Book Review : Computing Your Artificial IQ Scores

The Sachertorte Algorithm and Other Antidotes to Computer Anxiety by John Shore (Viking, $16.95)

John Shore wrote this book about computers on a computer. The review you are reading was not, as it happens, written on a computer but it was edited on one. All of which tends to support Shore’s skepticism that computers are or ever will be anything like brains. They no more think, he believes, than they love.

Some readers will object. The notion of an artificial intelligence is geared to not a few scientists and hackers. The point is that Shore’s computer did not object, and neither--assuming this makes it into print--did my editor’s.

Shore is a computer scientist whose passion for his subject has been given an odd and useful spin by his passion for George Orwell and Strunk and White’s “The Elements of Style.” His book attempts to reconcile computer literacy with literacy and it is only natural that the product of this quixotic effort should bear such a title as “The Sachertorte Algorithm.”

Humane and Sensible Book

That may sound fey, and three or four times in the course of this humane and sensible book there is a labored joke or a faintly tired bit or byte of sprightliness. But Shore’s Aunt Martl’s sachertorte does, in fact, provide a highly lucid illustration for the benefit of the intended layman reader.

The recipe reached the author in the form of a list of ingredients--butter, chocolate, three egg whites and so on--and an instruction that reads, in its entirety: “Bake slowly in not too hot oven, cake should be moist.”

Aunt Martl stands at one extreme of knowledge, you see. The computer stands at the other. In between stands the algorithm--the precise description of the procedure for solving a problem. And Shore proceeds to demonstrate the various ways in which this algorithm can be framed so that a computer could receive it. In the course of it, we computer illiterates get more than an inkling of just how such computer languages as Pascal (which Shore admires for its conceptual elegance) and Fortran (which he considers a nitwitted boor) really work.

Enormous Problems

Then, for our further instruction, he makes clear that even a chocolate cake recipe would, in practice, present enormous problems for computer programming. One programmer, he reports, spent 12 hours writing up Julia Child’s beef Wellington and then gave up, having produced 60 pages of program covering only three of the recipe’s 16 pages.

Shore is concerned throughout the book to moderate both exaggerated expectations and exaggerated fears of what computers are and can do. He begins with a phenomenon he calls “computer anxiety.” His physics professor had it and assigned Shore as his assistant, to study up on the subject so he could work some things out for him. There can be more instruction in a teacher’s ignorance than in his instruction. For Shore, there was a vocation.

He is good on computer anxiety, his secretary having fled in tears when he placed one on her desk and moved her typewriter to a side table. It taught him tact and an attractive skepticism. He is critical of the instant authority that the computer can wield. Scientists are now deluged by the equivalent of first drafts of papers produced cheaply and easily on the computer. Too cheaply and easily; formerly, a paper would be worked, re-worked, appraised and revised before anyone would go to the expense of printing it.

Computer Language

He notes the peremptory, even catastrophic character of computer’s language. “Illegal,” “Fatal Error”--such phrases cause the novice all kinds of guilt when the fault is frequently in the program. Computers being relentlessly small-minded, they will reject perfectly reasonable instructions if the program is not written with a decent regard for the user. Shore recalls following his father’s detailed driving instructions to a wedding, going briefly astray over an ambiguous direction, and picking up the trail by a little common sense reconnoitering. If he were a computer, he writes, he would have telephoned his father, called out “Job Aborted--Fatal Redundancy Check,” and hung up.

It is a promising frame of mind for leading non-initiates a considerable way into what computers can and cannot do, and into the current advances and problems in the field. Some of the chapters threaten to become technical, but none require more than a reasonable concentration. Shore’s fault, if any, is to let his discursiveness occasionally outrun his concrete examples. Generally, though, he holds the reader by writing well and by demonstrating an agreeably contradictory quality of being open minded and opinionated at the same time.

He writes in some detail about what he calls the software crisis. Computer programs, he reports, are widely susceptible to errors. So much so, in fact, that what in the trade is called “maintenance” in fact means that the manufacturer undertakes to conduct periodic corrections of the errors. By car manufacturing standards, “most software programs are lemons.”

Some Serious Errors

It is in the big programs that errors are most frequent and serious. Shore cites a foul up in computing a Quebec election, and the failure of a back-up computer in the space shuttle program. “The complexity of the computer programs we write has grown faster than our ability to write them correctly,” he argues.

In a section devoted to the prospects for artificial intelligence, he makes useful distinctions between what computers can do and what the brain does. Computers count much better than the brain; but they are not good at natural language, associating and finding patterns. And unless we discover just how these rich and mysterious processes work, there is no way that computers can learn them. Artificial intelligence work, he writes, quoting a fellow expert, “concerns the automation of things we know how to do, not the automation of things we would like to know how to do.”

The computer, in short, is a tool, “important not because it allows us to do less but because it helps us to do more,” he concludes. It is a marvel, at that, and Shore has a marvelous phrase for it. He quotes Santayana’s “To turn events into ideas is the function of literature” and adds that the function of computer programming is “to turn ideas into events.”