SEAC and the Start of Image Processing at the National Bureau of Standards

Testing the Computer

The Maintenance Crew was prepared to claim several years of fault-free operation of the SEAC when they moved the computer to a different part of the NBS campus. This required disassembling the hardware and reconnecting it according to the design diagrams (and the excellent memory of some of the staff). Upon reconnecting the computer, it was discovered there was a wiring error in the original design which had survived about a decade of scrupulous diagnostic checking and would result in a definitive error every time that a suitable program was run. It had never occurred in a decade of round-the-clock operation! From this, one may conclude, throughout history, there probably has never been a faultless computer built, since all computers have always had known malfunctions with associated "workarounds" in the software, if not in the hardware.

Such assiduousness in tracking down malfunctions is seldom practiced, if ever, today. But in the process of testing the computer after a suspicious fault had been detected, the crew would confirm the correct operation by running some of its own research programs. It turns out that many people in the early days of computing had this same practice. For example, Arthur Samuel at IBM, who wrote the first artificial intelligence program for checker playing, was the manager for the Poughkeepsie plant. He had masses of IBM 701 computers playing checkers every night supposedly in order to check out the computers, but of course, really to accumulate learning experience on his program. This also was the source of similar computer time that Kirsch used to accumulate his own artificial intelligence research results (6).

These modes of operation permitted testing ideas for which there were not necessarily any stated applications. And the great size of the "government market" enabled any innovative computer use to be successfully placed somewhere in a government agency.

Among these innovative uses of computers there was the searching of chemical structures (5), artificial intelligence (6), information retrieval (7), language processing (8), and many others in SEAC's 14 years of productive computation. We discuss, below, the four decades of one of these applications, image processing, as it developed at the NBS.

<< previous next >>