Braille Monitor                                                                 February 1985


The Boundaries of Computerdom

by Kenneth Jernigan

There was a time when the National Federation of the Blind was a much smaller and less complicated organization than it now is. The goal was simple: Help the average blind person get enough to eat, strengthen the organization, do a little public educating, and fight the agencies. Much of that original objective is still valid, but the days of simplicity are decades behind us.

All of this was recently brought home to me in a very interesting and unmistakable manner. As the work of the Federation has increased, so have the number and size of our Committees. We deal with everything from parents to workshops, from lawyers to music, and from candy sales to computers. We are not just an organization established to advance the interests of the blind, nor (important as it is) are we even just a vehicle for the formulation of policy and expression of concerted action. We are a broad-based people's movement, touching every aspect of the daily lives of the blind.

One of our newer ventures is the Committee on Research and Development. As would be expected, it spends a great deal of time dealing with computers. Not long ago I asked Dr. T. V. Cranmer, who chairs the committee, whether it was possible (given the current state of technology) for a computer to translate from one language to another. I took the matter lightly and soon forgot it, but apparently he did not.

Tim Cranmer (as anyone will tell you) takes his computers seriously. A challenge is a challenge--and a question is a question. It deserves to be researched and answered. So he did--and it was. Thus, the following letter:

Frankfort, Kentucky
November 25, 1984

Dear Dr. Jernigan:

Alas, we must await the successful outcome of man's fevered efforts to create artificial intelligence before we can enjoy fast accurate machine translations of the natural languages. Your casual question as to the availability of software to translate English into Spanish or other language sparked my interest in finding out.

Recent papers in Scientific American convincingly argue that the process is beyond the present capabilities of our biggest computers. I am convinced that the task is a thousand magnitudes more difficult than, say, the conversion of straight text into grade two. For one thing, natural languages have evolved rather than having been planned. The result is that current usage and context plays a primary role in determining meaning. If the computer can deduce meaning from a phrase or sentence, it must then compose a sentence in the target language that expresses the same thought replete with subtle cultural meaning.

Since it is not my purpose to write a paper on the subject, I will be content to cite a few examples that illustrate just a tiny bit of the problem.

Go to the bank. Is this financial advice, or instructions to a fisherman? The chickens are ready to eat. Are the chickens food for a feast, or are the chickens hungry and waiting to be fed? He saw that gasolene can explode. Is this sudden insight to a property of gasolene by a chemistry student, or a declaration of a witnessed event? Here the word "that" and the phrase "gasolene can" have totally different sentence construction roles. There's a man in the room with a green hat on. We know that rooms don't wear hats. Computers don't know that.

Governments and industries throughout the developed nations are advancing huge sums of money to fund elaborate research aimed at creating artificial intelligence. Japan has the largest government-supported effort. West Germany, the Scandinavian countries (and somewhere in the race, American industry) are working to make what has come to be known as the fifth generation of computers. Most observers doubt that these machine giants will be realized in this decade. They may be available to the military by the mid-nineties. A similar computing power may be available to business and industry in the first decade of the twenty-first century. We'll have to wait until sometime after that to dictate our letters to speech input typewriters, instruct our personal robot to mow the lawn, wash the windows, or go to the handy market and fetch a liter of milk. Then, too, we will be able to speak to a machine and have it transcribe or reutter our thoughts in the foreign language of our choice.