The characteristics of the NIPT Program: information in the actual world, and (1) The massively parallel system, neural system, and the integrated system are constructed and are composed of millions of element processors. Robust system composing technology and the adaptive control technology for load distribution to the element processors are needed. passes appropriate and advanced judg- (2) Important subjects are to develop autonomous manner. (3) In order to construct the system composed of millions of element processors, consideration should be made to the technology of integrated circuits to carry plural element processors on the wafer and the application of the optical device technology with ability of massively parallel transmission and processing of information. In my earlier reports I have commented repeatedly on the vagueness of the proposed NIPT (RWC, 4DC) program. Some of this vagueness may have resulted in a rather imposing view of what might be accomplished. For example, in the NTW mentioned above, it is written that "the Japanese are looking for ways to make a million processor, ultra-supercomputer, artificially intelligent thinking machine." MITI's own documentation hopes that this program will be "a means of realizing real-world computing that flexibly deals with imperfect, ambiguous, and changing The workshop began with a half day of plenary talks in which the four subgroup moderators gave their views of the eventual research thrusts in their areas, followed by 2 days of four parallel sessions, one on each of the four subgroup topics. On the last day a general session gave the moderators an opportunity to present a very rough master plan for each group. These were done under a great deal of time pressure (the session moderators are to be congratulated for achieving as much as they did), but these plans had not been coordinated and I felt that they needed much further refinement. I should mention that although referred to as proposals, in fact most of the presentations were more like extended abstracts, without budgets, specific task descriptions, milestones, evaluation criteria, etc. Individual proposal descriptions were labelled as not for distribution. However, pro labelled and these are in the Appendix. Interested readers are encouraged to contact the authors directly for more detailed information. In this workshop proposals were broken into one of four subgroups, Theory and Novel Functions for Applications, Massively Parallel Systems, Neural Systems, and Optical Computing Systems (corresponding to T, M, N, and O, respectively). I believe that this stems from NIPT views that show optical technology as a fundamental tech-posal titles and authors were not so nology underpinning both a massively parallel and (perhaps equivalent) neural system which, in turn, uses new theory to help develop new and important applications. Earlier documents had listed eight basic areas with some overlap; at the moment these have been reaggregated into four. With the help of the authors, the organizers tried to place the proposals fairly, but several placed in T could have as easily been put into N. Nevertheless, at this stage it is fair to say that there is very little integration between the four sections, and this will be a major task for the steering committee. (This opinion was quite generally shared by the participants; some even wrote as much in their proposals.) Perhaps it will be impossible to do more than give lip service to this integration, as the time constants for some of the research proposals were vastly different. Also, as yet there has been no attempt to integrate the proposals within a subgroup; certainly the Western scientists had no idea of what they would hear when they arrived. My understanding is that the steering committee will now attempt to decide which of these are to be moved forward, and these individuals will be asked to produce more detailed proposals. I think there will be many changes resulting from the current workshop. Some changes were announced even before the workshop was ended. Original plans were to have two additional feasibility workshops, in January and March 1992. It is now the intention to have one more workshop during the week of 2 March, at which time an integrated master plan will be presented. Project funding will begin the following month. Most of the decision making will go on in executive sessions, and the general feeling among the attendees was that individuals with whom collaborations were to be established had already mostly been selected. There will certainly not be a refereeing process in the Western sense; however, the committee members are very senior scientists and readers should expect that the projects funded will be internationally credible. It is not entirely clear, though, how new ideas that surface during the 10-year duration of the program will be handled. I mentioned above that participating in this project is a high prestige item. This was echoed by a German colleague who felt that the financial support, while welcome, was really incidental. It is presumably the reason that the Korean proposal in the Theory subgroup was presented with a clear statement that only technical collaboration, not financing, was being requested. In the Optical Computing Systems subgroup, another Korean proposal on Optical Neural Systems seemed to indicate an interest in obtaining advanced Japanese optical devices as part of the international collaboration. Similarly, a Singaporean was seeking funding but probably would be happy just to collaborate technically. There is also good reason to think that prestige, as well as being "plugged in," is a major factor in most Japanese company participation. MITI has made significant efforts to make this a program that is not only focused on Japanese industry. In July 1991 ["NIPT Feasibility Study and Workshops," Scientific Information Bulletin 16(3), 5-6 (1991)] I reported that there was a major change in the way ownership of intellectual property was to be handled, with equal sharing between the Japanese Government and the inventor. It is also important to note that for the NIPT project there is now the possibility that Japanese university researchers can obtain funding from MITI. This could supplement funding from the (much poorer) Ministry of Education. One Japanese professor told me that this possibility of new funding was the major reason he was participating. Another told me that this source of new Japanese Government funding for universities would be of most benefit to younger faculty whose connections have not been so well established. More senior faculty, he claimed, have many opportunities for support from government and industrial sources. At least one European observer felt that the Ministry of Education would not stand by while MITI infringed on its jurisdiction of university funding. The most concrete of the sessions concerned massive parallelism. Here I was told repeatedly by attendees that the focus is that ETL wants to continue the focus is that ETL wants to continue its research on generalizations of the dataflow model that it has been working on for some time and for which the laboratory has a great deal of hardware experience. They want to build a 10,000processor 1-TFLOP machine within 5 years as a prototype and a one million processor machine within 10 years. As there are already commercial machines with tens of thousands of processors (albeit simple ones), this does not appear to be a revolutionary step, although there are many, many problems to be studied concerning the computing model, reliability, and software development. The emphasis in this development will be a machine that can adapt to various required computing paradigms and so will function as a general purpose resource. Nevertheless, the working language will probably be some type of parallel extended object-oriented language. I was told that at least one Japanese company proposal would likely be withdrawn (although they would probably continue as an industrial partner), as it seemed clear to the proposers that ETL had its own view of how it wanted to proceed. After listening to the individual proposals, the session moderator, Dr. T. Shimada of ETL, presented a draft master plan (of his session) to the entire workshop. This draft does not suggest that optics will play a major role, except perhaps at the level of interconnects. This is sensible, as the hope is to begin building in the next fiscal year, whereas optical technology, while very promising, does not seem to be at the same stage of readiness. During the final plenary session, the draft plan actually did not explicitly highlight bias towards a dataflow machine but listed dataflow as one option in a list of options that needs more study. Expectations of the program, according to the steering committee, are still focused on integration of (imperfect) information, adaptation, learning, approximate optimization, etc. Specific functions that were mentioned in the overall plan include speech understanding based on constraint satisfaction, flexible action control in robots, as well as applications to massively parallel, integrated, neural, and optical systems, etc. The theory session was composed of about 20 essentially independent proposals, ranging from a 50-year proposal from Mitsubishi research for computer vision understanding; R. Eckmiller's proposal to deal with international issues such as smog modeling, global warming, etc.; down to a 3-year proposal from Singapore's Institute of Systems Science (ISS) for development of what will essentially be a product. Most of these papers represented good science, but even if all are funded and all were to succeed nearly to their proposers' expectations, the fundamental goals of the program may not be attained. Thus readers should not expect that whatever hardware and software are eventually built within the scope of this program will result in a system that is able to realize when we are telling a joke. Further, it does not seem that the most advanced, hence ambitious, of the research from this subgroup could have much direct impact on the design of the massively parallel system, probably coming too late in the cycle. I asked about computing resources that would be made available to collaborators in this project. (The a-building massively parallel system will not be ready at the early stages as a research tool for users.) The organizers have not settled this question but spoke of making commercial parallel systems available and also of providing a high speed network for communication and experimentation among the participants. The neural session resulted in a draft master plan stating that they hoped to build a 10-TCUPS machine in 10 years. Last spring, Hitachi demonstrated a working 2-GCUPS wafer scale neural system. Thus 10 TCUPS appears to be a realistic target within 10 years, and certainly not an unexpected one. One change coming out of the workshop was that this committee has decided not to validate the system on real applications but only on "novel functions for applications," i.e., application building tools. The optical computing systems session emphasized the optical interconnect technologies and the supportive device and materials processing. The sentiment was that this type of work needed to be done before realistic sizings of processor architectures could be made. Of the 13 Japanese proposals, 8 were really expansions of and capitalization on existing industry (6) and university (2) optoelectronic device R&D. There was a ground swell of opinion that the session should report back only on devices for interconnects and leave the other two assigned topics (optical digital and optical neural) for other subgroups (namely, massively parallel and neural systems). The session moderator agreed with the technical reasoning, but said it would have to be discussed after the workshop. Virtually all the major Japanese companies, except NTT, were represented with these types of optical device proposals. Appendix PROPOSALS PRESENTED AT THE FIRST NIPT WORKSHOP 1991 Session T1 Theory and Models (1) T1-1 Computational Learning Theory of Probabilistic Knowledge Representations T1-2 Ecological and Evolutional Computation Models for Massively Parallel/Distributed Systems T1-3 Pattern Recognition Using Neural Associative Memory Lui, Ho Chung (National Univ. of Singapore, Singapore) (ISSLHC@NUSVM.BITNET) Session T2 Theory and Models (2) T2-1 Fundamental Research on Information Integration Technology Based on Multi-Neural Network T2-2 A Vision Processor in Neural Architecture Banzhaf, Wolfgang (Mitsubishi Electric Corp., Japan) (BAN@RELACHE.ADM.CRL.MELCO.CO.JP) T2-3 Hyper-Heterarchical System for Auditory Scene Understanding Kawahara, Hideki (Nippon Telegraph and Telephone Corp., Japan) (KAWAHARA@SIVA.NTT.JP) Session T3 Recognition and Understanding T3-1 Sensory Integration Sugie, Noboru (Nagoya Univ., Japan) (SUGIE@SUGIE.NUEE.NAGOYA-U.AC.JP) T3-2 Research on Information Fusion Methods in Pattern Recognition/Understanding T3-3 A Multi-Modal Human Interface Chiba, Shigeru (Sharp Corp., Japan) (SCHIBA@TOKEN.SHARP.CO.JP) T3-4 Action Image Database Abe, Hasahiro (Hitachi Ltd., Japan) (ABE@CRL.HITACHI.CO.JP) T3-5 Recognition of 3000 Chinese Characters Handwritten Kanji Characters Muehlenbein, Heinz (GMD, Germany) (MUEHLEN@GMDZI.GMD.DE) Session T4 Multimedia Information-Base and Interface T4-1 Self-Organizing Information Base Hirayama, Masaharu (Mitsubishi Electric Corp., Japan) (HIRAYAMA@SYS.CRL.MELCO.CO.JP) T4-2 Multilingual Natural Language Interface/Interaction for Human and Computer Ahn, Byung-Sung (Electronics and Telecommunications Research Institute, Korea) (D6000@KIET.ETRL.RE.KR) T4-3 Some Novel Functions for the NIPT Computer Interface Inoue, Hirochika; Sato, Tononasa; Hirose, Michitaka (Univ. of Tokyo, Japan) (INOUE@JSK.T.U-TOKYO.AC.JP) T4-4 Self-Organizing Adaptive Mechanisms Based on Uncertain Information Taura, Toshiharu (Nippon Steel Corp., Japan) (TAURA@ELELAB.NSC.CO.JP) Session T5 Inference and Problem Solving T5-1 Learning Neural Computers for Global Forecasting and Control - Theory and Applications T5-2 Research on the Dynamic Inductive Inference Sakurai, Akito (Hitachi Ltd., Japan) (SAKURAI@HARL.HITACHI.CO.JP) T5-3 Coordinative Distributed Problem Solving Model by Structured Agents Fukumura, Satoshi (Kawasaki Steel Corp., Japan) (FUKUMURA@SYSTEM.KAWASAKI-STEEL.CO.JP) Session T6 Integrated Systems T6-1 Development of a Flexible Hand-Eye Robot in Neural Architecture (refer to the proposal of T3-5) T6-2 Human Interface Using Artificial Reality Kurihara, Tsuneya (Hitachi Ltd., Japan) (KURIHARA@CRL.HITACHI.CO.JP) T6-3 Research on Flexible Information Processing Models - An Autonomous Growth System Asakawa, Kazoo (Fujitsu Laboratories Ltd., Japan) (ASAKAWA@FLAB.FUJITSU.CO.JP) Session M1 Architecture (1) M1-1 Research and Development on Composite Super-Fine-Grain Architecture for Massively Parallel Computing Systems Hiraki, Kei (Univ. of Tokyo, Japan) (HIRAKI@IS.S.U-TOKYO.AC.JP) M1-2 A Dataflow Massively Parallel Computer Shimizu, Uasahisa (Sanyo Electric Co. Ltd., Japan) (SHIMIZU@EDDEN.HIRAKATA.SANYO.CO.JP) M1-3 Research and Development of a Massively Parallel Architecture Based on the Extended Dataflow Model Session M2 Architecture (2) M2-1 Adaptive Massively Parallel Machines Koike, Nobuhiko (NEC Corp., Japan) (KOIKE@CSL.CL.NEC.CO.JP) M2-2 A Massively Parallel Machine with Optical Interconnection Oyanagi, Shigeru (Toshiba Corp., Japan) (OYANAGI@ISL.RDC.TOSHIBA.CO.JP) M2-3 Massively Parallel Image Processing Sub-System Mizoguchi, Masanori (NEC Corp., Japan) (MIZO@PAT.CL.NEC.CO.JP) |