This interview was provided courtesy of the Computer History Museum.
Booch: You were starting to see computers around you. You had, you know, been involved in some of the analog computing, saw that kind of dissipating. What was the first computer that you had your hands on?
Humphrey: Well, let me move back to the story a little bit, how I got there, if you don't mind.
Humphrey: I decided to leave
Shortly after I got
Mrs. Humphrey: Maurice? First name is Maurice.
Humphrey: Maurice -- is it Wilkes?
Booch: Oh, Maurice Wilkes.
Humphrey: Yeah, Maurice
Wilkes. He was one of the professors teaching the course. So I walked into the
course and there registering everybody was this pretty young lady, and so I
spent my two weeks trying to figure out how I was going to meet this lady. Well,
I did chat with her that day, during a break, and it turned out she was from
Booch: I think it will stick. And was that her voice I heard in the background a moment ago?
Booch: Was that her voice I heard in the background a moment ago?
Humphrey: That's right, telling me Wilkes.
Booch: Well, congratulations. Wow, like 55 years you say, that's remarkable. Congratulations.
Humphrey: Yeah, I learned an awful lot in that course. I wrote a program for the Whirlwind and I got totally committed to computers and to Barbara. She had been a math minor and English major, and they wanted somebody who could help with writing and presentations. The scientists were close to illiterate, but not quite. She was helping them. But she got me a copy of -- was it Herman Goldstein and John von Neumann, if you've ever heard of it. Anyway, it was a book that they wrote on planning and programming computers.
When I started at
I decided, instead of trying to tell these guys what to do—since I didn't know what they should do-- I spent my time asking questions. I was quite frank with everyone. If I didn't know something, I'd say, "Okay, why does that work that way?" And so I treated my job as a learning experience. What surprised me was how much the engineers loved it. They loved to talk about their work, they loved to explain what they were doing, and they didn't look down on me for a minute, even when they knew something I didn't. It didn't bother them, they were perfectly happy. They were proud that they could explain something to me. I learned more in that brief period than I ever did in college. And the team was just extraordinary. We got patents on some circuits and produced some great products. I learned a lot real quickly and was soon able to start helping the engineers when they had design problems.
I have followed this same management style ever since. It turns out that, when you manage thousands of people, you're going to have lots of people that know more than you do. With this style, you can work with people who know more than you do. If you can't do that, your future's terribly limited. I was enormously fortunate to learn that on my first real job in industry.
We had a great team. A
few months after I got to
So when I became program manager, they came back to me and said, "We've got to get the power requirements." I said, "We haven't designed the circuits yet. How can we give you the power requirements?" They said, "Well, if we're going to meet the schedule, we've got to get all the structural frames ordered and, to do that, we've got to order the transformers." So I said, "Okay, well, how long does that take?" Well, it turns out that if we didn't get the transformer order in next week, we couldn't make schedule. The transformer was a long-lead item. I said, "Oops." So we made a power estimate right then. We were very cautious and put a fair amount of fudge on it. We actually had enough, which was fortunate. At the same time, I said "Let's make a detailed schedule for the whole program."
We got with the
manufacturing guys and laid out the plans and put the whole schedule together.
We found that we barely had time to get the job done. Later, in the middle of
testing the hardware, the circuits were working and everything was great. We
were preparing for the final testing and for the Signal Corps to come up and
review the system in the next week or so. I was home one weekend and I got a
call from the guards; there was a hurricane coming. In those days we didn't
have hurricane warnings, and we were in a basement lab in downtown
Booch: Wow, that's remarkable.
Humphrey: Isn't it? Amazing.
Booch: So you were doing cryptanalysis work around then. I'm curious if you had any connections with Turing or had heard of the Colossus around that time? Was that on your radar?
Humphrey: No, that wasn't, but
when I got to
Humphrey: I was quite used to teaching because the professor I had at IIT, George Cohen, was wonderful. I took every course he taught in electrical engineering. On occasion, he would call me up and say, "Hey, look, I can't be there tonight, will you give the next lecture?" And so I generally arrived at class prepared to teach. I was working at the Chicago Midway lab in the day and taking these courses at night. I was used to teaching, even stuff that I didn't know. Since I had been able to do that, I decided that I could probably teach a course on computers.
I went over to the MIT library and the Harvard library. I had previously talked to Howard Aiken about taking a PhD at Harvard but he wanted me to take a lot of formal math that wasn’t the least bit interesting. So in preparing to teach this course at Northeastern on computer design, I read the von Neumann book, and I went to the libraries of Harvard and MIT and read everything written on computers. And there wasn't a whole lot. It is amazing that I managed to fall into the computer field at a time when you could read all the literature on a subject. Can you imagine doing that today?
Booch: It'd be tough.
Humphrey: Yeah. So I read
everything about it, everything they had. I got the Kiester, Ritchie and
Washburn book by the Bell Labs folks, which was about switching design. It was
a very good book, but it wasn't about computers. It was designing, essentially,
telephone switching exchanges. It had a lot of useful stuff in it. I put this
two-semester course together, and the first class started on the 13th of
September, in 1954. I had 13 students, and that day we moved into a house in
Booch: Marvelous. Those 13 students, have you had any further contact with them? Do you know where they've ended up?
Humphrey: I have not. Most of them worked for Honeywell and they were all designing the Datamatic computer. They probably knew more about computers than I did. But I followed the same kind of teaching strategy that I had used in management. I would have them explain stuff to me. They knew a lot that I didn't, and so we had a marvelous class. I learned a lot from that course. I then went through my notes and I put them together into a framework of the course the next year.
I had very detailed lecture notes and, at the end of the year, I decided to see if I could turn it into a book. So I contacted McGraw-Hill to see if they'd be interested, and they were. So I wrote my first book. The second year I taught the course, it wasn't as much fun because I knew so much more at this point than I had the year before, and the students treated me more like a professor than a colleague. I had notes for my manuscript, which I was teaching from now, and the third year I taught, I had the published textbook. I taught that for four years.
Booch: What was the title of that book, do you recall?
Humphrey: Switching Circuits with Computer Applications.
Booch: So mostly what you taught -- I would imagine that's out of print these days, I don't know.
Humphrey: It's out of print,
but it was used as the text at MIT and Purdue and some places in
Booch: In fact, thanks to the power of Google, I see that one can still buy some copies, so it's available.
Humphrey: Wow, it's still available.
Booch: So most of your course was certainly not programming it sounds like, but more on the machine side, the architecture of the computer. Would that be a fair characterization?
Humphrey: I had not done any programming, except what I'd done at that MIT course.
Booch: Got it.
Humphrey: And so I wasn't really a programmer at that point.
Booch: In your comings and goings, did you run across von Neumann or Eckert and Mauchly as well, too? Where were they in the midst of this?
Humphrey: I didn't run into
any of them. In fact, the people that I ran into at the
So I started to power it up. I didn't get very many of the units up before it blew a fuse, and so all their guys came over to see what was going on, and the building maintenance guy who was running all of that stuff said, "Well, we'll put in a little bit bigger fuse." And so they put in a little bit bigger fuse, and I went a little bit farther this time and blew that out. And the electrical engineers who were putting this stuff in, they came in and they kind of laughed and they said, "He said to put in a bigger fuse so we did." And I said, "Oops." And he said, "Okay fire it up." So I fired it up, and I got most of it up, not all of it, but most of it up, and it sounded like a bolt of lightning. I mean, when it blew that fuse, it really -- it blew the fuse box. The electrical guys came over and said, "Come over and see this, you've got to see this." So I went over with them. The whole wall was black. It had taken out the whole fuse box. I mean, this was a lot of power that I was drawing with this thing. So in any event, they got it all fixed ultimately, it took them a little while. And so we got it all back up, and I had been trying to figure out how we were going to test all this. This had been somewhat earlier. We were due to get a demo for the Signal Corps and for some other gentlemen, who arrived from an unnamed agency. And the question is, when you've got this random digital data stream, how do you determine that it's working. I mean, the circuits were working, but do you know if the logic works?
So I worked through the logic of this and figured out how I could get it by connecting the input to the output of the crypt box, how I could get it actually to fall into a pattern and so I did. I was able to work out a bunch of digital standing-wave patterns, and we used them for testing the system. When the Navy Signal Corps came to get a demo, I probably described this system for getting this to fall into these synchronized patterns, to self-synch, and demonstrated it. The Signal Corps guys were all excited. And these three men, in black suits, didn't say a word. And I realized later I had killed the system right then. I mean the crypto guys weren't the least bit interested in a system you could get to synchronize like that. It would have been tough to test if I hadn't. But in any event, the first system, we actually built it, shipped it off, and it was installed, apparently, in the Pentagon, linked to some unnamed location. I learned later than it ran for 5,000 hours unattended without an error.
Humphrey: For 5,000 tubes, that's pretty good.
Booch: That's pretty amazing.
Humphrey: Yeah. So those
circuit design kids -- they were all young, they did an extraordinary job. We
had it derated so you could bring the power down 50 percent, and you could do
all kinds of stuff with it. It would run from minus 30F to plus 130F in high
humidity. It was an amazing system and it just worked forever. So it was a
tragedy it didn't continue to get built. And at that time, I was running this
group now, I had several projects at the Waltham Lab at