Computer Pioneers: Pioneer Computers Part 1

ah hmm hello i'm gordon bell here at the computer museum in boston when i got into computing in the late 50s i met many of the pioneers and saw some of the first computers in action in fact the 1952 whirlwind behind me was the first computer i ever wrote a program for with this series i hope to give you a sense of what it was like to invent computing you'll see and hear the pioneers and their computers in vintage footage and in later day reminiscences i will be your guide in time travel from the dawn of computing in the 30s to its commercialization in the 50s in the 30s card calculating while in advance over pencil paper and mechanical calculators was labor intensive slow and unwieldy systems like this were used in the war effort including the design of the a-bomb at los alamos scientific problems were laboriously done on desk calculators took hours and days to do and were fraught with the errors of human input by 1937 five independent researchers were dreaming about computing machines four were driven by curiosity and frustration with the error-prone slowness of their own work they were interdriven to build a fast scientific calculating machine the first was conrad zusa in berlin while a student of civil engineering he started to work on the design of mechanical aids to stress calculations by 1936 zusa had a basic design for a computer whose operations were specified by a sequence of markings on a tape the same year in england alan turing's paper on computable numbers was published introducing the concept of a universal computing machine to an academic audience but none of the others were affected by his paper at harvard physicist howard aiken inspired by the designs of charles babbage was determined to build a machine to solve non-linear differential equations that he was encountering in his dissertation at iowa state university john adanasoft was consumed with building a machine to solve simultaneous algebraic equations for the solution of partial differential equations and finally at bell labs george stipitz was experimenting with the use of relays to build a reliable calculator for circuit analysis stibits was the first to get an operational machine his story is the btl mark 1. he tells us about this in a 1980 lecture at the computer museum there was little effective communication among the computer pioneers and each development with possibly one exception i think took off from uh very ancient beginnings in the case of the relay computers that i had a hand in the initiation was entirely a matter of curiosity about the way in which computers in which relay circuits could be combined and activated it was just a curiosity about the logical properties of the contacts which individually seemed so extremely simple and in collective masses could do some very remarkable things in the telephone business as i say this had no connection at all with the need for computers at that particular time i was a mathematical engineer at the bell laboratories everybody was an engineer then of some sort or else a girl but i noticed that the relay contact had only two stable conditions off and on and this i realized was true of the digits in binary notation and nowadays the binary notation is familiar to most youngsters and it's hard to realize that 1937 that i with a phd in applied mathematics had to stop and think back considerably to decide what a binary notation was i had a vague idea on looking it up it became familiar to me again and we decided to try it out as a means of employing of using the telephone relays and so late in november of 1937 i borrowed from the lab's scrap pile a couple of u-type relays and i assembled them as nearly like this as i can remember and it's designed to add two one bit numbers there is a and here's b if b is one the sum is one if a is one b zero the sum is one if a and b are both one we get a one and a i carry and a zero in the first bit i took it to bell labs and showed you some of my friends down there and they were satisfactorily amused by the idea that you could use uh binary notation from the old days to do arithmetic in the modern in the modern times well then i i wasn't entirely sure that it was so funny so i started to think about the possibility of doing a little more elaborate arithmetic than one plus one and i drew up some circuits at home for a calculator that would be something like a desktop calculator it's well known that you can easily connect electrical circuits with a soldering iron you don't have to do a lot of mechanical design to get them to work together and so i suppose i propose that maybe the relay circuits could be used to do the complex arithmetic so the project was started in september 1938 and construction began in june 1939 the design work uh assignment of relay types uh diffusing and all the multitude of details that have to be considered were taken care of by sam williams when i started to think about it decided against training operators to use the binary notation but i had considered the alternatives of converting all numbers into binary form inside the computer or converting decimal digits into binary form of the keys now because of the large amount of input and output we expected from this device we decided on the ladder that is the each key would convert its own digit decimal digit into binary form it was decided that when a meeting of the mathematical society was announced for september that we would demonstrate the machine and give a paper on the subject about why it was useful and how it worked a general i'd the original idea was that we would have a telephone line open between hanover where the meeting was to be held and new york where the machine was and someone at the meeting would call to the operator on the phone down in new york and ask her to put numbers in the computer and report back what it said and this scheme was very shortly modified so that the computer would up would respond directly to a teletype which is located in hanover and would send data back to that same telesite the model one the complex calculator was then studied by the group the bell laboratories and everyone was horrified to find that it had cost twenty thousand dollars including the development design construction and debugging well this is a number that simply wouldn't wasn't thinkable the idea that anybody should spend twenty thousand dollars for a mere calculator was something that was not acceptable and the labs decided that no more things of this sort would be built soon after the outbreak of world war ii the need for calculation prompted the government to fund the development of more machines at bell labs to give you a roadmap of each of these machines this graph shows their memory size versus speed do bell labs built six more relay machines before the end of the war the largest completed in 1946 had over 9 000 relays in several rooms consumed 20 kilowatts of power used floating point arithmetic and cost about 500 000 or roughly the equivalent of four million dollars today conrad zusa had many ideas that would lead to computing rather than just a better calculator his 1935 diagram of a universal calculating machine used to punch tape to feed information into a control unit that selected a memory cell and specified the arithmetic unit's operation he saw the need for large memories as a critical element of the computer because his interest came from civil engineering and performing stress calculations involving matrix algebra this is an and gate the inputs are on the left and on the bottom when they're both zero there's a zero on the output when the lower input becomes 1 there's still a 0. when the left input becomes a 1 we have a 1 on the output the z1 a 1938 prototype constructed in zeus's berlin apartment had a 64-word memory the z2 used relays for computation and was useful in convincing germany's experimental aircraft institute to fund the z3 the z3 was completed in late 1941 and operated until it was destroyed in world war ii a copy was made in 1963 for the deutsches museum from his 1941 patent application this model was still working in the early 1990s arithmetic was done in a polished postfix fashion by loading operands into its two registers and then specifying the operation floating point addition took about two seconds instructions were read from holes punched on used 35 millimeter film other data was read into and written from switches and lights in floating point decimal form thus the controller for doing the floating point arithmetic was non-trivial inventing the notation for designing the relay logic was key to the z3 zusa himself was very interested in the programming aspects of computation this talk is from a 1981 computer museum lecture surely at first i worked on paper but just right from the beginning i worked with a floating point principle i called it to that time the half logical form because at first i tried to make on computers which were working complete with but you know the difficulty is the addition yes i did not succeed to develop a good advice device for this from the mathematical point of view it's clear yes just from the beginning i was perfectly clear for the engineer only machine with floating point arithmetic will be sufficient and so from the first point i developed my whole mechanical machine with this principle and only that was the third one and the arithmetic part was working but it was not working so sufficiently that it could run day by day yes and to make it uh controlled by program and therefore decided to switch over to the relay construction and before i did make a whole computer in floating point arithmetic with relays i made a small test model that was a z2 that had 16 binary digits with fixed point i only had in mind to try how works this really technique at all i never had in mind that this this model could would work really really well when i began i had no knowledge of mathematical and symbolic logic and i developed for my own purposes so i called it conditional calculus i had my my own science for that my own representation i was was very proud proud to have found a new calculus ascended to my mathematical teacher of the school yes it was a very clever man and had contact with him and i sent the dependent i got a letter oh what you do is yeah that's a good old propositional calculus yes and at first surely i was to the point that i was bored i have found a nude or detected a new way but then as i thought now if i have what's the opinion that it was very good there existed already a well-proven and elaborated calculus for these things and i only had to translate the formulas of this calculus into switching algebra 936 or 37.

when i developed the switching algebra the machines i had made set three and set four may be the most sophisticated wired in logic which any machine i don't think that any machine built later on had so sophisticated built in conditions and so that i told you that i hesitated to take the line the wire from from right to left yes but i put all the logic in the circuitry of this arithmetic unit and when you solve a quadratic quadratic equation the cause of the calculation you can made in a making a string like package only on the end you have two values and the solution of the square root was perfectly wired in like division and multiplication all these things by surely including point arithmetic yes and so it was not necessary to have a machine with conditional orders to solve quadratic commissions surely it would have been possible to make conditional orders also to enlarge the machines four but i saw we were in a hurry the machine had to be ready where we had areas so there was not the time to change the principle of the machine when i would have gone to my uh the people who ordered the machine the ministry of the aircraft industry when i would have said that well i have a new idea i can make it change the the address and you must be ready next month when i began with the work and the 30 years i i constantly concentrated my ideas more on the relations between men and machine i saw that one time there will be that i didn't see any border between calculating and thinking surely to that time the computers we could make to that time or 20 years ago they were they were far away from being electronic brain today i hope you and my brains is ahead of the computers still today the z4 was constructed between 1942 and 45 and it was rebuilt in 1950. zusa also had to make a living and work for the air ministry lesser known are his fixed function calculators used between 1942 and 1944 for aircraft design that replaced 30 calculator operators by 1951 z4 had a 1000 word mechanical memory with one half second access time its logic was carried out with relays on its 32-bit floating point numbers it was modified in 1950 to perform conditional branching it remained in use until 1959 after the war zusa started a computer company that built a series of machines before it was ultimately absorbed by siemens john v adanassoff was perhaps the first to conceive of an electronic calculator using serial binary arithmetic in january 1940 adam nassoff and his research assistant clifford berry built a prototype to demonstrate serial binary arithmetic and the use of a capacitor store the attinasoft berry computer or abc was designed to solve 30 simultaneous linear equations to enable the solution of partial differential equations capacitors were mounted inside a drum with contacts that were used to read each bit serially as it rotated punch cards stored intermediate results adam nassoff also told his story at the computer museum in the fall of 1981. my mother had a large arithmetic which she studied in eighth grade it contained many other topics and those usually included in what you call arithmetic and what i would call arithmetic but it had a chapter on numbers with other bases than 10.

and i studied this with great care in partial differential equations you can't make much progress but trying to solve them you can't solve them in any ordinary sense but all you could do is to approximate the answer there are too many there are too many uh uh functions necessary to solve those equations they're all over the map things you've never heard of before and no one else has ever heard of before and you just can't uh get out the solution that way this is what actuated me really more than any other problem to get interested in the computing art there i was in 1936 turning my mind invent a new digital computer not knowing how it would be built or how it would work however i believe that it should be fast accurate and flexible of what stuff would my digits be constructed see we're going to use some quantity to represent digits and how what stuff would that be and would it be mechanical electrical or some other and what would the structure be and the only thing i could think of was to separate memory and computing and maybe it wasn't a bad idea and now i thought does bass play a role if we're building a digital computer what are we going to do about the bass and what determines this how should i do computing uh what was the computer going to be like and how should memory be constructed i had all these words at that time and uh for base two numbers i thought of a two-step mechanical system fair magnetic material going back to the wire you know and i thought that was an awful good idea in the end i didn't use it a two-state vacuum tube circuit and a capacity with two states of charge you should realize that these concepts represent a hodgepodge of ideas i tried again and again to sort them out nothing seemed to work after months of work and study i went to the office again one evening but it looked as if nothing would happen i was extremely distraught and then i did something that i did in those days but i've had to stop lately i got an automobile and started to drive i drove hard so i would have to give my attention to driving and i wouldn't have to worry about my problems i drove towards the east i'm driving a ford ford v8 and with a south wind heater i don't suppose you know what a south wind heater is uh pretty warm but the night was very cold it was the middle of the winter in 1937 and 38. and when i finally came to earth i was crossing the mississippi river 189 miles from my desk you couldn't get a drink in iowa in those days but i was crossing into illinois i looked ahead and there was light of course it was it and i stopped and got out and went in and hung up my coat i remember that coat and sat out the desk and got a drink and then i noticed that my mind was very clear and sharp and i knew what i wanted to think about and i went right to work on it and worked for three hours and then got in my car and drove slowly back to ames and i had made four decisions in that evening in the illinois roadhouse use electricity and electronics that meant of course vacuum tubes in those days use base two in spite of custom for economy use condensers but regenerate to avoid lapses compute by direct action not by enumeration you know the it's an easy thing if you if you know about number system to just have things that count a kind of an internal counter that gives you enumeration but i was going to do it but direct action the next thing is the adanas ferry computer this represents our main effort in computer construction it was designed specifically to solve the problem that bugged us the solution of linear algebraic equations we're building this machine to solve 30 30 equations and 30 unknowns to work on determinants to 30 places and we're going to use simultaneously 30 abstract mechanisms and june 13th 1941 dr john markley visited abac installations at the invitation of dr john villarrenasoft and he was shown the complete enterprise might tell you how things went about that computer the ohio state college staff members were not unduly enthusiastic corporations were not impressed and ibm said they'd never build such a machine since the beginning many men have worked on computing and many have furnished elements that were important this applies to markley eckert some members of the staff such as burks greater than others historically the list can should contain pascal napier and zuza credit must also go to the originators of our number system which began computing what each man accomplishes depends on his brains and energy but also on the surrounds in which he works in this timing is important pascal pascal could not have invented electronic digital computing although his brain brainpower was overwhelming in a larger sense no man invents anything he builds and extends a little with his friends and on the shoulders of others many are quietly involved in making this historical record clear and we should all support this effort when at nassau left iowa in 1942 to become part of the war effort the abc still lacked a reliable card punch it was not in service the parts were cannibalized and the only remnant is the drum at nassau's greatest contribution may have been helping break the broad eniac patent filed in 1964. he was clearly the first american to use binary arithmetic and vacuum tubes for direct digital computation and to describe and prototype ideas such as non-restoring divide a concept the eniac patent also claimed the last stories of this era involve thomas watson seniors funding of the harvard mark 1 and the ssec at columbia's watson laboratory the inventor of the mark 1 was howard aiken aiken was motivated to build a scientific calculator to evaluate integrals like those he was encountering as a phd candidate he alone of these pioneers was inspired by reading the works of charles babbage and his plan for an analytic engine by 1937 he had specified the architecture of a machine that he often referred to as a computing engine it was the imagination and drive of aitkin that created the ascc for automatic sequence controlled calculator or harvard mark 1 marking ibm's entry into computation that was not controlled by a plug board ascc was engineered built and tested by ibm operational in january 1943 it was moved to harvard and dedicated on august 7 1944 unfortunately howard aiken died in 1973 and never told his story for the camera but rear admiral grace hopper his co-author of the ascc manual vividly describes life working on the harvard mark 1.

i think we've totally forgotten the environment in which mark 1 appeared and what a difference it made in the reaction to her and the problem she ran 41 was pearl harbor by 43 we were in the thicker things i had long wanted to join the navy as well but i was in a classified occupation i was a college professor teaching mathematics at that time the teaching of mathematics was a classified occupation however i finally got permission to leave vasser and then i faced a second obstacle i was underweight for some unknown reason to this very day the navy considers that i should weigh 140 pounds all these years i've been trying to explain to the navy that um i had scott's ancestry and i'm lean and tough and i don't need to weigh 140 pounds i weighed 105 pounds my orders sent me to the um bureau ships computation project at harvard this was on a friday and i was supposed to report on monday morning so i had saturday and sunday off with my family and then i came to boston and then i started to find a bureau of computation project well i finally finally found the office of naval research they'd heard of it but they weren't quite sure where it was so about finally about two o'clock in the afternoon i found a bureau ships computation project in the basement of craft laboratory at harvard i walked in the door and commander aiken looked up at me and said where the hell have you been he then waved his hand at mark 1 all 51 feet over and he said that's a computing engine i think when he called it a computing engine what he meant was it was made up of different parts that performed different functions he informed me that he would be delighted to have the interpolation coefficients for the art tangent by next thursday well by this time i'd have been introduced to the two programmers who were ahead of me but it was a small group and a very big machine because mark one was 51 feet long eight feet high and eight feet deep and she was in a magnificent glass case eventually designed by nominal gettys which of course made it better than ever if you go back to that time and to the need for computation she begins to take her place in history in the first place she was the first large-scale digital computer in the united states she was automatically sequenced she was programmed step-by-step program exactly as we have today everybody talks about eniac because it was the first electronic computer but anyak was not programmed he used patch cords and put that together in in a simulation of the problem you wanted to solve it was not programmed when it was built so from that point of view mark one more closely resembled what we have today than any of those other early machines because it was sequentially programmed step by step one operation after another there were a lot of adventures with mark 1 some of which have become slightly historical at the same time we were running mark one we were starting to build mark ii for the navy and mark ii was built entirely out of relays in an awful rush because um the wall was on and um we were building mark 2 the summer of 1945 and naturally since it was world war ii we were working in a world war one temporary building the air conditioning wasn't very good the screens weren't very good and the windows were all open mark two stopped they finally found the failing relay and inside the failing relay bitten to death by the relay contacts was a moth about this big so the operator got a pair of tweezers and very carefully fished them off out of the relay put it in the log book put scotch tape over it and below it you wrote first actual bug found one of the reasons we so greatly needed computation is another thing most people have forgotten world war ii talk saw an almost complete change in our weapon systems up to that time when you had a mine the mine had ears sticking out of it and the ship had to hit the ears and break a connection before the mine went off all of we started suddenly started building acoustic mines listen for ships magnetic mines that waited for the heat to sense the metal of the ships well now we had to know how how big was the effect of the mine how much area was it going to detect ships so all the computations had to be gone over had to be created to tell us how much space around a mine with this mine effect so we know how close to sow the mines where to position them so there was a tremendous amount of computation to be done including the a-bomb and all the other things that happened during world war ii and the pressure for that computation was very great everything was hurry up do it yesterday so the pressure to keep the only computers we had running was very very strong and mark one ran 24 hours a day seven days a week which was a very rough task for a small crew mark one had one set of storage which was a bunch of switches it was a big panel now the mark one word consisted of 24 digits a sign in 23 decimal digits so across the panel were 24 switches and you could turn them to the different numeric positions and put constants into a problem by putting setting them up in the switches the instruction was essentially a single address instruction except i've never seen anybody call it that because we wrote in three columns one column was the out column one was the in column and the c column was the action column for instance a multiply instruction would be four three two out of counter four three two seven six one into the multiplier and seven to go on to the next thing or whatever and the in column was essentially the operation to be performed so it was really a single address code except that you named the quantity first then the operation and the third column was the one that for instance would tell a print counter to go ahead and print on a punch card encounter to hurry up and punch that card and so on various action codes were in the third column along with the automatic which moved it to the next step which was the seven so were the 72 of those storage registers each one of which was an adder the next important unit was probably the multiply divide unit the multiplicand went in first and then in the registers were built up the nine multiples of that quantity the multiplier came in and it called down the multiplicand the multiples of the multiplicand one after another and shifted each time and multiplied just exactly the way we do when we try to do long distance bulk multiplication so a very interesting thing started developing which led to other things later we began to have pieces of coding in our notebooks for instance stick had a nice little sign program for annuals and angles less than pi over 4 positive angles and when i needed it for another program i'd borrow it from his notebook and copy it into another program we didn't realize it but we developed subroutines and the whole future of beginning to get the compilers and the various high-level languages came from those things in the notebooks and both bob and dick had notebooks full of those pieces of code and believe me i borrowed them very frequently i also learned learned another lesson at that time which later forced me into the development of the compilers and that was that programmers can't copy things and that when you integrate a piece of code into another program you frequently have to add to all the addresses and programmers can't add either i don't know whether i've made it clear but the data and the sequencing were totally independent throughout the development of mark one and mark two and mark iii our decan always insisted that the data and the program must be independent and stored independently he was a tough task master i was sitting at my desk one day and he came up beside me and i got on my feet real fast and he said you're going to write a book i said i can't write a book he said you're in the navy now and so i wrote a book i have it here with me so that i can manage to answer any questions this is the mark 1 manual the entire bible of mark 1. it contains every circuit samples of all kind of programming coding every has an excellent bibliography in it on computation as timing charts for all the operations in the various circuits all the circuit diagrams are in here you could take this and build mark one over again if somebody felt like it but many of the things that we think of today common ordinary or forgotten where they came from arose in the development of the first programs that were written for mark 1. both ascc and colombia's much faster electronic ssc stand out because they have large memories like the z3 they were programmed from a sequence of instructions stored on tape ssec was a prototype for ibm's production model card program calculators these machines led to important patents for ibm because they could perform arithmetic on and then execute stored instructions the ibm engineers working on the sscc went on to build the highly successful ibm 650 the ssec programmers became members of the ibm 701 programming group but that's another story here's herb grosh one of the programmers of the ssec heroes of this story are primarily watson senior and wallace eckert this is tom watson senior as i first met him he furnished the money he furnished a large part of the inspiration for the activity through his support for what he thought of as scientific research what we would today call i think applied research this is the watson lab as we saw it in 1945 and 46 when we were moving in it was a former fraternity house which ibm remodeled at great expense and donated to columbia university and it was in those strange quarters that the romantic incidents that i'm going to relate to you occurred i came into this thing because i was trained as an astronomer my phd is in celestial mechanics and very early in my career as a graduate student i met many of the figures who were interested in doing this sort of thing somewhat more economically than the old-fashioned logarithm multiplication table hand cranked brunswick that had been used for years one of the chains of applications that made the watson laboratory possible was the motion of the moon among the stars in the early days we did not have chronometers of any accuracy in order to determine one's position in crossing the pacific you need to measure your longitude as well as your latitude without a chronometer you have to have some other kind of clock and you can't do it believe me with water clocks or or sand hourglasses or what have you then the war threatened the u.