Summary of the 3rd Meeting of the Fundamental Information Technology of

the Next Generation (Real World Computing Project) Promotion Committee

The 3rd meeting of the Fundamental Information Technology of the Next Generation (Real World Computing Project) Promotion Committee (chaired by Professor Hidehiko Tanaka, The University of Tokyo) was held at MITI on August 24, 1999.

This committee was established as a private advisory board for the Director-General of the Machinery and Information Industries Bureau at MITI to discuss and determine basic policies, implementations, and other important issues during the second phase (FY1998-FY2001).

In this meeting, the achievements during FY1998 were reported and the implementation plans for FY1999 were confirmed. MITI made a special briefing on the pending issue concerning the methodology for future development of the fundamental information technology of the next generation. The following is a summary of the comments and opinions of the committee members concerning MITIｹs statement. (The committee members who attended the meeting are listed at the end of this report.)

Opening Speech

Mr. Shin Yasunobu, director, Electronic Policy Division, Machinery and Information Industries Bureau (MITI), gave an opening speech at the beginning of the meeting.

[Summary]

I would like today's meeting to be conducted primarily as a free discussion.

The points that MITI is now reviewing in relation with not only RWC project but other technology development projects and policies are as follows:

• The advanced technology at the time the RWC Project started is now entering practical use. How should we deal with the situational differences?

• It was decided to reorganize the central government ministries and agencies in 2001, so the technology development policies must be clarified. In addition to defining the responsibilities of industry, academic, and government, MITI must express its views on these matters.

• MITI annually spends some 40 billion yen for the development of information-related technologies through various programs and systems. We have to assess on-going projects, consider how to implement them, and discuss what to do next after the projects.

I would like you to freely discuss the evaluation of the RWC Project and its future development. Your discussion will help MITI reconsider its policy and future direction as well as provide ministerial guidance for programs other than the RWC Project.

Achievements in FY1998

1. Introduction

During FY1998, we were actively involved in various research activities including multi-modal functions in the field of real world intelligence technology, and seamless computing technology in the field of parallel and distributed computing technology.

Research was carried out at Tsukuba Research Center and member laboratories of the Real World Computing Partnership, and the RWI Center of the Electrotechnical Laboratory.

The industrial sector has shown strong interest in the results of the project. For instance, the broadcasting industry has indicated their intention to market a video retrieval system.

The outstanding performance of the RWC PC Cluster system was highly evaluated at SC98 (High Performance Networking and Computing Conference 1998), a well-known international conference in the field of parallel computing. Some of its basic software programs have already been put to practical use by Japanese companies and US institutes.

Also, the PAPIA (PArallel Protein Information Analysis) system that uses the PC Cluster is available to all over the world through the RWCP web site. The number of accesses to the site is increasing daily.

2. Overview of Research Results

(1) Real World Intelligence Technology Field

We developed technology required for information integration and a learning type information processing system which can accept raw information as is from the real world, recognize or predict a specific environment or situation, and autonomously respond to it. Also, we compiled a database of real world information and a software library as the basis of research and development in this field, and identified some items that can be used as benchmarks and evaluation methods.

(2) Parallel and Distributed Computing Technology Field

We developed fundamental technology required for realizing the next-generation parallel and distributed environment to provide optimal parallel computing capability in response to increasingly diverse computing demands. Also, we established a methodology to improve the ability of developing the generation-after-next computing architecture.

3. Patent Applications, Presentations for Outside Occasions, and Awards

(3) Awards (in FY1998)

• JPSS `98 Best Paper Award (Joint Symposium on Parallel Processing) Parallel and Distributed System Software TRC Laboratory
• Recommended Lecture Award at the 4th Meeting of the Japan Society of Applied Physics Optical Interconnection Fujitsu Laboratory
• Best Presentation Award at the 1998 Japanese Society for Artificial Intelligence National Conference Information-Base Function TRC Laboratory and Multi-Modal Functions TRC Laboratory
• Best Presentation Award at the 1998 Japanese Society for Artificial Intelligence National Conference Jijo-2 Robot Lab and Learning and Information Integration Lab, Electrotechnical Laboratory
• Excellent Paper Award at Image Sensing Symposium Adaptive Vision Lab, Electrotechnical Laboratory
• IAPR MVA Organizing Committee Most Influential Paper of the Decade Award Electrotechnical Laboratory
• ICES98 (International Conference on Evolvable Systems 1998) Best Paper Award Evolvable Systems Lab, Electrotechnical Laboratory
• ICES98 Best Student Paper Award Evolvable Systems Lab, Electrotechnical Laboratory
• Excellent Researcher Commendation Award by the Director-General of the Agency of Industrial Science and Technology (MITI) Jijo-2 Robot Lab, Electrotechnical Laboratory

4. Research Results of Distributed Laboratories (omitted)

Research Implementation Plans for FY1999

1. Overview of research implementation plans

The research methodology should be explicitly defined so as to make this year a milestone of the second phase and to clarify overall the research results within this year, since this is the third year of the projectｹs second half.

2. Real World Intelligence technology field

(1) Overview of Real World Intelligence technology field

We develop the information integration technology and learning/self-organizing technology that are necessary for an information integration and learning-type information processing system which can accept raw information as it is found in the real world, recognize or predict a specific environment or situation, and autonomously respond to it.

Specifically, we constructed three kinds of system for confirming functions including the elemental functions of recognition/understanding, interaction, problem solving, control, using a parallel study method and attempted to establish technologies for information integration, learning/self-organization, and so forth in coordination with theoretical and fundamental research on algorithms.

In addition to developing real time adaptive processing devices which are useful for these technologies, we will upgrade the real world information base, tasks that become benchmarks, software libraries and so forth necessary for research and development.

(2) Implemented theme and contents

1) System for confirming functions

• Multi-modal function
  With the purpose of developing an agent-driven human interface for using information systems through a combination of speech and images, etc., during this fiscal year we develop a model of the positions and postures of hands and fingers and human behavior as well as a learning and recognition method based thereon.

• Autonomous learning function
  With the purpose of developing an agent system that can move around in a real environment autonomously and collect and learn information from the environment and its surroundings through sensing and interaction, etc., and which can act on its own, during this fiscal year we develop a method of separating sound sources from a sound background that contains more than one sound source, as a method of recognizing the environment in which the robot moves.

• Self-organizing information-base functions
  With the purpose of developing an agent system that can compile, summarize, retrieve and present various types of information in a self-organizing manner in the real world or on an information network, during this fiscal year we develop a user adaptation function in the self-organization and retrieval of large documents, retrieval of speech, motion images, and static images, and a method for 3-dimensional visualization.

2) Theoretical and algorithmic foundation

With the purpose of conducting research on the fundamental theories and common methods for information integration, learning/self-organization, optimization, and so forth, during this fiscal year we develop distributed and active learning methods, a method of evaluating knowledge, theoretical models, a reasoning algorithm and self-organization methods.

3) Real world adaptive device

With the purpose of developing the next-generation field programmable gate array (adaptive device) equipped with high-speed and adaptability in order to carry out real time processing, during this fiscal year we design and develop an LSI for a demonstration of adaptive devices.

3. Parallel and Distributed Computing technology field

(1) Overview of Parallel and Distributed Computing technology field

Besides establishing parallelizing software technology for effectively using the hardware performance of a multiprocessor system, we establish the fundamental technologies required for a parallel computing environment in which computing resources having various architectures can be used without needing to acknowledge the differences in architecture. In addition, we develop a state-of-the-art parallel application for demonstration.

(2) Implemented theme and contents

1) Seamless parallel and distributed computing technology

Using a representative structure of a distributed system as a platform, we dynamically re-construct a heterogeneous environment on which this distributed system exists. For research and development of diverse computational processing, during this fiscal year we construct a prototype system that provides a complete image for compilers, operating systems and optical interconnections.

2) Multiprocessor computing technology

Focusing on the software side of research and development of parallelizing elemental technologies such as parallelism extraction and application technologies, and scheduling technology which will be key technologies for developing next-generation high-performance multiprocessor systems and single chip multiprocessors (SMC), during this fiscal year, following on from last year, we reinforce the procedure-to-procedure data flow analysis technology and a parallelizing transformation technology geared toward the use of cache.

3) Parallel application technology

Based on the overall plan of a demonstrative approach in the development of parallel applications that efficiently operate on a parallel and distributed environment, focusing on research fields that could have a potential ripple effect in industrial fields other than information technology, during this fiscal year we demonstrate a compound analysis of protein structures, and investigate a compound-type function prediction technology in bio-infomatics.

4. Support for Research and Development

(1) Planning, coordination, and investigation for promoting research and development

We plan to make detailed implementation plans, construct a preferred implementation system, dispatch researchers and other personnel to investigate situations in other countries, and promote exchange programs by inviting researchers and technical personnel from domestic and overseas sources. Also, we will host and manage expert meetings and WGs in order to promote research and development.

(2) Improvement of the research and development environment

We will improve the computing, networking, and research support environments and maintain, operate, and manage them in order to facilitate research and development activities.

(3) Promotion of joint research activities at home and abroad

We will choose subjects from the real world intelligence technology and parallel and distributed computing technology fields considering whether or not they could be more efficiently conducted at university laboratories, if the environments of university laboratories are superior to others for the subjects concerned, and the degree of speciality of such subjects. Once we determine that a certain subject is suitable for a university laboratory, we will identify the most appropriate from 20 laboratories of 13 Japanese national universities, five laboratories of three Japanese private universities, and three laboratories of three overseas universities.

(4) Support for the U.S.-Japan Joint Optoelectronics Project

We plan to support the U.S.-Japan Joint Optoelectronics Project in their activities in the field of computing optoelectronics.

Development of Fundamental Information Technology of the Next Generation (Real World Computing)

Methodology for future development (pending issue)

1. Awareness of issues

The development of fundamental information technology of the next generation is scheduled to be complete in FY2001.

In FY2000, one year prior to the projectｹs final fiscal year, all laboratories will be required not only to make utmost efforts to finalize their research and development, but also to immediately take into further deliberation the following issues before the project year 2000 starts in order to make this project a real success and allow the extensive user of its results.

• Selective allocation of budget to those subjects that will produce outstanding achievements

• Policy to disseminate research results

• Final image of the project when it is complete

2. Selective allocation of budget

It is necessary to selectively allocate funds to those research subjects that will make remarkable research results so that their activities will be promoted more strongly. To this end, the criteria on which each research subject will be assessed and selected must be set. Our suggested selection criteria are as follows:

• Degree of achievement for research goals
• Technical novelty and originality
• Closeness to practical use
• Influence on society
• Degree of interest in Japan and abroad

3. Policy to disseminate research results

In addition to producing remarkable achievements, another objective of this project is to disseminate the research results. So, we have to establish a policy for that purpose and carry out specific steps in order of feasibility. The following are such feasible steps:

• Actively applying for patents and other intellectual property rights
• Making presentations at academic society meetings both in Japan and abroad
• Making full use of the mass media and the Internet

4. Final image

To date, all activities have been conducted according to the Master Plan and implementation policy. However, it is now necessary to project the extent to which each research effort will be achieved, taking into consideration their current status and present technology trends.

In other words, we should not be particular about element technologies but draw an entire picture of the project in which each elemental technology is systematically linked with others to produce significant results.

Summary of Discussion

The following is a summary of committee members' opinions and questions concerning the methodology for future research and development.

( Committee members, Representatives from MITI, RWCP, and RWI Center)

1. Dissemination of research results

What progress did you make specifically in the field of PC Cluster since last year? How about the dissemination of research results? In some cases, the United States has progressed quicker in terms of dissemination than Japan.

We have received an inquiry from Mitsubishi Electric Corporation for our PC cluster technology as an electric power control tool. We plan to exhibit it in SC99 (an international conference in the United States). According to our basic policy for disseminating this software, we provide it to some researchers and ask them to examine it for any problems and feed back their findings to us. We need a scheme that allows the technology to be used as free software.

If we acquire patents or other intellectual property right, it will be possible to transfer such rights from the government to the private sector after the enforcement of the "Special Measures Act for Revitalizing Industry", which is scheduled this October. This Act is, however, only applicable to rights obtained after the enforcement.

We can see some research results on RWCP's web site, but the web site exhibition seems to have a certain tendency.

Distributed laboratories of member companies have links with their supervising departments. It is possible that their research results will be disseminated or manufactured as a commercial product after the expiration of this project. On the other hand, RWCP Tsukuba Research Center does not have any means of commercialization, yet they are required to make their results accessible to the outside. Consequently, TRC's dissemination efforts are more public than those of distributed laboratories.

Could you spend 5 to 10 percent of the budget for dissemination activities?

The contract research fund from the government is basically for research activities and it is difficult to spend it on dissemination. However, the commercialization of the motion image retrieval system is funded by a different budget (the 3rd supplementary budget of fiscal 1998). Also, the commercialization of the text-base retrieval system is financed by a budget other than the research contract fund from the national government.

If it is to be disclosed as an open source, it can be presented to the rest of the world as a technology of Japanese origin. The past proves that technology will not be extensively used unless its source is made public.

Software programs developed by RWCP should be made available for free use in the same way as Linux, which is spreading all over the world. As the case of Linux indicates, software may gain wider popularity without organized promotion.

It is difficult to allow the free use of any software programs developed under research contract fund from the Japanese government. If Article 30 of the Special Measures Act for Revitalizing Industry is applicable to our case, it will probably be easier to make our software publicly available.

Since the National Property Act is applicable to computer software, users must pay for software programs owned by the government. Hence Article 30 of the Special Measures Act for Revitalizing Industry applies to our computer programs, intellectual property rights can be possessed by research developers and thus RWCP can choose to provide original programs with or without compensation. It is therefore necessary to create an effective mechanism for dissemination.

When an outcome funded by the national budget is to be commercialized, it should be considered as a return on investment. If it is guaranteed that business based on the results will grow, such results will be extensively used.

Subsidies must be returned in the end, but the results will belong to the developer. On the other hand, it is not necessary to return the contract fund, but the developer cannot freely use the results. When the Special Measures Act for Revitalizing Industry is applied and developers will gain the right to use what they have developed even in a contract-base system, the conventional subsidy system could lose its advantage.

2. Selective allocation of budget

Would it be realistic to selectively allocate budget to those research subjects that are not in the basic research stage but could be put to practical use?

How would selective allocation of budget be made consistent with the Master Plan?

• Although in the world of technological research some discrepancy often occurs between the current status and the initial project plan, we have to think of the budget allocation, keeping the original purpose of the RWC Project in mind.

• We have to clearly indicate the primary purpose of the RWC Project during its last two years and the selective allocation of budget must correspond with that objective.

• When we allocate budget, we should decide which subjects could be a great success if more effort was made. At the same time, we must take into account the possibility of asking them to review their plan thoroughly.

• When the project started, researchers were encouraged to choose difficult subjects and how well it could be attained was of secondary importance.

• If we put excessive emphasis on the commercialization of research results, their original value could be impaired. Although some social changes are inevitable, we should respect the initial policy of the project.

• A textbook on disseminating research results should be prepared. If researchers could have such a textbook and start their research based on the results of the RWC Project, that textbook itself would be a sort of an outcome. This should be considered in areas of theoretical research.

How should evaluations be conducted regarding the budget allocation for the following fiscal year?

Researchers will be required to submit their implementation plan for the following fiscal year. We would like them to specify in their implementation plan what the basic points of evaluation are. So, please indicate what your evaluation standards are.

As for commercialization, we should prepare evaluation criteria different from those of private companies.

• For private companies, the standard is whether or not it yields a profit.

• When a very effective system is developed, it is highly evaluated. But that is not the case in the business world.

• In addition, we should bear in mind the fact that really good technologies are not necessarily used by private companies.

You must avoid, after lengthy discussions, reaching a conclusion similar to the one you made last year. I would like you to allocate budget to each research subject on a case-by-case basis so that researchers feel their activities are evaluated properly.

The results of evaluation may differ from evaluator to evaluator. Iｹd like to emphasize that the way of using evaluation results is more important. You should be careful in using numerical evaluation results as well.

3. Final image

Besides evaluating the results of each research subject, you also need to selectively allocate budget to support the final image of the entire project.

We certainly have the final image of each subject. But the point is how we can establish the final image of the entire project.

Since you have been trying to change the world based on the purpose of the Master Plan, what counts more is substantial success rather than formal success.

Bearing in mind that substantial success is more important, we would like to create the final image of each research subject in the RWI Center and in the field of real world intelligence technology.

There were two interpretations about the master plan at the beginning of this project. One is that basic and explorative research should be conducted in the first phase and commercialization would be the purpose of the second phase. The other interpretation says that this project is exclusively for fundamental research.

• Which course should we take?
• A decision should be made equally considering both the "fundamenta" and "commercialization" approaches.

Since the RWC Project was regulated by GATT (WTO) when it started, we had no choice but to target fundamental research. However, the project became more commercial after it became free from WTO regulations.

Since the viewpoint for evaluation is different from that of private companies, we should take the difference into account when considering the final image.

• The purpose of the project is to create infrastructure for the development of technologies that could be commercialized. I would like you to discuss the final image of the project based on this purpose.

We will have to define the final image in the near future. We would like to clarify the image at meetings of the Promotion Committee next fiscal year. It is therefore an issue for future discussion.