Coping with the imprecision of the real world
Lotfi A. Zadeh · Communications of the ACM · 1984
The tools researchers use to probe certain AI problems, says this Berkeley professor, are sometimes too precise to deal with the "fuzziness" of the real world.Q. Professor Zadeh, in this interview today, you agreed to talk mainly about the limits of traditional logic in dealing with many of the problems in the field of artificial intelligence (AI) and your approach toward helping to overcome those difficulties.Before getting into those issues, though, could you first give our readers a brief overview of what you see as the major areas for computer applications in the years ahead?ZADEH.In the years ahead, there will be three major areas of computer applications.One, in the traditional vein, is the use of computers for purposes of numerical analysis.Numerical analysis will be very important in a number of fields--particularly in scientific computations and simulation of large-scale systems.For such purposes, there will be a need for larger and larger computers.This is especially true for applications in meteorology, in nuclear physics, in modeling of large-scale economic systems, in the solution of partial differential equations, and in the simulation of complex phenomena like turbuldnce, fluid flow, etc.Area number two will be concerned with masses of data--large databases.This is the sort of thing that is playing and will be playing an important role in banking, insurance, records processing, information retrieval, etc.What will be important in these areas is not so much number-crunching capabilities as the capability to store massive amounts of data and to access whatever data are needed rapidly and at a reasonably low cost.Furthermore, in these areas, computer networking, of course, will be playing an essential role.For you will have to have access not just to a single database but to a collection of interconnected databases.In response to this need, we will see many advances in computer networking during the next several years.The third major area for computer applications is what has come to be known as knowledge engineering.This area has received considerable publicity during the past few years, particularly since the Japanese have ©1984ACM0001-0782/84/0400-0304 75¢ highlighted it as an area of prime concern.This is a rapidly growing field in terms of importance and breadth of applications.Knowledge engineering is one of the major areas of AI.And within knowledge engineering, a field of primary importance is that of expert systems.True, there may be exaggerated expectations of what expert systems can accomplish at this juncture, but as Jules Verne observed at the turn of the century, scientific progress is driven by exaggerated expectations.So we have these three major areas for computer applications in the years ahead.All will be growing in importance.But knowledge engineering, I think, will be growing in importance more rapidly than the other two, because it is the youngest and, in a sense, the most pervasive of the three.In saying that knowledge engineering is going to become very important, I don't want to imply that the other two will become less important.They will become more important also.But in relative terms, knowledge engineering will certainly be much more important than it is today.Now, what I'm going to say will relate to this third area, rather than the first two.Q. Do you see supercomputers playing an important role only in the area of numerical analysis?ZADEH.Supercomputers pertain to all three areas: numerical analysis, large databases, and knowledge engineering.But they apply primarily to the first area: numerical analysis.There is at this point some controversy as to how the available research funds should be distributed between the efforts to build supercomputers and to build machines that will be AI oriented.These are somewhat distinct efforts.The Japanese are pushing both of them.And in the United States, the emphasis on AI-oriented types of computers is just beginning to become strong, largely as a reaction to the Japanese effort.As you may know, Edward Feigenbaum of Stanford University is a leading advocate of the establishment of a U.S. National Center for Computer Technology as a rallying point for the U.S. effort.