Inventing Fire by Frank Short

YESTERDAY. TODAY AND TOMORROW

Keynote speech -April 15, tadalafil 1991 -Imperial Hotel -Tokyo, ailment Japan

by Dr. J. Presper Eckert

Copyright © 1991 J. Presper Eckert, Jr.

10th anniversary of ERIC (Eckert Research International Corporation)

I am very happy to be back in Tokyo again and to be able to hear in person some of the things you in Japan are thinking about in computers. I am happy to talk to you about my favorite subject, computers. It has been several years since I have talked at any length to a large group about computers.

At a recent party, at my daughter’s house, I counted that over half of the people present used computers in their work and in many cases were dependant on them for their livelihood. Only a few of them were friends of mine, most of them were friends of my daughter and her husband and were of their generation.

The point I am trying to make is that computers are now approaching the universal nature of the printing press, which in many cases it has replaced. It is more universal than the book, the magazine and the newspaper because it now makes up a major part of our communication systems as well. In addition it can assist in retrieving and locating the information we need. Books are often much slower in this area.

The computer now forms a part of our automobiles, our microwave ovens, our robots and other forms of factory automation. To me, one of the most exciting uses of computers is in medicine. Computers have greatly improved the ability of physicians to diagnose the human body. We now have Cat Scan, Magnetic resonance, and Ultrasonic imaging devices.

Both John Mauchly, my partner, and I were frequently asked if, in the early days we could visualize all that was happening and would happen in computers. The answer is yes and no, we thought that much could be done both in science, engineering and business with the digital approach, because of its accuracy, flexibility and resistance to deterioration of the information involved. However, no one expected that the invention of the very large scale integrated silicon chip would occur as soon as it did and that it would prove to be as reliable and as inexpensive as it has turned out to be. But we always believed that the digital approach would wipe out the analogue approach, not only in most existing areas, but in many new areas as well. We thought it would aid many, if not all, areas of life. We thought all this would take time, considerably more time that it appears to be taking.

Answers to common questions

Two young men sent me a series of questions which were clearer and perhaps more to the point than those asked by many professional journalists. In any case their questions lend themselves to fairly flexible answers to some of the important things that happened to Dr. John Mauchly and myself in the pursuit of high speed electronic digital computers.

Their questions were:

QUESTION: 1) What gave you the idea to develop a computer?

ANSWER: Several things converged to give Mauchly and I the idea of building a computer. We were both working at the University of Pennsylvania during World War II. The army was using two of the three Differential Analyzers in the world at that time. The first one was developed by Dr. Vanavar Bush at M.I.T. and was being used by the Navy. All three were being used to do various ballistic calculations. The two machine the Army were using were improved copies of the M.I.T. Bush machine and were built in the 1930’s to provide employment, in the depression, to engineers and designers. They were built at government expense. They were rather slow and not sufficiently accurate for most ballistic work. The one percent (1%) accuracy achieved was however suitable for many engineering problems at the University before World War II.

Working with Dr. Cornelius Weygandt and other university employees we managed, by the addition of about 400 vacuum tubes and a considerable number of special motors and generators, along with photo-cell devices and light polarizing devices, to speed up the Differential Analyzer by about ten times and at the same time to achieve about ten times the accuracy (0.1%) “Annie”, as we nicknamed the machine, was in spite of the added electronics, still an analogue device and was not only severely limited in speed and accuracy but was only useful on certain kinds of differential and integral equations. Not only did the research on government weapons require more speed, accuracy and flexibility than this, but many scientific and engineering problems also required more of these abilities.

John Mauchly and I discussed making electronic integrators using electronic pulse counters. I had worked at the University of Pennsylvania on some digital counters for a special high accuracy Radar project and on a Mercury Delay Line timing device (which I invented). We saw how some of these ideas could be applied to an electronic integrator. The mechanical integrator was the weakest part of the differential analyzer. We decided that while the electronic integrator would improve the Differential Analyzer it would still be severely limited in speed by the remaining gears and shafts that interconnected the system as well as by the electromechanical servomechanisms that now drove them.

We rather quickly concluded that any sensible improvement in computing equipment would not only have to be all electronic, but would have to encode the transmitted information in a far more efficient manner than simple strings of pulses, such as a pulse integrating circuit might employ. We thus decided that our first approach should be all digital, using codes that would allow the use of existing punch card equipment for input output purposes.

QUESTION: 2) How many attempts did you make before the final project was achieved?

ANSWER: We really made only one carefully organized attempt to build the original ENIAC. We made no false starts in designing equipment. We refined our original ideas, as outlined above, in the early discussion stages. We studied over half a dozen counter circuits. These circuits were obtained largely from the work of physicists who used them for counters in radiation and cosmic ray studies. We found the physicists’ designs were generally too slow and too unreliable. We then worked out a fairly fast and quite reliable design of our own based in part on some circuitry that had been developed for timing purpose in measuring projectile motion in ballistic tests.

We went to the RCA tube research laboratories and obtained clues as to how to operate vacuum tubes to achieve long life and trouble free operation. We conducted tests on thousands of tubes and set up standards to design circuits that would allow proper operation in spite of large variations in the tubes, resistors and capacitors, that comprised the circuits. We also designed circuits and power supplies that were quite insensitive to power line voltage variations. In addition we used a voltage regulation system on the power line to gain added reliability.

We made a few “bread board” models of some of the key circuits and built a “prototype” model of two accumulators (which were really electronic adding machines). They were of the design which we would use in the final machine. With a few changes and improvements these prototypes worked fine. All this took about one year from when we started. We then built twenty accumulators as well as the rest of the machine.

In the next year and a half we built and tested thirty-seven electronic panels, each two feet wide and eight feet high. Each contained about 500 vacuum tubes (or radio tubes). We also built three relay panels, using very reliable telephone relays for buffering the data to and from the punch card machines. These buffers allowed us to read cards, punch cards and compute; all at the same time. These three panels were the same size as the electronic panels and contained about 1500 relays or about 500 relays per panel. In addition we designed and built a very extensive centralized six panel power supply system. These panels were larger and deeper than the forty panels, which made up the machine proper. A single special capacitor panel was placed in an adjoining room to keep these power supply filter capacitors as cool as possible.

Additional punch card machines, for creating input keypunch data, sorting data, printing data, etc. were put in a hallway leading to the main thirty by fifty foot room in which the ENIAC was housed.

QUESTION: 3) How long did it take to achieve your goal?

ANSWER: It took two and one-half years to design and build the machine from when we started. But it took months after this to work out small problems as they showed up in various test problems, and to start to understand better how to program the machine. Three years is probably the best answer as to how long it took. Various improvement and additions were made to the machine for some years to follow. The machine remained in use at the Ballistic Research Laboratory at Aberdeen, Maryland for about 10 years.

QUESTION: 4) How did the public respond to the development of the ENIAC?

ANSWER: The news media kept referring to the machine as a “giant brain”, rather than as an automatic machine for performing arithmetic on many numbers in accord with a predetermined sequence and/or a set of rules for manipulating the numbers. As a result an overly complex and misleading picture was formed in the public mind. Many people were however not confused by all this and interest grew quickly. We found little foresight in the minds of banking and financial circles and did not received help from the people who traditionally claim to encourage the future.

We worked on a number of government projects to get money to keep going. Some of this work probably held us back more than it helped us.

Many business men thought we would not be able to do what we said we could do. It was very tough to get started.

QUESTION: 5) How did your lifestyle change after the ENIAC was invented?

ANSWER: My life style was not really affected very much by the computer, except that it took all my time. But anything I got interested in would probably have done the same. My father made a fairly good income, I went to a private school (William Penn Charter School -now over 300 years old -it was started by William Penn and is I believe the oldest private boys school in the U.S.). I went to an Ivy League college, the University of Pennsylvania.

My father and mother traveled a lot. I was an only child. They always took me with them. By the time I was 12 years old I had traveled about 125,000 miles and had been in every state (48 then) and Alaska, most of Europe and part of Asia. We cruised almost to the North Pole. We had traveled by ships, cars, trains, horses, mules and even camels (no airplanes those days) . In the years that followed the ENIAC I frequently traveled 125,000 miles in a single year, mostly by airplane.

I met president Harding when I was 5 years old, many movie stars before I was 8, and was given a medal for inventing the computer by President Johnson in 1968 at age 49. My life has been exciting both before and after John Mauchly and I invented the computer.

QUESTION: 6) Although the ENIAC is perceived by most as the first computer, isn’t the ABC computer at Iowa State the first computer?

ANSWER: The work by Dr. Atanasoff in Iowa was, in my opinion, a joke. He never really got anything to work. He had no programming system. He tried for a patent and was told the work he has done was too incomplete to get a patent. A competitor in a patent suit convinced, what in my mind was a very confused judge, to believe Atanasoff ‘s story, even though it had no real relation to the case at hand. The judge wrote a confused and irrelevant opinion about the invention of the computer. We had convinced judges in cases with LB.M and A.T.& T. that we had invented the computer. The lawsuits got too expensive; the last one cost over one million dollars. We could have gone on to appeal but the patent would have run out before the matter would have ever been settled. We simply quit wasting time on a patent system that could not really understand what we had done at the time we did it. The U.S. government supplied patent attorneys were so inept at their work, in my opinion, that the work was delayed too long and not carried out at all well. This resulted in great confusion later.

Who invented the first of anything is always confused in history.  To make the point let me quote from the fourteenth edition of The Encyclopedia Britannica, 1929. Volume 14, Page 105.

“The earliest attempt at making an incandescent lamp was made by Dc la Rue in 1820. In 1840 Grove demonstrated his battery by lighting an auditorium with incandescent electric light.”

The encyclopedia goes on and mentions a British patent to Frederick De Moleyns issued in 1841. J. W. Starr patented a carbon filament lamp in 1845. Sir Joseph W. Swan manufactured incandescent lamps from 1848 to 1860. All in all about 100 people made various incandescent lamps before Thomas A Edison developed his world famous lamp.

It is clear that Edison did not invent the incandescent lamp. He lit up Menlo Park in December 1879 about 59 years after De la Rue started it all in 1820 and 34 years after the carbon filament idea he used was invented.

What then did Edison really do?

1)         He made a lamp with a long skinny carbon filament (also invented in England at about the same time by Swan)

2)         He used a completely glass sealed vacuum chamber.

3)         He used a rather high vacuum. Made possible by the vacuum pump work of Herman Sprengel in 1865 and the improved vacuum techniques of Sir William Crookes developed in 1875.

4)         He sealed the wires into the lamp with platinum wires which matched the temperature coefficient of the glass the lamp was made of to avoid the vacuum loss of earlier designs

He thus combined four known ideas into one and made a lamp which consumed 7.5 Watts, had a life of about 40 burning hours, and worked on a high voltage (110 volts). It used a very low current, so low that many bulbs could work in parallel on reasonable size (and cost) copper feed wire from a generator with reasonable size commutators. Earlier lamps used low voltages and high currents, frequently from batteries. They lasted perhaps two (2) hours before a replacement of the filament was required.

The earliest practical generator for lighting was developed by Gramme as a so-called dynamo in 1875 only four years before Edison’s lighting demonstration at Menlo Park.

Edison also needed a wattmeter if he was to sell a practical lighting system to people. He invented a chemical wattmeter in the same year as his lamp (1879) to solve this problem. A bit later a high school teacher at Central High School in Philadelphia, Professor Elihu Thompson invented a much better electromechanical wattmeter without chemicals and it properly completed Edison’s system. Thompson and his school teacher partner, Houston, had formed a company to make generators, wattmeters, etc. Thompson is also considered the inventor of electric welding. Thompson and Houston’s company later merged with Edison to form “General Electric”.

My father, when a teenager was taught to swim by Houston, who was much older and who lived as a bachelor in a large house in West Philadelphia.

The point of all this is that Edison had a complete lighting system and a practical incandescent lamp. Others had not achieved this. However the basic idea was quite old at the time of his invention.

Mauchly and I achieved a complete workable computing system. Others had not. In fact others had not even achieved the essential functioning parts of an electronic computing system.

If Edison is the inventor of the incandescent lamp it would appear that by the same yardstick Mauchly and I are clearly the inventors of the computer.

Frankly, I find the claims of Dr. Atanasoff and the Iowa people to be absurd and inaccurate to say the least; furthermore it took them years to come up with their claims well after the computer was a fully accepted idea.

One could carry the discussion of what constitutes an invention on with a lot of additional examples. The Wright brothers’ invention of the airplane is another good example. Here again the Wright brothers developed a complete and working system. The idea of wings and a propeller driven by some kind of engine was already old when they started their work, they were late comers, just as Edison was with his light bulb. Dr. Langley had built a model airplane driven by a steam engine about ten years before the Wright brothers had their success. It was in its way successful; it made a flight of about 4000 feet (over 1000 meters).

But the Wright brothers were not satisfied, they felt that with their bicycle making ability and some hard work that they could make a machine to carry a man and fly it much further than others had been able to. They studied the design ideas of Dr. Laungley and were not satisfied with his calculations. To learn more before they made a final design they spent two years building and experimenting with various wing shapes in their experimental wind tunnel starting in 1901. They found a lot wrong with the approach of Dr. Langley and others and they developed their own approach.

They went further than anyone else had and concluded that airplanes were not inherently stable and needed more control from the pilot to properly maneuver them. They need more than just a rudder. They fastened ropes to the wings of their early planes so they could bend the wings to achieve the desired flying stability and avoid a crash. They had thus invented the first form of the “aileron” control, an essential ingredient to a practical and complete flying system. still it took them two more years before they found out how to avoid a “tail-spin” in making a turn.

Without going on and on about inventors it can be seen that inventions of importance are not just a “flash of genius”, as many think, they are the incorporation of such flashes into a complete working system. Before the Wright brothers and the aileron control pilots had attempted to stabilize flight by throwing their body weight around to control the planes stability. Can you imagine doing this in a modern airplane?

The contributions of Edison can still be recognized in the modern light bulb. So can the aileron control of the Wright brothers still be recognized in a modern airplane?

Can the contributions of Eckert and Mauchly to the original ENIAC and later the Univac still be recognized in the modern computer? I hope and believe they still can be.

Of course few in the world will ever compete with the mathematician Euclid whose historic work on geometry remained nearly unchanged for over 2000 years. It is only somewhat over one hundred years ago that alternate, non-Euclidian, geometries were developed, even then they have not really displaced Euclid’s work.

ENIAC Firsts

I have now attempted to draw up a list of what I think are the important “firsts” that came about through the development of the ENIAC.

1) I think the most important item in the ENIAC was the control of the subroutines in programming. This idea was first proposed to me by Mauchly, and it became immediately clear that it was absolutely essential to the design and construction of the ENIAC. While neither Mauchly nor I knew it at the time, an electromechanical machine, constructed of I.B.M. tabulator parts, and called the Mark I, was being built at Harvard University by I.B.M. Both Dr. Aiken at Harvard and I.B.M. seems to take credit for its design. It is not clear to me how the design was arrived at.  I found out later from Dr. Grace Hopper, a mathematician who worked with us (and who was I feel the world first programmer, at least on the Mark I) that, as originally conceived and operated, the Mark I had no real sub-routine control ability. I f one wanted to perform a repetitive set of operations on different sets of data or of subsequent data in a successive approximation process, it was necessary to punch out the instruction routines instructions, over and over again, perhaps hundreds of times on a tape.

If we had employed this “linear string of instruction” approach on the ENIAC it would have taken over one million tubes in the programming or sequencing control system of ENIAC instead of a few thousand of the total of eighteen thousand tubes in the machine. This says to me that Mauchly’s idea was not obvious and was very clearly a new and important idea. I was told many years later by various people who had studied the work of Babbage that he and Lady Lovelace also had this subroutine idea. It is not mentioned in the Encyclopedia Britannica which was current at the time Mauchly and I developed the ENIAC and if Babbage and Lovelace had this idea they had not gotten the world to know about it. Of course Babbage never got any of his more ambitious machines and ideas to work, so if he had good ideas they died with him for all practical purposes.

2) The second important idea in the ENIAC was the idea of a General Purpose Register (GPR) which could be used for many purposes and which could be read into and out of, at electronic speed. I.B.M. in their tabulating equipment had electromechanical registers which tended to be specialized in their use. They were usually used to take data from cards and accumulate totals which could be printed on paper and or punched into other punched cards. The Mark I appeared to make better use of its counter ­registers but odd restrictions still remained. ENIAC ideas are the origin of our modern electronic Random Access Memory or RAM.

Both the Mark I and the ENIAC had banks of switches which provide much of the high speed memory required. These switches could be set by hand and then read out electrically at high speed as needed. Again these ENIAC ideas are the origin of our modern electronic non-volatile Read Only Memory or ROM.

3) The concept of rerouting the sequencing process by examining the value or sign of a particular number and then choosing an appropriate subroutine as a result allowed the ENIAC to make decisions based on numbers it had calculated and this feature gave it both great programming power and flexibility.

4) The concept of nesting and interlooping subroutines to produce complex results with comparatively little program switching equipment was intrinsic in the design of the ENIAC program system. It was common to have sub-routine loops within other sub-routine loops and so on, to avoid longer and more complex subroutines that would otherwise been required.

5) ENIAC had the ability to stop the process after each pulse time, after each addition or data transfer period, or at special points introduced in the subroutine process and in accord with some set of conditions or rules. The purpose of this was to facilitate trouble shooting of both the hardware and the software and to allow for human intervention in the problem solving or decision resolving processes being investigated.

6) ENIAC had the ability to provide automatic input, on demand from the process, from a stack of punched cards placed in a suitable machine. This provided a large inexpensive memory for some forms of problem, as well as a convenient data input device. Cards could also be punched up manually on a standard key punch. Cards could be sorted or collated on standard punch card equipment and thus the overall system combined the advantages of the new and the old, at least until new ideas could beat out the old ideas in all areas of data manipulation. Sorting turned out to be the hardest to perform economically on electronic computers but the problem was ultimately solved.

7) ENIAC also had similar output ability and could punch cards from data in the machine. Since these cards could be fed back into the computer a large memory for problems with a systematic arrangement of the data was provided. These output cards could be sent to a tabulator for printing and for certain types of data checking which would otherwise have tied up the ENIAC itself.

8) Partly because of the way the ENIAC grew in complexity as the project progressed and because the demands and expectations of the Army grew, and partly to provide as much flexibility as possible, each section of the machine, each register etc. of the machine had its own program control system built into it. This allowed each unit to be tested without to much dependence on other units. But it also allowed operation of several sections of the machine at one time. Most problems were of sufficient complexity that this parallel computing ability was seldom used. However the ENIAC allowed for parallel operation of several processes at the same time.

9) There were also banks of buffer relays between the punch card machines and the rest of the ENIAC which allowed data to be put in and data to be taken out and data processing to occur; all three at the same time in order to save time.

10) The ENIAC was broken up into forty (40) main panels plus seven (7) power supply panels for servicing and manufacturing reasons. This was not too uncommon in radio transmitters and telephone work those days. But our panels differed in that they were largely made up of several dozen small chassis, each containing 12 to 28 vacuum tubes. This was unusual and the number of tubes is a panel, about 500, greatly exceeded the total tube count in anything built up to that time. These small chassis or “plug in units,” as we called them, were plugged in or out of the “back panels” of a panel unit with only the use of a special set of handles to force the units to plug into or out of the sockets on the panels as required. No screwdriver to disconnect wiring from wiring terminals was ever required only special handles which added greatly in removing the plug in units or modules was required.

There were a limited number of types of plug in units and one or more spares, depending on the number of that type used in the machine, were kept on hand. Today’s printed circuit cards that plug into mother boards are direct descendants of this idea, an idea not really exploited in any electronic equipment before ENIAC.

11) All the circuits of the ENIAC resulted from very carefully calculated design studies. Nearly all electronic design work either for radios or industrial control devices were not given much in the way of a calculated analysis to find out the effect on circuit operation of value variation in the components employed. On radio production lines those days, units that did not seem to work were pulled off at the end of the line where bench testing was done by a few special technicians. By trial and error and changing parts the technicians would get the units to work. On the ENIAC project we set up test limits for every important characteristic of each tube used and for each electronic part used. We first tested many tubes and other components to make sure high yields of these parts, which were hard to get during a war, would result. We worked on the design of the circuit until they would tolerate the wide tolerances we found. We would redesign until we had fairly non critical circuits. Some people later called this approach worst-worst case design since many parts could be at their worst limits at the same time in a given circuit. The theory was all available, but it was not being used as it should have been in 1943.

This idea of testing parts and then designing to tolerate the variations in economically produced parts is still a part of the computer business today and this approach really did not adequately exist to any great extent in consumer electronics before ENIAC.

12) ENIAC accomplished all of the above goals at speeds much in excess of all past human experience for devices which carried out complex sequential processes. The only digital computing machines that came earlier, and they were approximately 1000 times slower, were the electromechanical MARK I at HARVARD, and the relay “complex number multiplier” and the relay “interpolator”, these last two both from Bell Laboratories.

Most of the above 12 items are still found, in “astronomically improved” form, in today’s computers. We certainly are doing poorer than Euclid who had ideas that lasted, almost without change, for a couple of millennium. But it is hard to find anyone else whose “ideas” have held up without change as well as Euclid ideas.

The Stored Program

The big item not found in the above list is the concept of using a single high speed memory for both data and program instructions. Without this the modern computer would be a bust. Mauchly’s big idea, in my mind, was the subroutine control concept.

My big idea was the idea of the stored instruction sequence or program, using a single fast memory for both data and instruction with no distinction between registers used for many purposes.

At the time I first thought of this idea, (in January 1944), I knew of no good way to provide the required memory registers. My first idea was to use magnetic disc such as Bell Laboratories had used in some telephone sound recording work. But first let me tell you how I thought of the idea itself.

While we were building and testing the ENIAC there were periods, usually waiting for some wiring to be completed by a technician or some circuit to be “debugged” when I had time to think. My thoughts usually turned to what the next machine should be like. I thought it should have a magnetic wire or tape for input and output and probably a greater use of binary as opposed to decimal or coded decimal arithmetic. If we were free of punch cards there was no longer a good reason to stick to the decimal system.

The problem on which I always got stuck was the question of how to spend money on the various high speed electronic memories. In ENIAC we had one type of memory for numbers or data, another for instructions, another for fixed set of numbers and another for input output buffering. The problem was that each different problem in the future would require a different mix of these things. It occurred to me that most of the great mathematicians and scientists tried to avoid solving specific problems and tried to find general solutions to broad areas of problems. I felt we had to find a general, rather than a specific solution to the memory problem. Once you take this position you have no choice but to say: let’s have only one kind of memory for almost all our high speed purposes, except for some very few “working registers”.

I wrote, in January 1944, a memo proposing magnetic discs as memory for all forms of storage required. I later realized that a variation of a “mercury tank” which I had invented and developed for several radar problems at M.I.T and Harvard could be modified to be a random access memory and would be the best bet for our first stored program machine, the BINAC and our first commercial machine, the UNIVAC I. We also performed considerable work on an electrostatic storage tube approach.

I told these ideas to Mauchly, Goldstine and all the others on the staff of the ENIAC project. I also told these ideas, many months after January 1944 to John von Neumann at the suggestion of Goldstine who was the Army representative on the ENIAC project. Goldstine brought von Neumann, his “hero” at that time, to see us and to discuss the ENIAC and some of our newer and better ideas.

A few months later both Mauchly and I were shocked and really horrified to find that von Neumann had taken these ideas of ours, described them in terms of “neurons” instead of electronic circuits and presented them at public meetings; giving the impression that they were his ideas. We asked him about it and he said he was just writing up and discussing our ideas with others to clarify them in his own mind.

We were also shocked when Goldstine backed up von Neumann’s claims, even though I had explained these ideas to Goldstine well before von Neumann arrived on the scene. I could not believe Goldstine had not understood my explanation of these ideas presented to him well in advance of von Neumann’s arrival.

Neither Mauchly nor I were free to present a public paper on the subject because the computer work was still under Military classification. In my opinion we were clearly “suckered” by John von Neumann who succeeded in some circles at getting my ideas called the “von Neumann architecture”. Till his death von Neumann would never admit what he had done. He tried to patent my ideas on the subject and was told he had only the vague workings of the idea not enough to patent. Most engineers think I invented the current computer architecture ideas, however many mathematicians and programmers thing von Neumann did.

Many other stories about similar dealings with other people’s ideas have come to light in recent years about John von Neumann.

The Present and the Future

Well enough of the past. What about the present? One of my complaints about current computers is that they have too little built in checking. “Univac I” was quite full checked, by parity in the memories and in the instruction circuits and by duplication of the fairly simple serial arithmetic circuits. “Univac II” was similarly checked. By the time we got to the UNIVAC SOLID STATE machines and UNIVAC III we dropped checking, except for the parity checking of the memory devices. The rational was that with magnetic amplifiers and solid state diodes and later with transistors, the inherent reliability would be so high that checking would no longer be needed. In UNIVAC LARC built around 1960, a bit earlier than UNIVAC III we again had complete checking. L~C had over 60,000 transistors and over 200,000 diodes. More parts were used in a second LARC.

The real reason checking was dropped by UNIVAC was that I.B.M. never did much checking in their machines and to compete cost-wise with I.B.M. we had to degrade our machines and leave checking out of UNIVAC III and all of our later machines until fairly recently when UNIVAC again used dual processors in some systems to regain full checking, but only when it became very economical to do so using very large scale integrated circuits.

In our small machines today the processor is, or should be, a small part of the cost. The INTEL monopoly, on some of the latest chips, has kept their cost artificially high but, depending on the outcome of some current legal battles, many think this will change.

My guess is that soon full checking, by having two, or better yet self correcting with perhaps three processors, will only add 10% or 20% to the cost of a small computer, depending on how future fancy chip prices fall and how much self correction ability is put in the machine. The cost increase percentage is much smaller, just a few percent, if all the other items, printers, software and maintenance costs, etc. of a complete system are included in the comparison. Several Japanese companies have recently announced fairly small machines that include self-correction ability.

With full checking, when something goes wrong, you know immediately whether it is the software or the hardware causing the problem. I believe the savings with a simpler software burden for checking and the saving of maintenance time would much more than offset the small cost increase. Of course it depends on the application, but in networks and in systems where a number of terminals depend on a file server the effect would be most noticeable, especially if self correction with perhaps three (3) processors is employed.

Of course some main frame systems do have redundant equipment for checking, but few small systems do. I see the future, as I have for the last 25 or 30 years, dominated by small systems leaving the very large super systems for certain scientific and engineering problems. However most of the money will be in the small systems.

Despite the millions of people working on software all over the world, software, not hardware, is still by far the weakest part of a computing system as it has been since we first built the ENIAC. This may always be true unless a new way to program is invented.

People ask me if things have gone the way I expected in the early days. The answer is much better that expected on the hardware and much worse than expected on the software. The software situation is particularly distressing in the area of teaching software.

A friend in this field finds that educational software “kills itself” to be “video game like” and fun. He finds the standard business software, not only to be of better quality, but frequently better for the job. Of course you give up the little “icons” and pictures. The newer “window” programs may cause all this to be less of a problem as time goes on but they use up a lot of extra time and memory for a small return.

The whole question of how to teach mathematics and computers is confused by the fact that we once thought arithmetic, algebra and geometry were the real basics and all else could follow from there. Then teachers got into set theory but never taught it in enough depth to make sense. We have I believe in recent years backed away from the so-called “new math” anyway.

Computer graphics has put a whole new light on geometry and especially on topology more than perhaps in other areas of mathematics. As a result many think we no longer really know what the basics subjects that we should teach are.

In any case we should be teaching values and not just reading, writing and arithmetic. Perhaps we should teach the young much more about our present world problems. Not just about computers and programming but their use in simulation and modeling of problems. Maybe some history and other educational courses from the past will suffer but the idea that if we do not fully understand the errors of the past we will be condemned to make the same mistakes in the future has never made much sense to me. If John Mauchly and I had known a great deal about Babbage, would it have helped us or hindered us? I think the loss of time locking into Babbage, who had a totally different technology at his disposal, would have just wasted our time and hindered us.

Computers should not be seen by the young as “magic boxes” or just labor saving devices, but as a very powerful tool for solving problems of the future, a tool that we never had up until now and a big advantage in helping to understand how to do things like saving the environment and also to help us learn how the environment works. The environment is everyone’s and we should get some real understanding of the problems involved, perhaps even while in high school.

We have got to get our youth interested in making future progress and point out to them that many times the computer may be their greatest helper in understanding the future. We in the United States are shocked when American soldiers are killed in a war. Yet little is done to reduce the roughly 50,000 deaths a year from highway accidents. If we were to go back to the greater use of railroad trains to carry the long haul freight we could probably clear a great deal of the most dangerous kind of traffic from our highways. We would lower the maintenance cost of the highways and save tens of thousands of lives and considerable medical expense each year. The money saved would probably more than write off the cost of improving our rail service. As a bonus a railroad train takes about one-third (some say one-fifth) the fuel oil per pound carried per mile, compared to a large truck. It will take a lot of computer simulation to get this case across to the public and to the politicians (so something can be done about it).

Only the young will have the guts to build the case against the oil companies, the truck manufacturers, the highway builders, etc. and the labor unions, all of which profit from the way things are done now, with long haul large trucks and trailers on the highway. Recent legislation is aimed at allowing even more freedom to have larger trucks and trailers on longer hauls. The problem of “highway pollution” is getting worse not better.

In recent years computer have become several thousand times less expensive and several thousand times faster, all in the same machine, due to the large scale integrated circuit.

Yet in spite of all this we still have failed to have computers which are effective in language translation, an area where there was much optimism 30 years ago at our largest educational institutions.

We have failed to bring robots and automation, in any important way to the everyday problems of cutting the grass, cleaning our house, preparing our food and driving our cars. These things are little different than they were before computers. It is clearly not hardware that is limiting our solution to these problems. It is new ideas and new approaches that we need. We have barely scratched the surface of what can be done and should be done.

No one is setting any clear priorities on what the big problems of the world are. The first electronic digital computer, the ENIAC, came about as more or less of an accident, a war provided a problem and the money to start a project. What was done could have been accomplished 15 years earlier. There was little in the technology of vacuum tubes that improved, in any really important way, from say 1928 to 1943 when we started. Even early punched card equipment was available well before 1943.

The situation is the same today. Our efforts are mostly going into better telephone systems, better high fidelity sound systems, high definition television systems and better robots and other automation for industry.

Less work is going into how to get people fed, how to stop the destruction of our environment, how to better educate our young people without subjecting them to more and more strain in the process. Even with reasonable powerful computers Russia was not able to solve its economic problem and is now going to a free market economy.

Yet one would think that with modern computers something could be done to plan our economic progress and avoid peaks and valleys in our economies. But so far we have largely used computers to help big investors play the stock market. This has not helped our economy and sometimes may have hurt matters.

Regulatory laws have helped to stabilize the economies of the world but these started back in the 1930’s, long before computers. Look at the banking disaster in the United States. Was the U.S. government using their computers to properly monitor the banking system that they had undertaken to insure?

We need to put a lot more effort into simulating the future with computers and even more effort into getting leaders into government who will listen to the results obtained from the computers. In spite of several energy crises in the United States, we are still dependent on foreign oil to too great an extent. The various conservation efforts have not gone forward as they should. The most recent U. S. energy policy is little more than hope, it is not a real energy plan. We do not need an energy czar, but we do need a better energy plan and some real leadership.

In the years since I was born, 72 years ago, it has become more and more evident that we cannot just continue to populate the world in an ever-expanding way without solving the population and environmental problems that have been created. Most wars grow out of economic problems and yet no form of government has yet succeeded in knowing how to manage the problems.

How much worse have things got to get before our leaders get better. This must all start with the young and with education. While computers are not the solution, plans for the future are, and computers and people who can use and write programs for computers are needed to aid the future leadership of the world.

In spite of all the problems of the world, which I think the computer can help more than anything else, working on computers has always been a lot of fun and I feel very fortunate that I have been able to be a part of it.

It has also been fun to travel to Japan and talk to the Japanese people about computers.

As I mentioned earlier, I have given few speeches in recent years, the last one I remember was a few years ago at the Computer Museum in Boston, Mass., at the 40th birthday celebration of the ENIAC.

I wish to end my speech with the same line I ended that speech with:

How would you like to have most of your life’s work end up on a square centimeter of silicon?

YESTERDAY. TODAY AND TOMORROW

Keynote speech -April 15, cialis sale 1991 -Imperial Hotel -Tokyo, viagra Japan

by Dr. J. Presper Eckert

Copyright © 1991 J. Presper Eckert, Jr.

10th anniversary of ERIC (Eckert Research International Corporation)

I am very happy to be back in Tokyo again and to be able to hear in person some of the things you in Japan are thinking about in computers. I am happy to talk to you about my favorite subject, computers. It has been several years since I have talked at any length to a large group about computers.

At a recent party, at my daughter’s house, I counted that over half of the people present used computers in their work and in many cases were dependant on them for their livelihood. Only a few of them were friends of mine, most of them were friends of my daughter and her husband and were of their generation.

The point I am trying to make is that computers are now approaching the universal nature of the printing press, which in many cases it has replaced. It is more universal than the book, the magazine and the newspaper because it now makes up a major part of our communication systems as well. In addition it can assist in retrieving and locating the information we need. Books are often much slower in this area.

The computer now forms a part of our automobiles, our microwave ovens, our robots and other forms of factory automation. To me, one of the most exciting uses of computers is in medicine. Computers have greatly improved the ability of physicians to diagnose the human body. We now have Cat Scan, Magnetic resonance, and Ultrasonic imaging devices.

Both John Mauchly, my partner, and I were frequently asked if, in the early days we could visualize all that was happening and would happen in computers. The answer is yes and no, we thought that much could be done both in science, engineering and business with the digital approach, because of its accuracy, flexibility and resistance to deterioration of the information involved. However, no one expected that the invention of the very large scale integrated silicon chip would occur as soon as it did and that it would prove to be as reliable and as inexpensive as it has turned out to be. But we always believed that the digital approach would wipe out the analogue approach, not only in most existing areas, but in many new areas as well. We thought it would aid many, if not all, areas of life. We thought all this would take time, considerably more time that it appears to be taking.

Answers to common questions

Two young men sent me a series of questions which were clearer and perhaps more to the point than those asked by many professional journalists. In any case their questions lend themselves to fairly flexible answers to some of the important things that happened to Dr. John Mauchly and myself in the pursuit of high speed electronic digital computers.

Their questions were:

QUESTION: 1) What gave you the idea to develop a computer?

ANSWER: Several things converged to give Mauchly and I the idea of building a computer. We were both working at the University of Pennsylvania during World War II. The army was using two of the three Differential Analyzers in the world at that time. The first one was developed by Dr. Vanavar Bush at M.I.T. and was being used by the Navy. All three were being used to do various ballistic calculations. The two machine the Army were using were improved copies of the M.I.T. Bush machine and were built in the 1930’s to provide employment, in the depression, to engineers and designers. They were built at government expense. They were rather slow and not sufficiently accurate for most ballistic work. The one percent (1%) accuracy achieved was however suitable for many engineering problems at the University before World War II.

Working with Dr. Cornelius Weygandt and other university employees we managed, by the addition of about 400 vacuum tubes and a considerable number of special motors and generators, along with photo-cell devices and light polarizing devices, to speed up the Differential Analyzer by about ten times and at the same time to achieve about ten times the accuracy (0.1%) “Annie”, as we nicknamed the machine, was in spite of the added electronics, still an analogue device and was not only severely limited in speed and accuracy but was only useful on certain kinds of differential and integral equations. Not only did the research on government weapons require more speed, accuracy and flexibility than this, but many scientific and engineering problems also required more of these abilities.

John Mauchly and I discussed making electronic integrators using electronic pulse counters. I had worked at the University of Pennsylvania on some digital counters for a special high accuracy Radar project and on a Mercury Delay Line timing device (which I invented). We saw how some of these ideas could be applied to an electronic integrator. The mechanical integrator was the weakest part of the differential analyzer. We decided that while the electronic integrator would improve the Differential Analyzer it would still be severely limited in speed by the remaining gears and shafts that interconnected the system as well as by the electromechanical servomechanisms that now drove them.

We rather quickly concluded that any sensible improvement in computing equipment would not only have to be all electronic, but would have to encode the transmitted information in a far more efficient manner than simple strings of pulses, such as a pulse integrating circuit might employ. We thus decided that our first approach should be all digital, using codes that would allow the use of existing punch card equipment for input output purposes.

QUESTION: 2) How many attempts did you make before the final project was achieved?

ANSWER: We really made only one carefully organized attempt to build the original ENIAC. We made no false starts in designing equipment. We refined our original ideas, as outlined above, in the early discussion stages. We studied over half a dozen counter circuits. These circuits were obtained largely from the work of physicists who used them for counters in radiation and cosmic ray studies. We found the physicists’ designs were generally too slow and too unreliable. We then worked out a fairly fast and quite reliable design of our own based in part on some circuitry that had been developed for timing purpose in measuring projectile motion in ballistic tests.

We went to the RCA tube research laboratories and obtained clues as to how to operate vacuum tubes to achieve long life and trouble free operation. We conducted tests on thousands of tubes and set up standards to design circuits that would allow proper operation in spite of large variations in the tubes, resistors and capacitors, that comprised the circuits. We also designed circuits and power supplies that were quite insensitive to power line voltage variations. In addition we used a voltage regulation system on the power line to gain added reliability.

We made a few “bread board” models of some of the key circuits and built a “prototype” model of two accumulators (which were really electronic adding machines). They were of the design which we would use in the final machine. With a few changes and improvements these prototypes worked fine. All this took about one year from when we started. We then built twenty accumulators as well as the rest of the machine.

In the next year and a half we built and tested thirty-seven electronic panels, each two feet wide and eight feet high. Each contained about 500 vacuum tubes (or radio tubes). We also built three relay panels, using very reliable telephone relays for buffering the data to and from the punch card machines. These buffers allowed us to read cards, punch cards and compute; all at the same time. These three panels were the same size as the electronic panels and contained about 1500 relays or about 500 relays per panel. In addition we designed and built a very extensive centralized six panel power supply system. These panels were larger and deeper than the forty panels, which made up the machine proper. A single special capacitor panel was placed in an adjoining room to keep these power supply filter capacitors as cool as possible.

Additional punch card machines, for creating input keypunch data, sorting data, printing data, etc. were put in a hallway leading to the main thirty by fifty foot room in which the ENIAC was housed.

QUESTION: 3) How long did it take to achieve your goal?

ANSWER: It took two and one-half years to design and build the machine from when we started. But it took months after this to work out small problems as they showed up in various test problems, and to start to understand better how to program the machine. Three years is probably the best answer as to how long it took. Various improvement and additions were made to the machine for some years to follow. The machine remained in use at the Ballistic Research Laboratory at Aberdeen, Maryland for about 10 years.

QUESTION: 4) How did the public respond to the development of the ENIAC?

ANSWER: The news media kept referring to the machine as a “giant brain”, rather than as an automatic machine for performing arithmetic on many numbers in accord with a predetermined sequence and/or a set of rules for manipulating the numbers. As a result an overly complex and misleading picture was formed in the public mind. Many people were however not confused by all this and interest grew quickly. We found little foresight in the minds of banking and financial circles and did not received help from the people who traditionally claim to encourage the future.

We worked on a number of government projects to get money to keep going. Some of this work probably held us back more than it helped us.

Many business men thought we would not be able to do what we said we could do. It was very tough to get started.

QUESTION: 5) How did your lifestyle change after the ENIAC was invented?

ANSWER: My life style was not really affected very much by the computer, except that it took all my time. But anything I got interested in would probably have done the same. My father made a fairly good income, I went to a private school (William Penn Charter School -now over 300 years old -it was started by William Penn and is I believe the oldest private boys school in the U.S.). I went to an Ivy League college, the University of Pennsylvania.

My father and mother traveled a lot. I was an only child. They always took me with them. By the time I was 12 years old I had traveled about 125,000 miles and had been in every state (48 then) and Alaska, most of Europe and part of Asia. We cruised almost to the North Pole. We had traveled by ships, cars, trains, horses, mules and even camels (no airplanes those days) . In the years that followed the ENIAC I frequently traveled 125,000 miles in a single year, mostly by airplane.

I met president Harding when I was 5 years old, many movie stars before I was 8, and was given a medal for inventing the computer by President Johnson in 1968 at age 49. My life has been exciting both before and after John Mauchly and I invented the computer.

QUESTION: 6) Although the ENIAC is perceived by most as the first computer, isn’t the ABC computer at Iowa State the first computer?

ANSWER: The work by Dr. Atanasoff in Iowa was, in my opinion, a joke. He never really got anything to work. He had no programming system. He tried for a patent and was told the work he has done was too incomplete to get a patent. A competitor in a patent suit convinced, what in my mind was a very confused judge, to believe Atanasoff ‘s story, even though it had no real relation to the case at hand. The judge wrote a confused and irrelevant opinion about the invention of the computer. We had convinced judges in cases with LB.M and A.T.& T. that we had invented the computer. The lawsuits got too expensive; the last one cost over one million dollars. We could have gone on to appeal but the patent would have run out before the matter would have ever been settled. We simply quit wasting time on a patent system that could not really understand what we had done at the time we did it. The U.S. government supplied patent attorneys were so inept at their work, in my opinion, that the work was delayed too long and not carried out at all well. This resulted in great confusion later.

Who invented the first of anything is always confused in history.  To make the point let me quote from the fourteenth edition of The Encyclopedia Britannica, 1929. Volume 14, Page 105.

“The earliest attempt at making an incandescent lamp was made by Dc la Rue in 1820. In 1840 Grove demonstrated his battery by lighting an auditorium with incandescent electric light.”

The encyclopedia goes on and mentions a British patent to Frederick De Moleyns issued in 1841. J. W. Starr patented a carbon filament lamp in 1845. Sir Joseph W. Swan manufactured incandescent lamps from 1848 to 1860. All in all about 100 people made various incandescent lamps before Thomas A Edison developed his world famous lamp.

It is clear that Edison did not invent the incandescent lamp. He lit up Menlo Park in December 1879 about 59 years after De la Rue started it all in 1820 and 34 years after the carbon filament idea he used was invented.

What then did Edison really do?

1)         He made a lamp with a long skinny carbon filament (also invented in England at about the same time by Swan)

2)         He used a completely glass sealed vacuum chamber.

3)         He used a rather high vacuum. Made possible by the vacuum pump work of Herman Sprengel in 1865 and the improved vacuum techniques of Sir William Crookes developed in 1875.

4)         He sealed the wires into the lamp with platinum wires which matched the temperature coefficient of the glass the lamp was made of to avoid the vacuum loss of earlier designs

He thus combined four known ideas into one and made a lamp which consumed 7.5 Watts, had a life of about 40 burning hours, and worked on a high voltage (110 volts). It used a very low current, so low that many bulbs could work in parallel on reasonable size (and cost) copper feed wire from a generator with reasonable size commutators. Earlier lamps used low voltages and high currents, frequently from batteries. They lasted perhaps two (2) hours before a replacement of the filament was required.

The earliest practical generator for lighting was developed by Gramme as a so-called dynamo in 1875 only four years before Edison’s lighting demonstration at Menlo Park.

Edison also needed a wattmeter if he was to sell a practical lighting system to people. He invented a chemical wattmeter in the same year as his lamp (1879) to solve this problem. A bit later a high school teacher at Central High School in Philadelphia, Professor Elihu Thompson invented a much better electromechanical wattmeter without chemicals and it properly completed Edison’s system. Thompson and his school teacher partner, Houston, had formed a company to make generators, wattmeters, etc. Thompson is also considered the inventor of electric welding. Thompson and Houston’s company later merged with Edison to form “General Electric”.

My father, when a teenager was taught to swim by Houston, who was much older and who lived as a bachelor in a large house in West Philadelphia.

The point of all this is that Edison had a complete lighting system and a practical incandescent lamp. Others had not achieved this. However the basic idea was quite old at the time of his invention.

Mauchly and I achieved a complete workable computing system. Others had not. In fact others had not even achieved the essential functioning parts of an electronic computing system.

If Edison is the inventor of the incandescent lamp it would appear that by the same yardstick Mauchly and I are clearly the inventors of the computer.

Frankly, I find the claims of Dr. Atanasoff and the Iowa people to be absurd and inaccurate to say the least; furthermore it took them years to come up with their claims well after the computer was a fully accepted idea.

One could carry the discussion of what constitutes an invention on with a lot of additional examples. The Wright brothers’ invention of the airplane is another good example. Here again the Wright brothers developed a complete and working system. The idea of wings and a propeller driven by some kind of engine was already old when they started their work, they were late comers, just as Edison was with his light bulb. Dr. Langley had built a model airplane driven by a steam engine about ten years before the Wright brothers had their success. It was in its way successful; it made a flight of about 4000 feet (over 1000 meters).

But the Wright brothers were not satisfied, they felt that with their bicycle making ability and some hard work that they could make a machine to carry a man and fly it much further than others had been able to. They studied the design ideas of Dr. Laungley and were not satisfied with his calculations. To learn more before they made a final design they spent two years building and experimenting with various wing shapes in their experimental wind tunnel starting in 1901. They found a lot wrong with the approach of Dr. Langley and others and they developed their own approach.

They went further than anyone else had and concluded that airplanes were not inherently stable and needed more control from the pilot to properly maneuver them. They need more than just a rudder. They fastened ropes to the wings of their early planes so they could bend the wings to achieve the desired flying stability and avoid a crash. They had thus invented the first form of the “aileron” control, an essential ingredient to a practical and complete flying system. still it took them two more years before they found out how to avoid a “tail-spin” in making a turn.

Without going on and on about inventors it can be seen that inventions of importance are not just a “flash of genius”, as many think, they are the incorporation of such flashes into a complete working system. Before the Wright brothers and the aileron control pilots had attempted to stabilize flight by throwing their body weight around to control the planes stability. Can you imagine doing this in a modern airplane?

The contributions of Edison can still be recognized in the modern light bulb. So can the aileron control of the Wright brothers still be recognized in a modern airplane?

Can the contributions of Eckert and Mauchly to the original ENIAC and later the Univac still be recognized in the modern computer? I hope and believe they still can be.

Of course few in the world will ever compete with the mathematician Euclid whose historic work on geometry remained nearly unchanged for over 2000 years. It is only somewhat over one hundred years ago that alternate, non-Euclidian, geometries were developed, even then they have not really displaced Euclid’s work.

ENIAC Firsts

I have now attempted to draw up a list of what I think are the important “firsts” that came about through the development of the ENIAC.

1) I think the most important item in the ENIAC was the control of the subroutines in programming. This idea was first proposed to me by Mauchly, and it became immediately clear that it was absolutely essential to the design and construction of the ENIAC. While neither Mauchly nor I knew it at the time, an electromechanical machine, constructed of I.B.M. tabulator parts, and called the Mark I, was being built at Harvard University by I.B.M. Both Dr. Aiken at Harvard and I.B.M. seems to take credit for its design. It is not clear to me how the design was arrived at.  I found out later from Dr. Grace Hopper, a mathematician who worked with us (and who was I feel the world first programmer, at least on the Mark I) that, as originally conceived and operated, the Mark I had no real sub-routine control ability. I f one wanted to perform a repetitive set of operations on different sets of data or of subsequent data in a successive approximation process, it was necessary to punch out the instruction routines instructions, over and over again, perhaps hundreds of times on a tape.

If we had employed this “linear string of instruction” approach on the ENIAC it would have taken over one million tubes in the programming or sequencing control system of ENIAC instead of a few thousand of the total of eighteen thousand tubes in the machine. This says to me that Mauchly’s idea was not obvious and was very clearly a new and important idea. I was told many years later by various people who had studied the work of Babbage that he and Lady Lovelace also had this subroutine idea. It is not mentioned in the Encyclopedia Britannica which was current at the time Mauchly and I developed the ENIAC and if Babbage and Lovelace had this idea they had not gotten the world to know about it. Of course Babbage never got any of his more ambitious machines and ideas to work, so if he had good ideas they died with him for all practical purposes.

2) The second important idea in the ENIAC was the idea of a General Purpose Register (GPR) which could be used for many purposes and which could be read into and out of, at electronic speed. I.B.M. in their tabulating equipment had electromechanical registers which tended to be specialized in their use. They were usually used to take data from cards and accumulate totals which could be printed on paper and or punched into other punched cards. The Mark I appeared to make better use of its counter ­registers but odd restrictions still remained. ENIAC ideas are the origin of our modern electronic Random Access Memory or RAM.

Both the Mark I and the ENIAC had banks of switches which provide much of the high speed memory required. These switches could be set by hand and then read out electrically at high speed as needed. Again these ENIAC ideas are the origin of our modern electronic non-volatile Read Only Memory or ROM.

3) The concept of rerouting the sequencing process by examining the value or sign of a particular number and then choosing an appropriate subroutine as a result allowed the ENIAC to make decisions based on numbers it had calculated and this feature gave it both great programming power and flexibility.

4) The concept of nesting and interlooping subroutines to produce complex results with comparatively little program switching equipment was intrinsic in the design of the ENIAC program system. It was common to have sub-routine loops within other sub-routine loops and so on, to avoid longer and more complex subroutines that would otherwise been required.

5) ENIAC had the ability to stop the process after each pulse time, after each addition or data transfer period, or at special points introduced in the subroutine process and in accord with some set of conditions or rules. The purpose of this was to facilitate trouble shooting of both the hardware and the software and to allow for human intervention in the problem solving or decision resolving processes being investigated.

6) ENIAC had the ability to provide automatic input, on demand from the process, from a stack of punched cards placed in a suitable machine. This provided a large inexpensive memory for some forms of problem, as well as a convenient data input device. Cards could also be punched up manually on a standard key punch. Cards could be sorted or collated on standard punch card equipment and thus the overall system combined the advantages of the new and the old, at least until new ideas could beat out the old ideas in all areas of data manipulation. Sorting turned out to be the hardest to perform economically on electronic computers but the problem was ultimately solved.

7) ENIAC also had similar output ability and could punch cards from data in the machine. Since these cards could be fed back into the computer a large memory for problems with a systematic arrangement of the data was provided. These output cards could be sent to a tabulator for printing and for certain types of data checking which would otherwise have tied up the ENIAC itself.

8) Partly because of the way the ENIAC grew in complexity as the project progressed and because the demands and expectations of the Army grew, and partly to provide as much flexibility as possible, each section of the machine, each register etc. of the machine had its own program control system built into it. This allowed each unit to be tested without to much dependence on other units. But it also allowed operation of several sections of the machine at one time. Most problems were of sufficient complexity that this parallel computing ability was seldom used. However the ENIAC allowed for parallel operation of several processes at the same time.

9) There were also banks of buffer relays between the punch card machines and the rest of the ENIAC which allowed data to be put in and data to be taken out and data processing to occur; all three at the same time in order to save time.

10) The ENIAC was broken up into forty (40) main panels plus seven (7) power supply panels for servicing and manufacturing reasons. This was not too uncommon in radio transmitters and telephone work those days. But our panels differed in that they were largely made up of several dozen small chassis, each containing 12 to 28 vacuum tubes. This was unusual and the number of tubes is a panel, about 500, greatly exceeded the total tube count in anything built up to that time. These small chassis or “plug in units,” as we called them, were plugged in or out of the “back panels” of a panel unit with only the use of a special set of handles to force the units to plug into or out of the sockets on the panels as required. No screwdriver to disconnect wiring from wiring terminals was ever required only special handles which added greatly in removing the plug in units or modules was required.

There were a limited number of types of plug in units and one or more spares, depending on the number of that type used in the machine, were kept on hand. Today’s printed circuit cards that plug into mother boards are direct descendants of this idea, an idea not really exploited in any electronic equipment before ENIAC.

11) All the circuits of the ENIAC resulted from very carefully calculated design studies. Nearly all electronic design work either for radios or industrial control devices were not given much in the way of a calculated analysis to find out the effect on circuit operation of value variation in the components employed. On radio production lines those days, units that did not seem to work were pulled off at the end of the line where bench testing was done by a few special technicians. By trial and error and changing parts the technicians would get the units to work. On the ENIAC project we set up test limits for every important characteristic of each tube used and for each electronic part used. We first tested many tubes and other components to make sure high yields of these parts, which were hard to get during a war, would result. We worked on the design of the circuit until they would tolerate the wide tolerances we found. We would redesign until we had fairly non critical circuits. Some people later called this approach worst-worst case design since many parts could be at their worst limits at the same time in a given circuit. The theory was all available, but it was not being used as it should have been in 1943.

This idea of testing parts and then designing to tolerate the variations in economically produced parts is still a part of the computer business today and this approach really did not adequately exist to any great extent in consumer electronics before ENIAC.

12) ENIAC accomplished all of the above goals at speeds much in excess of all past human experience for devices which carried out complex sequential processes. The only digital computing machines that came earlier, and they were approximately 1000 times slower, were the electromechanical MARK I at HARVARD, and the relay “complex number multiplier” and the relay “interpolator”, these last two both from Bell Laboratories.

Most of the above 12 items are still found, in “astronomically improved” form, in today’s computers. We certainly are doing poorer than Euclid who had ideas that lasted, almost without change, for a couple of millennium. But it is hard to find anyone else whose “ideas” have held up without change as well as Euclid ideas.

The Stored Program

The big item not found in the above list is the concept of using a single high speed memory for both data and program instructions. Without this the modern computer would be a bust. Mauchly’s big idea, in my mind, was the subroutine control concept.

My big idea was the idea of the stored instruction sequence or program, using a single fast memory for both data and instruction with no distinction between registers used for many purposes.

At the time I first thought of this idea, (in January 1944), I knew of no good way to provide the required memory registers. My first idea was to use magnetic disc such as Bell Laboratories had used in some telephone sound recording work. But first let me tell you how I thought of the idea itself.

While we were building and testing the ENIAC there were periods, usually waiting for some wiring to be completed by a technician or some circuit to be “debugged” when I had time to think. My thoughts usually turned to what the next machine should be like. I thought it should have a magnetic wire or tape for input and output and probably a greater use of binary as opposed to decimal or coded decimal arithmetic. If we were free of punch cards there was no longer a good reason to stick to the decimal system.

The problem on which I always got stuck was the question of how to spend money on the various high speed electronic memories. In ENIAC we had one type of memory for numbers or data, another for instructions, another for fixed set of numbers and another for input output buffering. The problem was that each different problem in the future would require a different mix of these things. It occurred to me that most of the great mathematicians and scientists tried to avoid solving specific problems and tried to find general solutions to broad areas of problems. I felt we had to find a general, rather than a specific solution to the memory problem. Once you take this position you have no choice but to say: let’s have only one kind of memory for almost all our high speed purposes, except for some very few “working registers”.

I wrote, in January 1944, a memo proposing magnetic discs as memory for all forms of storage required. I later realized that a variation of a “mercury tank” which I had invented and developed for several radar problems at M.I.T and Harvard could be modified to be a random access memory and would be the best bet for our first stored program machine, the BINAC and our first commercial machine, the UNIVAC I. We also performed considerable work on an electrostatic storage tube approach.

I told these ideas to Mauchly, Goldstine and all the others on the staff of the ENIAC project. I also told these ideas, many months after January 1944 to John von Neumann at the suggestion of Goldstine who was the Army representative on the ENIAC project. Goldstine brought von Neumann, his “hero” at that time, to see us and to discuss the ENIAC and some of our newer and better ideas.

A few months later both Mauchly and I were shocked and really horrified to find that von Neumann had taken these ideas of ours, described them in terms of “neurons” instead of electronic circuits and presented them at public meetings; giving the impression that they were his ideas. We asked him about it and he said he was just writing up and discussing our ideas with others to clarify them in his own mind.

We were also shocked when Goldstine backed up von Neumann’s claims, even though I had explained these ideas to Goldstine well before von Neumann arrived on the scene. I could not believe Goldstine had not understood my explanation of these ideas presented to him well in advance of von Neumann’s arrival.

Neither Mauchly nor I were free to present a public paper on the subject because the computer work was still under Military classification. In my opinion we were clearly “suckered” by John von Neumann who succeeded in some circles at getting my ideas called the “von Neumann architecture”. Till his death von Neumann would never admit what he had done. He tried to patent my ideas on the subject and was told he had only the vague workings of the idea not enough to patent. Most engineers think I invented the current computer architecture ideas, however many mathematicians and programmers thing von Neumann did.

Many other stories about similar dealings with other people’s ideas have come to light in recent years about John von Neumann.

The Present and the Future

Well enough of the past. What about the present? One of my complaints about current computers is that they have too little built in checking. “Univac I” was quite full checked, by parity in the memories and in the instruction circuits and by duplication of the fairly simple serial arithmetic circuits. “Univac II” was similarly checked. By the time we got to the UNIVAC SOLID STATE machines and UNIVAC III we dropped checking, except for the parity checking of the memory devices. The rational was that with magnetic amplifiers and solid state diodes and later with transistors, the inherent reliability would be so high that checking would no longer be needed. In UNIVAC LARC built around 1960, a bit earlier than UNIVAC III we again had complete checking. L~C had over 60,000 transistors and over 200,000 diodes. More parts were used in a second LARC.

The real reason checking was dropped by UNIVAC was that I.B.M. never did much checking in their machines and to compete cost-wise with I.B.M. we had to degrade our machines and leave checking out of UNIVAC III and all of our later machines until fairly recently when UNIVAC again used dual processors in some systems to regain full checking, but only when it became very economical to do so using very large scale integrated circuits.

In our small machines today the processor is, or should be, a small part of the cost. The INTEL monopoly, on some of the latest chips, has kept their cost artificially high but, depending on the outcome of some current legal battles, many think this will change.

My guess is that soon full checking, by having two, or better yet self correcting with perhaps three processors, will only add 10% or 20% to the cost of a small computer, depending on how future fancy chip prices fall and how much self correction ability is put in the machine. The cost increase percentage is much smaller, just a few percent, if all the other items, printers, software and maintenance costs, etc. of a complete system are included in the comparison. Several Japanese companies have recently announced fairly small machines that include self-correction ability.

With full checking, when something goes wrong, you know immediately whether it is the software or the hardware causing the problem. I believe the savings with a simpler software burden for checking and the saving of maintenance time would much more than offset the small cost increase. Of course it depends on the application, but in networks and in systems where a number of terminals depend on a file server the effect would be most noticeable, especially if self correction with perhaps three (3) processors is employed.

Of course some main frame systems do have redundant equipment for checking, but few small systems do. I see the future, as I have for the last 25 or 30 years, dominated by small systems leaving the very large super systems for certain scientific and engineering problems. However most of the money will be in the small systems.

Despite the millions of people working on software all over the world, software, not hardware, is still by far the weakest part of a computing system as it has been since we first built the ENIAC. This may always be true unless a new way to program is invented.

People ask me if things have gone the way I expected in the early days. The answer is much better that expected on the hardware and much worse than expected on the software. The software situation is particularly distressing in the area of teaching software.

A friend in this field finds that educational software “kills itself” to be “video game like” and fun. He finds the standard business software, not only to be of better quality, but frequently better for the job. Of course you give up the little “icons” and pictures. The newer “window” programs may cause all this to be less of a problem as time goes on but they use up a lot of extra time and memory for a small return.

The whole question of how to teach mathematics and computers is confused by the fact that we once thought arithmetic, algebra and geometry were the real basics and all else could follow from there. Then teachers got into set theory but never taught it in enough depth to make sense. We have I believe in recent years backed away from the so-called “new math” anyway.

Computer graphics has put a whole new light on geometry and especially on topology more than perhaps in other areas of mathematics. As a result many think we no longer really know what the basics subjects that we should teach are.

In any case we should be teaching values and not just reading, writing and arithmetic. Perhaps we should teach the young much more about our present world problems. Not just about computers and programming but their use in simulation and modeling of problems. Maybe some history and other educational courses from the past will suffer but the idea that if we do not fully understand the errors of the past we will be condemned to make the same mistakes in the future has never made much sense to me. If John Mauchly and I had known a great deal about Babbage, would it have helped us or hindered us? I think the loss of time locking into Babbage, who had a totally different technology at his disposal, would have just wasted our time and hindered us.

Computers should not be seen by the young as “magic boxes” or just labor saving devices, but as a very powerful tool for solving problems of the future, a tool that we never had up until now and a big advantage in helping to understand how to do things like saving the environment and also to help us learn how the environment works. The environment is everyone’s and we should get some real understanding of the problems involved, perhaps even while in high school.

We have got to get our youth interested in making future progress and point out to them that many times the computer may be their greatest helper in understanding the future. We in the United States are shocked when American soldiers are killed in a war. Yet little is done to reduce the roughly 50,000 deaths a year from highway accidents. If we were to go back to the greater use of railroad trains to carry the long haul freight we could probably clear a great deal of the most dangerous kind of traffic from our highways. We would lower the maintenance cost of the highways and save tens of thousands of lives and considerable medical expense each year. The money saved would probably more than write off the cost of improving our rail service. As a bonus a railroad train takes about one-third (some say one-fifth) the fuel oil per pound carried per mile, compared to a large truck. It will take a lot of computer simulation to get this case across to the public and to the politicians (so something can be done about it).

Only the young will have the guts to build the case against the oil companies, the truck manufacturers, the highway builders, etc. and the labor unions, all of which profit from the way things are done now, with long haul large trucks and trailers on the highway. Recent legislation is aimed at allowing even more freedom to have larger trucks and trailers on longer hauls. The problem of “highway pollution” is getting worse not better.

In recent years computer have become several thousand times less expensive and several thousand times faster, all in the same machine, due to the large scale integrated circuit.

Yet in spite of all this we still have failed to have computers which are effective in language translation, an area where there was much optimism 30 years ago at our largest educational institutions.

We have failed to bring robots and automation, in any important way to the everyday problems of cutting the grass, cleaning our house, preparing our food and driving our cars. These things are little different than they were before computers. It is clearly not hardware that is limiting our solution to these problems. It is new ideas and new approaches that we need. We have barely scratched the surface of what can be done and should be done.

No one is setting any clear priorities on what the big problems of the world are. The first electronic digital computer, the ENIAC, came about as more or less of an accident, a war provided a problem and the money to start a project. What was done could have been accomplished 15 years earlier. There was little in the technology of vacuum tubes that improved, in any really important way, from say 1928 to 1943 when we started. Even early punched card equipment was available well before 1943.

The situation is the same today. Our efforts are mostly going into better telephone systems, better high fidelity sound systems, high definition television systems and better robots and other automation for industry.

Less work is going into how to get people fed, how to stop the destruction of our environment, how to better educate our young people without subjecting them to more and more strain in the process. Even with reasonable powerful computers Russia was not able to solve its economic problem and is now going to a free market economy.

Yet one would think that with modern computers something could be done to plan our economic progress and avoid peaks and valleys in our economies. But so far we have largely used computers to help big investors play the stock market. This has not helped our economy and sometimes may have hurt matters.

Regulatory laws have helped to stabilize the economies of the world but these started back in the 1930’s, long before computers. Look at the banking disaster in the United States. Was the U.S. government using their computers to properly monitor the banking system that they had undertaken to insure?

We need to put a lot more effort into simulating the future with computers and even more effort into getting leaders into government who will listen to the results obtained from the computers. In spite of several energy crises in the United States, we are still dependent on foreign oil to too great an extent. The various conservation efforts have not gone forward as they should. The most recent U. S. energy policy is little more than hope, it is not a real energy plan. We do not need an energy czar, but we do need a better energy plan and some real leadership.

In the years since I was born, 72 years ago, it has become more and more evident that we cannot just continue to populate the world in an ever-expanding way without solving the population and environmental problems that have been created. Most wars grow out of economic problems and yet no form of government has yet succeeded in knowing how to manage the problems.

How much worse have things got to get before our leaders get better. This must all start with the young and with education. While computers are not the solution, plans for the future are, and computers and people who can use and write programs for computers are needed to aid the future leadership of the world.

In spite of all the problems of the world, which I think the computer can help more than anything else, working on computers has always been a lot of fun and I feel very fortunate that I have been able to be a part of it.

It has also been fun to travel to Japan and talk to the Japanese people about computers.

As I mentioned earlier, I have given few speeches in recent years, the last one I remember was a few years ago at the Computer Museum in Boston, Mass., at the 40th birthday celebration of the ENIAC.

I wish to end my speech with the same line I ended that speech with:

How would you like to have most of your life’s work end up on a square centimeter of silicon?
Is man’s best friend still a dog?  Or are you more likely to choose a laptop as your desert island companion?

February 15, pharmacy 1946 is the day that ENIAC was shown to the world—and the world is still giddy with the trillion ways your life is better with a computer.

See today’s Philly Post: Happy Birthday ENIAC. It’s a great overview of the ENIAC (and our fight to give it the respect it earned!).

Also new: Marty Moss-Coane on WHYY Radio Times did a nice hour program called The ENIAC Anniversary.   It featured an interview with Mitch Marcus from Penn and a few playful pokes at Jane Smiley, who called in to try to defend her minority report.

Also: A TOP TEN LIST in TechnicallyPhilly.
 

by Bill Mauchly

March 23, seek 2011 – Jean Jennings Bartik, nurse pioneer software engineer, died at age 86.  She was one of the original programmers for the ENIAC, and continued in a career in the computer industry.  She worked closely with Eckert and Mauchly at the Eckert Mauchly Computer Corporation in late 1940’s, an environment of innovation she referred to as a “Technical Camelot.”

I am going to miss her.  Jean was a ball of fire; she was always full of energy, enthusiasm, good stories and an uncensored opinion.  “The ENIAC was a son-of-a-bitch to program,” she once explained.  And she wasn’t being sensational – she was trying to be honest!  She knew – it was her job.  By 1947 she was head of a group of programmers at the Moore School writing code for the new-and-improved stored-program version of the ENIAC, an enhancement that was yet to be built.  I believe that this makes her the first modern software engineer – a person who’s job is to write code for a (stored-program) computer.

I got to know Jean Bartik in the last ten years.  My mother, Kay Mauchly, was also one of the original six women programmers for the ENIAC.  She and Jean had started getting together to give talks about their experiences at the dawn of the digital era, working alongside Pres Eckert, John Mauchly, and John von Neumann, among others.  After my mother died in 2006 we stayed in touch and did a few things socially.   Jean came to see my jazz band play an outdoor concert;  think about that – an 80 year old woman dragging a friend, a bottle of wine and a beach chair out to see live music.

My wife and I went to the Computer History Museum to celebrate with Jean and her family when she was inducted into the Hall of Fame.   It was gala event and Jean was in top form.  She is a great – no, I guess I have to say she was a great public speaker.  Jean was so engaging, so authentic.  A natural story teller, but they weren’t just stories – she lived every bit of it, and still remembered all the details of her long career in computing.  Linus Torvalds and Robert Metcalfe were also getting honored that night, but they had a hard time getting their share of the limelight in the presence of Jean’s natural Missouri charm.

Jean Bartik will be remembered in many ways.  Just a month ago I was proof-reading her autobiography, which, with luck, will be published this year by Northwest Missouri State University.  They have also established the Jean Jennings Bartik Computer Museum there in her honor.   She is, along with the other human “computers” and ENIAC programmers, part of a new documentary film  “Top Secret Rosies.” She is a role-model for women in technology.   Those of us who knew her will remember her genuine warmth coupled with razor sharp intelligence.  But I have to admit that I might just remember her wit above all.

Jean Bartik’s advice to women in the workforce:

“Look like a girl, act like a lady, and work like a dog.”

 

I guess the ENIAC was the first computer to calculate a lot of things, cure since it was the first computer to calculate.  But Pi has a special place around here.  Pocket-Lint did a nice write up about the whole thing.