Toward the Information Systems Melted into Daily Life

Hidehiko Tanaka Chairperson, Promotion Committee of the Project for Fundamental Information Technology of the Next-Generation, Ministry of International Trade and Industry (MITI) (Professor, University of Tokyo)

Today, I'd like to give a speech under this title (Fig. 1).

I'll start with my views on daily life as it is today, and then proceed to describe how the RWC Project could affect it (Fig. 2).

Now we are in the year of 2000, many things are said - What environments would be new for information systems? What types of systems would be required? - which made me think of essential technologies when considering researches in RWCP. First, let's consider modern society and computers (Fig. 3). I'm sure all of you already use business tools such as PCs, and a variety of underlying systems are used behind our daily life. But although PCs have been made simple to use, only 30% of us use them after learning. And another set of underlying systems are interconnected as critical infrastructure systems behind the scenes.

When viewing computers from bottom up, they are still unreliable (Fig. 4) - PCs still cause many headaches, and when large systems go down, we get upset.

Next are the communication bottlenecks - the communication bandwidth is not wide enough to exchange data freely, and mobile communication is really unstable. Also, information devices are difficult to use. OS's are changed or upgraded every couple of years, forcing us to buy the latest one. Moreover, there are too many types of PCs, and we have UNIX as well as mainframes. They are not easy to use, they are code-incompatible, and their capacity is not sufficient in many ways.

The storage capacity of current PCs is quite large, several powers of 10 over those in the past. But let's calculate roughly how much data is memorized in one person's brain over several decades from birth. Scenes are instantly stored as images; assuming that 100 images are memorized in a single day, approximately 10 terabytes of images are stored in 50 years. This volume can be achieved by current computers, but only in large systems. So storage capacity is still insufficient if we want to use computers as our personal secretaries. This is also true for processing speed. When attempting something new, many applications still require the speed to be several powers of ten.

Next, there are five main new environments frequently mentioned (Figs. 5 & 6).

First, the electronic society, such as e-commerce and the electronic community. The latter means that locality- and blood-based communities are being replaced with network-based communities. E-business consists of content, communities, and commerce, in which the provision of information and commercial transactions are mixed, while consumers and providers are also mixed. For instance, B-to-B systems for exchanging data are estimated to be worth up to 70 trillion yen by 2003. B-to-B systems are those among enterprises, while B-to-C refers to systems linking consumers and enterprises, which are estimated to become a 100 trillion-yen industry within several years after 2003.

Second, internationalization. Multi-lingual ability will be required, and as Mr. Yasunobe said, society is becoming increasingly environmentally aware. It is predicted that by around 2010, revenues in the environment industry will be 35 trillion yen, based on an electronic society in which things transferred physically will be transferred logically or more efficiently through information processing. In this sense, it may be considered problematic to keep workstations or PC's powered up.

Next comes the aging society as outlined in the former speech (Fig. 6). Again in economic terms, the health care and welfare industries will be worth some 90 trillion yen in 2010. However, this is not necessarily bad, merely meaning that many experienced labor markets will emerge. But again, the information industry will be at the core.

Finally, "Method of Handing Down Knowledge to Future Generations" means that the conventional use of paper for this purpose will be replaced with electronic media. Of course, knowledge and intelligence are not solely conveyed on paper, but "implicit knowledge" is carried in other forms, for instance, skills, schools, and work itself. Electronic media will also be able to help these to be conveyed successfully. Thus, the electronic information industry will also play a very important role. In brief, all of the above constitutes a new environment.

Our daily life will rely heavily on information systems (Fig. 7). Therefore, information systems must be reliable. Large-scale computation and super-high speed are essential and usability is the key as the systems will be used every day. Most of us will use them, but 70% are novices to information processing. Also, seamless connection is a key to build them into every device, and networks are the key for location independence.

Considering the above, information systems in the future must provide everybody with accessibility to computing resources (Fig. 8).

Here, "computing resources" cover the whole range of process performance, software, data, and hardware. Of course, which of these has the greatest weight depends on the level of life, level of business or something else. Anyway, accessibility is a must. Anywhere, anytime, various resources must be seamlessly accessible.

This idea was first conceived as a resource-sharing network around 1969, when the first computer network called ARPA Network emerged in the USA. Finally some 30 years later, the general public can use it. Figure 9 shows such systems which provide people with a multitude of resources.

Image of future information system

Fig. 9

At the top is those with the highest performance and capacity, and at the bottom are those which focus on ease of use. "Server Systems" in the second row may be better described as new mainframes. The third layer contains advanced PCs, and the fourth layer contains mobile Internet, where many accesses are made very easily via the Internet. Embedded systems to the right refer to various built-in information devices. Between (3) and (4) lie home terminals. The further we go to the right, the more machine-oriented, and the further we go to the left, the more human-oriented. They are all connected via the Internet.

What properties are required for such systems (Fig. 10)?

The properties correspond to the respective layers, and first up are reliable fundamental systems. Fundamental systems must be reliable because everyone relies on them. Next, as shown at the center of the Figure 9 as PCs, are usable personal assistant devices, which correspond to PCs. "Usable" does not mean usable when learned but that users find them easy to use. These include devices associated with keywords such as user-friendly, easy-to-use business tools, robustness, and electronic personal secretary. Corresponding to "Embedded Systems" on the right of Figure 9, they are embedded into every device so that the Internet can be used anywhere.

I will describe them more in detail. First, for reliable fundamental systems (Fig. 11), the system structure must be solved.

"Small reliable open kernel" means that certain OS's for fundamental systems have a long history since the mainframe age. In addition, UNIX-based systems have proliferated as their reliability has improved. Currently, Windows is particularly popular. However, operating systems, especially the functions of their kernels, are already understood by everybody. It seems strange to me that the people who made these earned huge sums. OS's are known, which makes them open. OS's are now in the public domain as common property, on which a wide variety of software will be implemented. If their interface is known to everybody, a LEGO structure can be built, as somebody else's work can be used as a base for further building. This is the key. OS's in the future may have a known functional structure augmented by improved reliability. For instance, reliability may be supported by reliable non-volatile high capacity memory.

Usable personal assistant devices (Fig. 12) focus on assisting not only business but also daily life at the individual level. While solving the problems of current PCs, such devices must be enhanced by implementing personal assistant functions and giving them intelligence to make them easier to use.

For personal assistant devices (Fig. 13), a personal secretary function is desirable so that everybody has his or her own computer secretary. "Tagging Function: Event Log" means that we need systems which simplify the tagging and logging of events, i.e., our activities, to record what we do in many places and store them as past experiences. And it is natural to have personalized portal functions.

This leads to the importance of security (Fig. 14). For true personal assistance, we will be able to acquire knowledge easily from anywhere, and so the ability to use that knowledge will be more important than the amount of knowledge. Computers must support individuals for a long period of time instead of having to be replaced frequently every two years as at present. Long term will become a key word. It's troublesome to have to replace a PC frequently and to transfer data from one to the other - it would be more convenient, as mentioned earlier, to have a function for storing and summarizing experience as knowledge.

For ease of use by making PC's intelligent (Fig. 15), artificial intelligence will be put into practical use. As we shift from conventional research on the potential of AI to usable AI, we need to use knowledge by restricting domains, applications, and individuals.

Embedded devices (Fig. 16) mean that every device needs real-time operations and seamless connectivity.

Mobile Internet (Fig. 17) is an information system for the general public at the bottom and provides an entry to sophisticated electronic business systems. Currently in Japan, there are 50 million cellular phones and PHS's in use. We need to take advantage of these units. Mobile Internet may quickly become a reality, especially in under-developed countries where investment in cables is difficult. In Japan, mobiles are popular, and their popularity will continue to grow.

Figure 18 summarizes the important technologies in the future.

First, a global architecture, which explicitly defines the data structures determined by many kinds of applications such as integrated use of three layers and the interface, is required. It must be published to ensure standardization. For instance, when designing a system, currently we buy PCs and servers and connect them. The structure of data flowing among them is determined by, for instance, Word or Excel, when buying PCs. This should not be the case in the future. Instead of PCs and/or servers determining the data structure of a system, the selection of applications should also determine the data structure, which should be standardized. After that, PCs and servers can be specified and a system designed in the reverse order of the current design, which at present hinders development.

Technologies to improve reliability are also very important.

Figure 19 is a simplified view with many terminals and PCs at the bottom, which are used to access various services shown above a network. At the center is an access network. The idea is that the S1-Sn systems above also use systems below as a loop. Services above are implemented on systems S1-Sn and also created by using other systems via the access network. This loop is essential. How this loop is created freely is a key to developing the system. The data structure must be standardized in this sense.

Fig. 19

Other important technologies (Fig. 20) include easy-to-use computers, seamless computing, and communication networks with megabit bandwidth per user, PC performance of 10 gigaflops, multimedia databases, and usable intelligent systems.

Figure 21 shows the importance of these technologies over time.

Fig. 21

The vertical line indicates the year 2000 (now) and the horizontal axis is the time in years. The black solid line indicates operating systems, the black dotted line business software, the blue solid line Internet tools, and the blue dotted line knowledge functions. Since operating systems are prevailing knowledge, they will become an affordable commodity, dropping most quickly. Business software will shift its focus onto techniques for integrating existing software, so it will drop next. Various Internet tools will flourish. And knowledge functions, still in their infancy now, will emerge as a mainstream technology. This is my view of the years to come in terms of these four technologies.

From this point of view, with easy-to-use reliable computers becoming the norm as additional feature, the current cost-based PC competition may move back to feature-based competition (Fig. 22).

Thus, small ideas may amount to major opportunities. Another thing to consider is system construction based on seamless computing or an open data structure, and this policy should be promoted.

Based on my view of the world as I've explained so far, next I'd like to outline researches conducted in the RWC Project (Fig. 23). They can be roughly divided as shown in Fig. 23. Details will be given in subsequent speeches together with demonstrations, so only the keywords and fields are covered here.

First, Parallel and Distributed Computing, and second, Real World Intelligence (Fig. 23). Among them, several technologies are emerging as readily available.

Examples (Fig. 24) include the PC Cluster System and Evolvable Hardware. PAPIA is a practical parallel application and is already available.

One of the important fundamental technologies of the future (Fig. 25) are those which enable high-performance parallel systems.

The open MP compiler, parallel library, and programming environment are all feasible technologies, and switch networks and optical devices are key technologies for high-speed links. And various real-world databases called Intellectual Resources are emerging, and will be used in many R&D activities.

In addition, to create a more natural interface, there are technologies for multimedia processing, recognition, interaction, and integration (Fig. 26). Recognition and interaction will be a key word for future products and key technologies to build a more natural interface. And learning, recognition, and self-organization are also important fundamental technologies for intelligent functions. Fundamental technologies for handling vast amounts of data include self-organization, feature extraction, and data mining. And then there's robotics. Various real-world technologies are in fact basic technologies for robotics. Intelligent systems with hands and legs as well as heads are typical robots, in which sense, real-world intelligent technologies constitute their foundation.

Right now everybody related with the RWC Project is working toward producing the final results (Fig. 27). In the 10-year projects, the remaining 2 years should be a concluding period, but in fact in today's fast-moving world, two to three years are long enough to see many changes, so we can say that actually we still have two years. This means that we can do many things if we want to, so let's do our best.

One issue that we must address is to show our technologies. We need to show our technologies to many people to identify the potential. This should lead to future technological progress. The rough stones created thus far in the RWC Project should be polished to reveal any luster they may contain.