In the glass-enclosed machine room of the Harvard Computing Center, a dozen men in shirtsleeves dash about--consulting a stream of paper tumbling out of one machine or filing a stack of cards into the entrails of another. The banks of the IBM 7094 churn through hours of work in minutes, lights flashing and tapes spinning into a blur, jerking to a stop, and spinning the other way.
Norman Zachary, director of the Center, and Anthony G. Oettinger '51, professor of Linguistics, are preparing a report which will indicate the need for a large-scale expansion of computer facilities--Zachary predicts that by 1971 the University will be spending $15 million a year on computers, more than ten times this year's budget. Most of the expansion will be outside the walls of the Center itself--in graduate schools, Houses, and laboratories. At present, about 95 per cent of Harvard's computer equipment is housed within the Center.
In the planned expansion, the Center will get a second 7094 within the next few weeks. By 1971, Zachary hopes to have a 360/91--a seven million dollar contrivance which is so new that it will not be delivered anywhere for two or three years.
The budget Zachary projects is largely independent of Harvard money. Between 70 and 80 per cent of the Center's funds comes from federal grants: NSF, NASA, AEC, National institute of Health, and the Defense Department. When students or professors use the computers for projects not sponsored by Federal grants, their department pays the bill. This accounts for most of the rest of the Center's income. The University as a whole ordinarily gives no money, al this year Harvard granted for some specific projects.
The Center has been growing since 1964, when it switched passive" to "active" service. When it was established in 1962, the original idea was to offer a computer to anyone in the University and say "use it," Zachary recalls.
When Zachary became director in 1964, the Center began to provide more than just computer time and minimal technical assistance. "Now," explains, "if someone asks for help, we're building up a staff of programmers who can give it to him.". This means that non-scientists can make use of the Center without knowing computer language.
Even on the present budget, the center has developed new uses for computers. The machines analyze the structure of English sentences, they solve intricate differential equations, they can flash mathematical curves on a screen at the touch of a button. All one has to do to make them work for him is learn their argot. Automation of the Widener files is one new use that will be possible with the expanded facilities.
Another new project will be the scanning of chromosomes in the study of congenital birth defects and hereditary diseases. The computer will examine hundreds of thousands of chromosome configurations in a few hours, separating out the abnormal ones. At this point, the scientists will take over, but they will have been saved thousands of bleary hours over the microscope.
It has become quite fashionable to explain that computers are not actually very intelligent. That's true. But--as the chromosome project shows--they are fast, and willing to do the dirty work. One current Center project which utilizes computer serfdom fully is a linguistics study, in which Oettinger has done much of the fundamental research.
In diagramming a sentence, the machine must decide about meaning by using context. Often this is difficult or impossible for the computer to do--which only means that we do it by intuition, but cannot transfer our intuition to the machine.
Useful Dictionary
Programmers therefore give the machine a dictionary of words that can be more than one part of speech. For all such words, the computer flashes the choices on a screen. A trained observer points to the correct choice in that context with a special "light pen" and the computer eradicates all other choices.
Laziness breeds genius, so linguists have developed a way to avoid some of the work in this process. When the same word appears again later on in the same work, the machine assumes that the same meaning is most likely to be desired; so it projects this choice more brightly than all the others, and will automatically use that meaning unless the observer indicates otherwise with his pen.
Take the sentence: "Time flies like an arrow." Instead of having the machine say, "time: subject, verb, adjective," and having the observer choose "subject" for this particular context, why can't the machine be instructed to "figure it out?" "Time flies like an arrow" is not really very different from "Fruit flies like a banana," but their diagrams are at opposite poles. In the latter, "fruit flies" are a species of fly and "like" is a verb. Why shouldn't the machine say that "time flies" are another (admittedly rarer) species?
When the machine consults this dictionary and flashes the choices, it is doing what it can do best: looking things up in a reference and displaying them. And the observer is doing what man does best: using his intuition.
