Training for the Future

Raising his voice over the clicking of an IBM 7044 computer, Hosakere Mahabala said happily, “I worked for months on a kind of computer-robot being built at Massachusetts Institute of Technology for an artificial intelligence project, and I never got it to perform as reasonably as my two-year-old daughter!”

Mahabala, an associate professor of electrical engineering at the Indian Institute of Technology in Kanpur, takes his relationship with computers seriously but always with a light heart. “What we always have to remember,” he continued, “is that basically a computer is an idiot. Of course, this is an idiot that can do its work very rapidly—a million calculations in 10 seconds, in some cases—but still it can’t do any more than carry out the orders you give it.”

llT’s 7044/1401 system was carrying out its orders with a busy clatter in the Institute’s new Computer Centre, a modern structure of stark but strong architectural lines. The largest computer facility in any Indian school, the IIT Centre trains not only IIT students in computer use but short-term students from educational, research and industrial organizations. All 300 undergraduates in the third year of their B. Tech. program are taught the use of computers. Most of the 500 postgraduates use computers, and on a typical day the Centre runs the programs of about 200 students, chiefly in their project and research work.

A graduate program with an option in computer sciences is offered in the engineering department. Both programming techniques and computer design techniques are taught in this course. There are in addition courses in computer application.

The Centre also rents its computer facilities and services to research and development establishments, although 90 percent of the computer’s time is still reserved for faculty and student projects. So far over 100 outside organizations have used the Centre’s facilities. Their needs fall chiefly into three categories: engineering design of systems like turbines, electrical generators, bridges and buildings; operations research, such as the location of petroleum distribution points and forecasting demand; and tabulation and analysis of data collected in surveys, such as in city planning and market research.

Watching the 7044 complex at work is much like watching babies in a hospital nursery. The observer peers at the machines through windowed walls while the proud engineer at his side points out tapes, discs, memory banks, calculators, processors and print-out devices. Inside, on a reinforced floor (the weight of these mighty brain cells is considerable) stand several dozen cases and consoles containing hundreds of miles of wires strung in intricate patterns, each of which produces a certain electronic result. Tending this set-up is a corps of high-school graduate operators hired by the Institute. An aptitude test is given to applicants, and if selected, they are given on-the-job training.

The 7044 computer does not work alone but is fed basic data by its companion, an IBM 1401 input-output satellite. First, program data already punched onto cards by human programmers is fed to the 1401. Each card contains a statement about the problem to be solved or instructions leading to its solution. These are stated in numeric form, the computer being able to understand only mathematical terms. The 1401 records these statements on magnetic tape at the rate of 10 cards of information, or about 800 characters, per inch of tape. With a tape speed of 70 inches a second, it takes the 1401 less than a second to read and record 10 cards. The 1401 feeds to the 7044 the finished tape on which the problem and its data have been stated. The 7044 translates it into its own “machine” language and proceeds to calculate. Within seconds it produces the answer and writes it on to a tape for the 1401 to read and print out on paper.

The computer itself works much more rapidly than its auxiliary reader and printer, the 1401. For example, 2,400 lines can be calculated in a minute but only 600 lines can be printed in that same minute. Therefore all jobs must be preprogrammed to feed the 7044 continuously. To allow the 7044’s expensive electronic brain cells to stop functioning for even a few seconds is uneconomical. In fact, the computer itself—as though aware of the value of its time—immediately rejects a program and moves on to the next if it finds a programming error which prevents it from continuing its calculations.

Mahabala, with his unfailing zest for joking about his computer family, confided that despite the sophistication of its machines, the Centre is still at the mercy of the gods of electricity. “There were times when we’d have two and three failures a day, but in the last year the Institute has improved the distribution system on the campus so blackouts aren’t as frequent as before.

“The heat of a Kanpur summer is also a dangerous thing for computers,” continued the energetic engineer. “That, plus the fact that the machines themselves generate a lot of heat.” This necessitated installing probably one of the largest air-conditioning systems in India, built here with financial assistance from the U.S. Agency for International Development (USAID).

Across the corridor from the 7044’s area is a smaller windowed room containing the Institute’s first computer, an IBM 1620, which was obtained in 1963 under a USAID grant. One hundred times slower than the 7044 but still useful, it is used today to train IIT students and non-Institute personnel in computer programming methods. According to Mahabala, almost every 1620-computer center in India has IIT-trained people on its staff.

A computer’s most productive years are its first five, after which retirement looms. So as computers go, the 1620 is considered obsolete. Mahabala explained why. “The electronic circuitry in a computer doesn’t age but the mechanical parts do. And since computer manufacturers keep building new models, after a while you can’t get spare parts for an older machine. You can continue to use a computer like the 1620—especially in India where computers are not yet that common—but eventually an old machine isolates you so that you can’t communicate with other computerized firms or institutions. At IIT we’ve been using the 1620 mainly as a calculator but we want to convert it strictly to input-output activities in conjunction with the 7044. I’m also interested in getting it to display answers on a television screen and to write music and produce graphic designs.”

Down the hall in a remarkably uncluttered office sat the bead of IIT’s Computer Centre, 36-year-old Dr. V. Rajaraman. The young director’s list of scholastic credits is long and impressive: degrees of ascending levels from Delhi University, Bangalore’s Indian Institute of Science, America’s Massachusetts Institute of Technology and the University of Wisconsin; faculty positions at the universities of Wisconsin and California and IIT-Kanpur.

Would Dr. Rajaraman explain in basic terms how a computer functions? As if all in a day’s work, the professor gladly began explaining in layman’s terms. “Originally, computers were thought of as calculating machines, but later it was found that they could be made to do more than calculate. A computer can be made to solve complex problems provided the programmer can break down the problems into simple steps.

“As Dr. Mahabala likes to point out, computers are idiots. Laymen sometimes think of them as having human qualities but I suppose that’s an emotional response to a process they don’t understand completely. Take, for instance, the act of baking a cake. A recipe assumes a certain set of preliminary operations: when it calls for two cups of flour, it doesn’t also tell the baker to go to the cupboard, take down the flour, open the container and measure it into cups. Those steps are assumed by the baker because it’s common sense to make that assumption. Since a machine doesn’t have common sense, it doesn’t assume anything; if a machine were baking the cake it would have to be instructed to go to the cupboard for the flour, and so on. For the computer to function, a human must give it an exact recipe plus all the other factors which man would just automatically assume to be present in a given situation.

“So, after you decide what you want to tell the computer, you decide how you’ll communicate that message. The only thing a computer understands is numbers. But on the other hand, it must be able to understand various specialized vocabularies—medical, library science, business and so on. So each of these vocabularies has to be translated into a distinct mathematical language. Most computers can digest several languages by means of internal dictionaries; the machine translates the new language into its own numerical language, does its calculations that way, and translates its work and results back into the input language. For example, the 7044 came with a built-in vocabulary that uses just two units—zero and one. To the 7044 a set of 20 zeros means something. And in its ‘memory’ it can store about 32,000 words, or configurations using zeros and ones.

“Today there are hundreds of computer languages, all adaptations of a basic algebraic language by which the programmer presents a problem to the computer in the form of numbers. Something called SNOBOL is used by librarians. ALGOL is the international computer language used to express scientific and engineering problems.”

Dr. Rajaraman then explained the meaning of the computer terms “hardware” and “software.” When computers originally appeared, it was the machinery itself, the “hardware,” that was expensive. They didn’t contain today’s transistors and integrated circuitry which are compact and long-lasting. And, being relatively unsophisticated in the early days, the machines couldn’t handle complex programs—“software.”

“It would be quite possible,” commented Dr. Mahabala, “for India to develop a computer-software industry. Software orders from nations where programming costs are high could be filled here at much less cost and shipped back to the States or Europe and the Indian supplier could still make a very good profit.”

How “intelligent” are computers of the future going to be? “Well,” reflected Dr. Rajaraman, “the basic relationship—the machine’s dependence on man—will remain the same. But the machine has become much more rapid and precise than the men who program it; it can do a thousand computations in two-tenths of a second. To avoid waste the computer of the near future will be serviced by 20 programmers; it will absorb data from all 20 in a few seconds, work for four seconds on these 20 problems—.2 second on each—and keep repeating the cycle of ingestion and calculation until the answers are produced. Because input and response between each individual programmer and the computer are practically instantaneous, the programmer will not really be aware of any disruption in his relationship with the computer.

“So, certainly what we call third-generation computers are going to be able to handle more work more quickly than older models. But the computer, per se, has its limitations. It doesn’t have hunches, for example. It can only solve problems that can be broken up into a limited number of steps. That has been illustrated by a computer’s ability to win at checkers but not at chess, which has an infinite number of possibilities.

“’Engineers have built artificial intelligence simulators but in order to work successfully, the machine has to be supplied with methods of solution and also with solved examples or procedures; unless it has an example, it can’t choose the correct method from its stored supply of methods. When a computer was first taught to play checkers, it was programmed with the basic rules of the game. A programmer would play against it and at first the machine would always lose because it had no experience stored up. When it was beaten in a certain situation the programmer would give it negative marks for making the moves that beat it. Afterwards it would never repeat that combination of losing moves. After years of playing, it rejected all the losing combinations in given situations and so it always won.

“At any rate, humans shouldn’t worry about being replaced by computers. I think you can draw a comparison from the old joke about the General who sends the enlisted soldier to the post office to get him some stamps. The man doesn’t come back with the stamps, so the General goes to the post office to look for him. He’s standing there with the stamps in his hand. Naturally the General wants to know why he didn’t bring him the stamps, and the soldier replies that his orders didn’t mention anything about bringing the stamps back to him. Computers aren’t much different. They always follow your orders but don’t do any thinking on their own.”

Originally published in August 1970

---

---