Computers And Society

David D. Thornburq, Associate Editor

The Fifth Generation

I can hardly resist the temptation to point out that Orwell's vision for 1984 is (thankfully) not going to come true. It is interesting to note that, as with many other futurists, Orwell overestimated the amount of social change that would occur by 1984, and seriously underestimated the amount of technological innovation that will have been developed by then. While it is true that office workers in Orwell's novel dictate their letters into a "speak write," an automated stenographer/printer, much of the remaining technology is neither advanced nor inspiring.

I was reminded of the impact of technological advances as I created the first draft of this month's column on my Brother EP-20 battery-operated electronic typewriter. This marvel of design is quite compact, fits on an airplane tray table, and is almost silent. Since it retails for about $200 and allows the user to correct up to 16 characters of text before it is printed, I would not be surprised to see this device open up whole new markets for typewriters. I never used a typewriter for rough drafts before, simply because they were too bulky. Now, this device has become my portable work-station (sadly missing the storage that would make it a terminal for my word processor), and I take it everywhere.

Is it significant that this innovation was developed by a Japanese company? As we look at the computer industry, it is clear that it is taking on a decidedly international flavor. And yet, so far, the big names in personal computers are definitely American (TI, Commodore, Atari, Apple, IBM, etc.).

KIPS Super Computer

A recently published book, The Fifth Generation (Addison-Wesley, $15.95), suggests that we must be much more aware of Japanese advances in computer technology if we are to survive as a technological nation. Far from being a "scare" book designed to erect protectionist trade barriers, The Fifth Generation is more a call to arms. Its authors are Edward Feigenbaum, a pioneer in the field of artificial intelligence, and Pamela McCorduck, a science writer who has written extensively on computers and intelligent behavior in machines. The authors say that Japan has embarked on a ten-year crash program to develop a new type of super computer – a "fifth generation" machine that is called a Knowledge Information Processing System (KIPS). The KIPS is expected to be markedly different in architecture from the computers in use today. Furthermore, it is expected that users of the KIPS will interact with it very differently from the way people use computers today.

What is a KIPS? While most of today's computers are used for data processing and, with the exception of languages like LISP and Logo, most computer languages are geared towards data processing tasks, the KIPS is an optimized blend of hardware and software, tailored to perform general symbol manipulation and symbolic inference. This shift in emphasis recognizes that most of our work is nonmathematical in nature. Much of our work involves reasoning, not calculating.

A Reasoning Machine?

Can one build a "reasoning" machine? According to Feigenbaum and McCorduck, the Japanese lack our preoccupation with this question. From their perspective, it is sufficient to note that computer systems powerful enough to be fifth generation machines will function at a level far beyond that with which we are presently familiar.

Modest projects in the development of systems that outperform human "experts" are an important result of research in artificial intelligence. For example, programs that perform certain types of medical diagnoses, analyze and propose synthetic pathways in the creation of new chemical compounds, and predict the location of geological deposits have already been implemented on existing commercial computers using languages such as LISP. Such programs must operate with both a "knowledge base" and a set of "inference procedures." To read a map, for instance, one must have both maps to read and a procedure for reading them.

Intuitive Solutions

The fifth generation KIPS will be built around the collection of vast amounts of data and the collection of problem-solving techniques that range from rigid deterministic methods to those that mimic the human ability to act on "hunches." You need not become embroiled in the machine intelligence controversy to appreciate that such systems have the potential to completely redefine computers, their use, and their place in society.

In order to create the KIPS, advances are required in both computer hardware and software. The computers we are familiar with operate in serial fashion. Instructions are executed one at a time. This type of computer architecture was developed by John von Neumann, and speed limitations in such computer systems are caused by the "von Neumann bottleneck" – processing instruction by instruction, byte by byte. In order to create faster computers, the fifth generation machines may favor a system using many processors in parallel.

A Billion Inferences Per Second

To appreciate the need for this approach, you should remember that the KIPS is to be used primarily for the linking of a knowledge base by symbolic representations (e.g., a sparrow is a kind of bird), or for the representation of rules (e.g., if the temperature is over 400 degrees, then the boiler must be turned down). To be used effectively, a problem-solving program must scan its library of "IFs" to find one relevant to the problem at hand. Finding this needle in the knowledge-based hay-stack of the size anticipated by the Japanese will require much more computational horsepower than we have seen to date. For example, today's big computers are capable of executing no more than 100,000 logical inferences per second (LIPS). (One logical inference corresponds to one IF/THEN statement.) A personal computer such as an Apple II might execute (depending on the language chosen) about 100 LIPS. The KIPS will be designed to execute up to a billion LIPS.

Such achievements are not the result of hardware alone. Interestingly, the language of present interest to the KIPS project leaders has already been developed by the Europeans – PROLOG.

How feasible is this project? There is much diversity of opinion on this topic, but there is consensus that, even if the project goals are not met in the allotted ten years, the interim results will most certainly change the nature of computers and computing. As Feigenbaum and McCorduck say:

Word literacy has given us power, access to an opulent, soaring world of mind – an alteration of thought processes – that is denied the illiterate. Computing literacy, even in its present form, opens still another world, one that all eventually may enter as routinely as they enter the world of letters, and it will confer perhaps even more power than the mighty pen and press have already given us. This is not idle promotion. As human muscle-power has been amplified by many special-purpose machines, so human mind-power will be amplified. The computer will change not only what we think, but how.