What Can You Do with a Broken Computer?
Russell A. Kirsch
Director of Research, Sturvil Corporation
President Lettieri called the 2079th meeting to order at 8:17 p.m. on October 17, 1997. The Recording Secretary read the minutes of the 2078th meeting and they were approved.
The speaker for the 2077th meeting was Russell A. Kirsch, Director of Research, Sturvil Corporation. The title of his talk was, “What Can You Do With A Broken Computer?”
Mr. Kirsch stated that we usually assume that computers behave according to their design specifications. If not, we consider them to be broken and try to fix them. But there are deep issues here that have consequences for our understanding of the behavior of computers. Some of these consequences effect our understanding of science and the arts, and the trust that we put in scientific investigations. Our computers are already senescent as tools and rigor mortis has set in. Consider the Y2K problem. To save space, early programmers stored only they last two digits of the year. Consequently, computers conclude that year 1999 is followed by 1900, not 2000. These early programmers failed to understand the consequences of their decision. Worse, these original programmers can't remember how they wrote the code and consequently can't fix it.
Mr. Kirsch encountered computers in the early 1950's with the SEAC computer. This was more than a passing relationship. Mr. Kirsch knew the complete wiring of the entire SEAC computer and in the 1960's he had to move SEAC to make way for “progress.” Fearing SEAC would never run again after it was taken apart and hauled across the campus, Mr. Kirsch personally disassembled SEAC and lovingly carried the components to their new home. However, to his ultimate horror during the disassembly work, he discovered SEAC was wired up wrong — and had been for more than ten years. Therein lies the beginning of the tale.
Out of this SEAC experience, Mr. Kirsch concluded that all computers are broken — all the time. If you hesitate in accepting this, consider the license disclaimers that accompany all computers and software compared to a guarantee for your new car. The software disclaimer is equivalent to the automaker saying, “We don't guarantee your car will run. You must figure it out yourself. Further some things (left unidentified) will make it not run.” Is there an alternative to this unsettling situation? Can we have reliable computers and what does that question really mean? The question - and the answer have a practical aspect and a theoretical aspect.
On the practical side the answer is “yes we can have reliable computers.” Mr. Shannon showed how relay type components can be made reliable and Mr. Von Neumann did the same for neuron-like elements.
The theoretical requires an answer to the next question, “What is a computer?” This implies models of computers with relays and neurons. One model is the finite state machine exhibiting dull, repetitive behavior. A second model is the Turing machine reflecting most programmers' perception that they can program anything they can understand.
The models reveal a deep asymmetry between the computer's hardware design and its operation and between the software code and its behavior. Briefly it revealed the asymmetry between the computer structures and functions.
It's easy to understand the design structure issues — the hardware design and the software code. This resolves the practical question of computer reliability. It's not so easy to resolve the function issues - computer operation and behavior. These are at the root of the long-winded disclaimers accompanying all software and hardware. It's also at the root of misleading computer demos and coding-creativity problems, where occasionally predatory programs are nefariously created which are extremely difficult to analyze and clean out. Thus our computers will reliably give you the right answer — as long as you know what the answer is beforehand. However, when you are relying on the computer to produce the right answer — when you haven't got a clue, the reliability issues come to the forefront.
Although we all know many startling new uses of computers, Mr. Kirsch described more for us where the reliability issue of new results will rear its ugly head. These include nonmathematical applications in linguistics, archeology, architecture, fine arts, and metallurgy. Mr. Kirsch then slipped up and described a complex computer analogy to a checkerboard. After cordially and knowledgeably responding to a multiplicity of questions about the analogy, Mr. Kirsch promised the Society he would never ever again mention a checkerboard in his talk.
Mr. Kirsch then closed his presentation and kindly responded to questions from the audience. President Lettieri thanked Mr. Kirsch for the society and announced the next meeting. He then made the usual parking announcement and adjourned the 2079th meeting to the social hour at 9:40 p.m.