Bonus- Gordon Bell - Building Blocks of Computing

Gordon Bell has had a long and varied career. An early employee of Digital Equipment Corporation, designer of many of DEC’s PDP series of minicomputers, leader in computer science at the National Science Foundation, long-time proponent of the Computer History Museum in Mountain View, California, and the creator and funder of the Gordon Bell prize, which annually the ACM awards for improvements to parallel and scalable computing. Today he’s a researcher emeritus at Microsoft. I recently met with Bell at the Computer History Museum in front of a fully restored and operational PDP-1. Well the PDP-1 is the first computer that I really did engineering on, maybe it’s the second. Because I got the Digital from MIT. where they had a machine called the TX-0, which was the forerunner, in fact maybe the prototype, for the PDP-1, that was designed at Lincoln Laboratory. First transis, one of the first transistorized computers. And I had built a tape control unit for this. What I needed was all these modules, flip-flops, AND gates, OR gates, all the component things that are so low-level to a computer scientist. These are these are, you know, tiny areas of a piece of silicon. Those were the components, and so I went to Digital, who was selling components, to buy those. And so I bought, I don’t know, 50 of them and put them together to make a tape control unit. At that time I was starting down the Ph.D route. I was a staff engineer doing speech research and so we needed the tape unit just to get more speech and more I/O, but what I did was a speech unit analysis system called Analysis by Synthesis, which is still, the paper by the way was the first one of these and still referred to because it’s a technique for, you know, now you’d call it an AI technique. It was a way of understanding what was going on in the vocal tract, and, anyway, what I learned from that was that basically I didn’t want to be a researcher. That I really was an engineer, and certainly above all, I didn’t want be a speech researcher, because I said, you know, it was, I, I’ve experienced this with two or three speech researchers since that time. One of them was telling, describing this, as an AARPA contractor he was going to conquer speech, and anybody who prize, looks at speech and you say my god, I can kind of recognize that that, why I should be able to write a program. And so I basically said okay, where, I’ll, where do you want us to send the Nobel Prize when you do that? And there was just, and so I got out. I said it was going to take 20 years to do it, and I was wrong by 40 I was wrong. It took about, actually we did make progress by 80s, by early 80s there was some voice recognition. Some early generation PCs had some voice recognition. Yeah, in about that time, but to get to the point where we are now is taking. So you were buying components We were buying. How big was a flip flop? Was it like this big? A flip flop, actually a flip flop was on a board about this big. Or, I think, depending on the speed and what it would do. But, I think we could get maybe two to four flip flops. If they were shift registers you could get more, but if they were, you know, JK, RS, the full bore, you know, maybe just get one of them. And those are running at, the DEC modules at that time were running at at five megahertz. So these are all five megahertz, I enjoyed the, the building of the program, to do analysis by synthesis, the building of the hardware, making all of that work, and then the building of the tape unit. And, but I had really wanted to do the engineering aspect and so I’ve basically spent my life mostly thinking as an engineer. In fact, I’d say over the last 10 years I’ve been an accidental researcher. I did, you know, I’ve been responsible for, for managing research in a, you know, overall I was head of R and D at Digital Equipment Company. But fundamentally I, my aspirations and my activities were all directed to getting, to building things in volume that people would want to use. So tell us a little bit about this PDP-1 and how its building sort of triggered, I mean was it, did it end up triggering things that you didn’t intend? I think any time you build a new component, a new component is introduced, it opens up totally new doors. The interesting thing about the PDP-1 as the, as I’d say it’s a forerunner for the minicomputer and we could go look at what a minicomputer. What I define a minicomputer is. This is 1960. By 1965, we built a thing that was, I would, I, that I claim is the first classical minicomputer, and the reason it’s that is although this was a component, was used as a component in some sense, the classical mini was actually built as a component. It was to be used for something else. And I’d say the big, big thing you should take away from that, this is, that in every generation, you get a machine, a computer that is a component. The microprocessor, as it was first came out, was a component. Once it got, the microprocessor got a little bit bigger, the PC, that began to be used as a component. In fact, what we just talked about was this guy here. The cell phone is now a component. This is the most amazing component you could ever get. You know, we would never have thought of that as a component. Because what’s in it is, gee, I’ve got, maybe 32, 64 gigabytes of secondary storage that’ll store everything. It’s got enormous processing power. It’s got miniradios in it. . And it’s got an accelerometer. It’s got a, a GPS unit and guess what? Who wants one of those? Well the, all of these nCopter things, the, the Geocopter three, the three-blade thing, the four-blade, the six-blade. There’s a, I saw a 12-blade that carries people the other day. All of those things are controlled by this being the central part of that. In a sense, when you first build something like that one I did. You need to keep it flexible. Yeah. Because you don’t know what exactly its purpose is. Absolutely, this. The beautiful thing that, the fun we had with the PDP-1 and I can, could go back and open the door there and show you, was the I/O system. The thing that, in fact the first documentation I did on, about the PDP-1 was I wrote a book, not a book, a 32-page manual, about how to connect stuff to the PDP-1 because I, that we were enamored with. How do you connect to that? And then the difference between the PDP-1 and then subsequent machines that we built were really, the I/O got much more flexible, it got easier to use. So that when somebody could read that manual they’d say, oh my god, I’m going to connect that to a process control, I’m going to connect it and make it an oscilloscope, I’m going to make it a pulse height analyzer. And so in a sense, what the minicomputer did it enabled the computer to be used as a component in sort of everything, and that’s where it was. And you know, I just wrote an article about the birth and death, birth and, I guess, I don’t remember, rise and fall, rise and fall of the minicomputer and basically the editor, you know I went back and forth with the editors but what happened was I had to, to say look, the function that we were creating here had not, there was nothing like it in the past. It wasn’t, wasn’t a scaled-down mainframe that you could interact with. That wasn’t the goal that we had. It wasn’t a, so it wasn’t a record keeper. It wasn’t a computer in the sense of a supercomputer. We, we weren’t after numbers, and we weren’t after bits and keeping records. We were after a component that could be used in any number of ways and it ended up being used to do message switching that have telegraph lines coming in and, and telegraph lines going out, which is the core part of, of networking today. The Internet, it all works on having computers that have bits coming in and out and moving them around. And it was also the ability to have analog sensing information from some of the first things, we were sensing body information. So we were sensing things that people were carrying on them. I normally wear a strap here that has, senses my my heatflux and skin resistivity and stuff. A thing called a BodyMedia. Right. So that. A body bug. The underlying thing is these are flexible systems. To be hacked in ways that you couldn’t anticipate. That’s right. And the idea was to make these systems as flexible as possible, and so that they could be interconnected to, and so. A toy chest for engineers, basically. Yeah. So basically, I. The thing that’s been the most, I’d say the, I’ve had the most fun with is sort of thinking about buses and connecting things over my lifetime. So, in this one we had a, a core, you know, processor memory, and some I/O, and then wires radiated to connect to other things. And then made a, and actually, the PDP-5 which was this, I’d say the forerunner of the 8, was basically a single wire that ran out, and you attached things to that. And then when we made the PDP-11 we had a single wire for everything, into all all memories and processors and everything connected to that, to that wire. And by that way, when we introduced Ethernet, because Ethernet, the first Ethernet had was a single wire and you connected computers to that wire. And I said, basically at that time I said the Ethernet is the unibus of the 80s. In fact, that was and so it was a two-and-a-half kilometer wire and, and you ran that wire all around and you just connected things to that wire and, and did your, and, and made this big system. Talk about how this became perhaps the first video game, would you say it is? Yeah, well, the interesting thing about that as being the first video game was, Harlan Anderson, one of the founders I think, had discovered at the end of the year that he could give a computer away and get total tax deduction for that and I think it, I don’t know that, whether that was the thing that made us profitable that year or not. But in fact we were, we did, we gave. I don’t know, a number, I don’t know what number it was, but a very early PDP-1 to MIT. And so right next door to the TX-0, the transistorized computer’s ancestor, we put a PDP-1 in that, that next door room. And, and all of the students came in there. Steve Russel and Peter Sampson, Allen Kotok came in and took a lot, took a lot of the software ideas that had been developed for PDP-1, and transformed them onto the that was on the TX-0, put them on PDP-1, and that, things like all the first editor, which we called Expensive Typewriter, debugging, interactive debugging, all the light pen kinds of interactions, connections, and then the, I think the most famous one that is still in a sense in use is, was Spacewar. And Spacewar has been used for years to arbitrate different legal aspects of, you know, when did you, you know, a lot of people say, I invented this by that time, and you generally will find that’s, find the, some function like that in Spacewar.