the project is reformed as shown (Fig. 14).

For the previous five years, the project focused on the research and development of one system with real world intelligence technology on top of parallel and distributed processing technology; in the coming years, the research and development of those two will be conducted concurrently. The current idea is to focus on realistic elemental technologies rather than to integrate them.

In conclusion (Fig. 15), the general objectives are to show the clear-cut image of technologies and to show the direction of the research and development in the coming years.

The ensuing five-year period may be too short for open-ended research and development, yet is suitable for research and development with narrowed objectives. Some say the dynamism in research should be more seriously considered. This is also very important. It is not always adamant that we must treasure initially established goals to the end. The ensuing five years will be spent creating a great stream with dynamism in selecting better directions based on surrounding situations.

The fundamental information technology to be developed in the coming years will be common property shared among humanity. I sincerely hope that we are able to establish such fundamental technology in cooperation with people around the world.