This morning we will ask each witness to make an initial statement. This afternoon we will ask our guests to participate with members of the subcommittee in a general discussion.

Our leadoff witness is Dr. Robert Fano of the Massachusetts Institute of Technology. Dr. Fano, a native of Italy and an American citizen since 1947, is one of our leading electrical engineers and computer experts. He is the associate head of the Electrical Engineering Department at MIT in charge of computer science and engineering.

Our second witness is Dr. Donald L. Bitzer of the University of Illinois at Urbana. Dr. Bitzer is also an electrical engineer and is well known for the work he has done as director of the Computer-Based Education Research Laboratory at Illinois. I might add that in 1964 Dr. Bitzer served as a computer expert for the Agency for International Development in India.

Dr. Bitzer has brought a team with him, and at the appropriate time, I will ask him to introduce his associates.

Our third witness is Dr. Paul Zukin, a medical doctor who is with the School of Public Health at the University of California in Los Angeles.

Our last witness this morning is Dr. Weston Vivian who is an electronics engineer and is well known to most of the members of the subcommittee because he was a colleague of ours in the 89th Congress. Dr. Vivian is a native of Canada who has lived in the United States since 1929. He has had broad experience in the practical application of information and communications technology.

We certainly do appreciate you gentlemen coming to help us out with our problems. First, we would like to hear from Dr. Robert Fano. I think it might be helpful if all of the witnesses come forward and sit at the table. We might as well get acquainted with you now. [Off the record discussion.]

Mr. MOORHEAD. Dr. Fano, you may proceed as you see fit.

STATEMENT OF DR. ROBERT M. FANO, ASSOCIATE HEAD, COMPUTER SCIENCE AND ENGINEERING, MASSACHUSETTS INSTITUTE OF TECHNOLOGY

Dr. FANO. Mr. Chairman, and members of the subcommittee, I welcome this opportunity to make a few introductory remarks on the information needs of society and on how information technology might be exploited to meet them. I hope they will be of some help in establishing a suitable frame of reference for the specific examples that will be presented by the other members of this panel and for the topics that will be discussed in the following hearings.

Communications technology has extended the range of our senses by enabling us to reproduce at a distance printed text, then sound, and finally pictures. The much older technology of printing had already enabled us to record, and transmit over time our collective knowledge, and to provide convenient and widespread access to it. These capabilities have greatly facilitated the communication process, in a broad sense, on which the progress and the very functioning of organized society depends. Distance and time are no longer serious limitations;

instead, communication within society is primarily limited by the ability of people to identify, comprehend and utilize the information available to them. Now it is the human mind that needs to be extended.

This limitation of the human mind has been largely circumvented in the past through the assistance of what I call knowledge specialists-I mean by that, lawyers, teachers, physicians, et cetera-but it is doubtful that the needs of society can be adequately met in this manner in the future. In the first place, the information needs are increasing rapidly because of the growing interdependence and operational complexity of society, and also because of the arising expectations in our society. In addition, it is becoming increasingly difficult for even specialists to keep abreast of their own fields of specialization, and they are forced in the direction of knowing more and more about less and less, which means we not only need more specialists, but also more specialties which compounds the problem. This suggests that society may not be able in the future to provide for itself the necessary knowledge-based services in the present manner. We must develop means of increasing in a major way the productivity in this sector of our economy. We must also enable individuals to do more for themselves, by utilizing more effectively the knowledge that is, in principle, accessible to them. The tools for achieving these goals must come from the information technology that is developing around the general purpose digital computer.

A first step in this direction took place in the sixties with the development of ways and means of making computers conveniently accessible to people as individuals, of making them useful as personal assistants in a variety of fields, and of interconnecting such computers into networks stretching from the east coast to the west coast. I would like to say that much credit for these developments should go to the leadership and support provided by the Federal Government, and in particular by the Advanced Research Projects Agency of the Department of Defense. We are fortunate that the needs of the Defense Department are very similar to those of the civilian economy. A whole new service industry has emerged from these developments. Yet much remains to be done.

Part of the problem that we face in the service sector of our economy is that knowledge and experience cannot be readily transferred from one person to another. A substantial fraction of our life is spent in acquiring them, in spite of the fact that many other people already possess them. Could computers be made to provide assistance to people where significant knowledge and real intellectual skills are required? Could computers be made to act as knowledge brokers, capable of acquiring knowledge from some people and making it available to others or even utilizing it directly in their behalf!

A positive answer to these questions is suggested by recent research. An example is provided by an existing system which exhibits knowledge and skills in an important area of mathematics comparable to those of a competent applied mathematician. A very different example is provided by a system which answers questions and executes instructions expressed in English. I have distributed a sample dialog with this system which is taken from a recent book by the developer of the svs

tem, a colleague of mine, Professor Winograd. It is clear that the system understands not only what it is being told, but also what it does, in the sense that it can explain its own activities. I must hasten to stress that it can do so only within the context of a very simple and very restricted world, namely, a world of children's playing blocks. Much further progress must be made before a conversation of this type can be carried out over a broad area of practical importance.

I would like to spend a minute making further comment on this dialog because there is a part of it that is of special social significance. Now, if you would turn to page 13 of what has been distributed, I want to read some of the dialog. Well, perhaps I had better not read it directly but comment upon it. This part of the dialog answers questions having to do with how the computer did certain things and why it did them. This is the first example I have seen of a computer system being able to explain to a person in an understandable way its innerworks, why it does what it does. This is extremely important because one of the dangers in the use of computers is to get to a situation in which computer systems are so complex that no one person any more understands how they work. This is a real possibility, in which case we may be faced with the choice of either having to accept without questioning at the conclusion reached by a computer system, or otherwise having to reject them. And I am afraid that we will not be able to afford to reject them. So, this is the first example that I have seen of something analogous to the conversation that goes on, say, between two people, one of whom is proposing to the other some course of action. Of course the latter person, if he knows his business, is going to ask how such conclusions were reached, and other questions intended to check that the proper factors, the proper constraints had been taken into account in arriving at the conclusion. So, I regard this first example as rather important.

Another aspect of the problem we are facing can be illustrated by analogy with the mass production of goods. The techniques of mass production have increased not only the quantity of goods available to us, but also their potential quality. Specifically, the quality of mass produced goods is limited only by the total knowledge and capabilities of the society, while the quality of goods produced by an artisan is strictly limited by his own knowledge and skill. While other people may have greater knowledge and skill, the person involved will have to learn and acquire the skills before those can be made useful in his production of artisan goods.

With respect to knowledge-based services, we are still at the artisan stage. Their average quality is far from satisfactory, and still their availability is inadequate, or, equivalently, their cost is excessive. What is needed, by analogy, is mass production of services, that is a way of providing services that utilize the total knowledge and capabilities of the society, in addition to increasing substantially the overall productivity. For this purpose, information, instead of matter, must be transported and brought to bear on a specific situation, where and when the need arises. Thus, a widespread and economic computercommunication network is clearly essential to the mass production of services, just as a widespread and economical transportation network has proved to be essential to the mass production of goods.

I believe that such a computer-communication network must ultimately appear to its users as a single system to avoid introducing artificial and possibly dangerous barriers between people and segments of society and between different aspects of their activities. I must stress this does not imply that all information networks must be owned or operated by a single organization, but it does imply that all such networks must be interconnected, and their users must be largely unaffected by the internal structure of the overall system. Where do we stand today with respect to our goals? The performance of computer hardware is increasing very rapidly and its cost is decreasing also very rapidly. Computer hardware is no longer a limiting factor. Our present communications network, which was designed and is still being used primarily for telephony, is not well suited to provide communication to computers and between computers, but I would not regard its characteristics as a major barrier to progress, at least not at this time. I must say, however, lower communications costs would be most welcome.

The real limitations in my view stem from the present state of our software technology.

In the first place, we do not know at this time how to implement practical knowledge-based systems of the type that I have suggested. Progress is being made, but the current level of effort is, in my view, inadequate in view of the difficulty and importance of the problems that confront us. In the second place, the development of computer software is still largely an art, rather than an engineering discipline. We do not know how to manage large software developments and there are large differences in productivity between people whose qualifications appear similar. As a result, the cost of developing computer software is inordinately large, and often much larger than expected. In addition, the probability of failure is still discouragingly high. I would like to add that I read recently a study done for the Air Force in which they have estimated that between 4 and 5 percent of the 1972 Air Force budget was spent on software development, and that was several times the cost of hardware. That gives you an order of magnitude of what is involved, and it seems that the situation is very similar in the civilian economy. I must conclude that substantial progress in software technology is needed, if we are to satisfy the information needs of our society. This is the most important block in my view.

Finally, I would like to stress that the social impact of information systems will depend strongly on the specific technology that we choose to develop. I understand that social impacts are going to be considered in a later hearing and I do not want to dwell on this. However, I would like to recite one example. A good example is provided by the protection of individual privacy. The development of means for protecting information within computer systems and for controlling its use is still in its infancy, largely because of past neglect by the industry. While a few systems exist in which access to individual files can be granted and withdrawn selectively, the vast majority of computer systems do not permit even this basic form of control. Yet much more elaborate forms of control are needed, including means for controlling

and auditing the controls themselves, according to externally established lines of authority and responsibility.

Finally, I would like to stress above all, and I really mean it, the tremendous importance of making computer-communication systems truly accessible and useful to the entire population. This is an extremely important matter in my view. The purpose of such systems is to facilitate access to knowledge, and knowledge, when restricted to any particular segment of society, inevitably becomes power on the part of that segment over the rest of society. I sincerely hope that we will be able to refocus our efforts toward helping people to cope more successfully with the tasks that confront them in their private and business lives, and succeed in narrowing the gaps of knowledge and power in our society that have already been created by computers. [Attachment to Dr. Fano's statement follows:]

SAMPLE DIALOG WITH A COMPUTER PROGRAM

(From the book by Terry Winograd, “Understanding Natural Language,” Academic Press, 1972)

This "dialog" exhibits highlights of results of research on how computers could be made to "understand" the material they are dealing with. The demonstration is genuine to the extent that the computer programs "understand" what the words mean, but only in relation to the very small "world" of simple physical blocks and towers. The system can handle, today, only a very limited vocabulary. But more important, it can handle only a very limited variety of meanings of the words.

Research on "meaning" is now developing in a healthy, useful way, but it is still in an early stage and much more must be learned. The remarkable smooth performance shown in the dialog should not mislead readers into thinking that such systems could be readily developed at this time for broad areas of practical importance.