Preface by Lucy Suchman, 28 February 2021
In 1980, at Xerox’s Palo Alto Research Center (PARC), we entered into a rather extraordinary conversation. Computer architect and electrical engineer Lynn was then head of the LSI (Large-Scale Integration) Systems area at PARC, while I was a recently arrived Research Intern at PARC and PhD student in Anthropology at the University of California at Berkeley, in the very early stages of formulating plans for my doctoral thesis. Lynn was emerging from a period of intensive activity focused on the ‘multi-university, multi-project chip-design demonstration’ (MPC79), an initiative involving the creation of a new pedagogy and associated design and manufacturing process for the production of very large-scale integrated circuits (VLSI). At Lynn’s suggestion, we began an exchange born out of her desire to think anthropologically about what she had just experienced, and mine to deepen my understanding of her sociotechnical imaginary and practice.
With Lynn’s endorsement, I recorded and transcribed our conversations, aided by an Alto computer and the text editor Bravo, both recently developed and in everyday use at PARC. The resulting text, which I printed out to share with Lynn and archived for herself in a 3-ring binder, was around 55 pages. We lost touch over the years following our respective departures from PARC until, in 2020, another PhD student, Philipp Sander, contacted Lynn to interview her. Engaged in doctoral work in Media and Cultural Studies at Leuphana University in Germany, Philipp was interested in understanding the place of Lynn’s VLSI work in the history of the microchip, and in the course of their exchange she pointed him to me as someone else he might talk with. Philipp’s consequent email sent me down to my basement to retrieve the binder and re-read the transcript. More importantly, it reconnected me with Lynn, and our conversation resumed with as much enthusiasm as it had begun 40 years earlier.
What follows is the original transcript of our conversations, preceded by a cover memo written at the time. Our conversations about that conversation continue (in part from comments on the transcript) as we, in Lynn’s words, grope our way into ever-clearer mutual understandings of the events surrounding that earlier encounter. As Lynn points out, rather than serving as any ‘real record’ of what we were actually thinking at the time, the transcript only just cracks open the door to shed light on what was going on in each of our worlds. For example, we've already noticed that rather profound social, political, and interpersonal forces were at play that were not only not discussed but were quite carefully kept between the lines of our text.1 It is those threads of the subtext that we’re now most keen to recover, to trace their effects and their wider resonance with other (his)stories of technoscientific research and development.
XEROX Palo Alto Research Center, Systems Science Laboratory
To: Conversations File
From: Lucy Suchman
Subject: Suchman-Conway Conversations
This file documents an ongoing project in ‘cross-cultural communication.’ It began with my interest in the sociological implications of Lynn Conway’s work in the field of VLSI. As an anthropologist, I shared Lynn's fascination with the issues of design, network collaboration and technical demystification raised by the MPC79 experience. Those issues were the focus of our first meeting, which we recorded and transcribed.
We found that the transcript provided an interesting artifact. It had several properties. First, by externalizing Lynn’s knowledge it immediately became a document of MPC79 and could be read as history. It could also, therefore, be read as a text, with all of the opportunities for finding topics and developing lines of discussion that a text provides. And finally, it provided a record of our first attempt at a dialogue.
More precisely, it suggested the terms for such a dialogue. We found that the issues of specialization developing out of the story of MPC79 had to do with just the work that we were ourselves engaged in, i.e., finding topics and concepts of mutual interest and usefulness across the boundaries of our respective fields. At the same time that the transcript talked about that project as a topic, it served as resource for carrying it out.
Our subsequent meetings have taken a more free-running and conversational form. Each time, however, we’ve tried to maintain some continuity with the one before. In that effort the transcripts have been very useful. A recurring topic of our talks has been how to find a framework for communication across disciplines. The object is to find the level at which specialized knowledge from one field can be presented to a novice from another so as to disclose its underlying simplicity, while not losing the depth and complexity that makes it genuinely special.
These discussions are not intended to be tightly ordered. At times they may wander, loop, or become downright silly. With that caveat, we invite you to read them in the same spirit of relaxed exploration by which they were generated. The hope is that whatever your field, there will be resonance at some point along the way.
c. Copyright 1980 by Lucy Suchman
Conversation I (MPC79), March 12, 1980, Filed on Maxc<Suchman>Conl.press
Conversation with Lynn Conway re MPC79
(S = Suchman, C = Conway)
S: The things that have particularly interested me about it, other than just the sequence of events – there are two major topics that I think are really interesting. One of them is the whole issue of the “network adventure,” as you call it. That kind of open collaboration.
S: An open collaboration that’s facilitated by having that electronic medium to communicate with. And the second thing is the demystification of what was traditionally a very specialized field of endeavor.
S: So maybe you could just go over the sequence of events, the way things happened. And then however those topics come into it – and whatever other topics you think are really important that those don’t include.
C: And while we're doing that why don’t you try to keep me on track or point the thing off in directions as they seem interesting.
C: The interesting thing, I think, is that the two things – the use of the network and the network adventure aspect has been coupled in a close way with the demystification. I’ll try to see if I can indicate how that happens.
A lot of this is intuitive. I think that what happens is that, how would I say it, the people involved in this have been – and when I say “this” I really mean the creation of the early course material, the work on the book,2 the simplification and streamlining of techniques of implementing designs once you have them, and then lately, this new system for collecting many designs and so forth, and sort of spreading that culture – I would say there’s one quality about the people involved. And that is that – I perceive this – that myself, Carver Mead, Doug Fairbairn, Dick Lyon, a bunch of Carver’s students and some of the others we’ve interacted with since very early, have all been sort of intellectual entrepreneurs. They’re interested in just making things happen, and having fun, and basically are a bunch of people that found themselves suddenly with a set of possible things to get going, and access to this network. And then, you know, just started operating. Operators might be a good word. Operators in the financial sense or business sense. Okay?
C: Which meant they would take risks, and just try out crazy things. Okay. But then also I think that some of us had an intuitive notion about how to use the networks to do some things. Being able to envision some mechanisms you might be able to get your hands on that would make things happen faster and go certain ways.
S: Does that, do you think, come from experience? Just having used the network for a long period of time for other –
C: Yes, yes. That’s a very good point. What I think happened was, a sort of critical point was, something triggered our starting to operate in the Arpanet just like we operate here on the second floor [of PARC].3 So it was probably that prior to all this most of us had been embedded here on the second floor in a variety of projects not really related to this, doing our own thing here, and had learned a style of joining, or forming and leading quick tasks pulled together where a bunch of people will go off and build a program, write a paper, do a thing in the network here on the second floor.
When this thing happened there was another counter pressure. There wasn’t much interest in this on the second floor. I hadn’t thought about that, that’s sort of old history. In fact the effort was sort of scoffed at as naive. It’s interesting. So then there was this reaching out looking for collaborators outside.
S: How long ago was that?
C: It first formed in – in fact maybe it would be interesting to go through just some of the chronology.
S: Yea, that would be great
C: This group was formed, really formed I guess as a group, in early 77. But the actual activity started in early 76. And right at that time a computer science department was forming at Caltech, under Ivan Sutherland, and this group formed here. And it was really just myself and Doug Fairbairn and then six months later, I think, Dick Lyon joined us. So initially it was just a very small group. And we were interacting with Carver Mead. We were learning how to do design in a way that Carver was teaching it in some courses at Caltech. We were computer architects and became very fascinated with this because it was clear that we could learn to do it and do it. And also that the way it was being done was quite simple. And we could see how to point towards more simplification. We were actually building our own by going out in the valley, just going through all the hassle of getting masks made and wafer fab done. Learning how to interface with those firms.
And so we were actually doing design and getting things implemented, seeing if they worked, kind of bootstrapping ourselves. And a number of simplifications occurred. In fact during that first year just in informal collaborations between here and Caltech – these were done not on the network – in fact the strange thing is that Caltech has yet to join the Arpanet community. So they’ve been sort of a stand-alone community. So what happened was there were courses that Carver ran there, Carver could interact with us over Xerox network facilities from Pasadena, so at one point there was all this exciting stuff happening. A lot of new ideas were boiling up, and I suggested to Carver that we ought to write a book on this. It was sort of a flip thing, okay. It was, you know, it was one of those things you just say. And he took it seriously and said let’s go write a book.
The way we chose to do that had two dimensions. One was – and this was sort of by accident –we had met Carlo Sequin who had just joined Berkeley as a professor coming from Bell Labs. And he had input some ideas on standards and stuff. He was going to start a course there. He wanted to teach something about integrated system design or circuit design. This was the summer of 77, that we thought we were going to start to write a book. We actually started at the beginning of August of 77. And Carlo was going to run a course starting in mid-September.
So we drafted three chapters of very primitive, basic material on how to get started doing design. And printed it up on the Dover here and bound them up nicely in little velo-bound things and gave them to Carlo. And he used that as his course notes. Then what happened was during Carlo’s course we just kept right on writing. And there was going to be a course run in the spring at several other places where we knew people. One was at, I think we did CMU that year, that was the target then. There were deadlines that got us to produce some stuff and just print it and hand it out.
The fact that the courses were going did two things. We were getting feedback on the material from the courses, and we were also working on deadlines. We’d have an arbitrary – now let’s get a couple more chapters done by the time [Bob] Sproull’s course starts. And we were just writing the stuff. But now we’re writing it basically in a community of people here, and some people at Caltech interacting over the facility from there. So we had a slight extension of the network. Basically to Caltech, and a number of people in the community outside to interact with. And then it turned out that Sproull at CMU was on the Arpanet so we started to get an informal, occasional message about what’s happening, and errors found or ideas of what should be in there.
Then in the spring of 78 another set of things happened. The material was evolving fairly rapidly. And it turned out Bert Sutherland was on the EECS visiting committee at MIT. He’s a member of a group of industrial leaders that come there periodically and meet and look at the programs they have and make recommendations and so forth. And he suggested to people there that gee, they ought to run a course on this there. Because, you know, a little one had been run at Caltech, and Berkeley, and CMU and it looked like interesting material. And it was in the process of being developed. And MIT, unlike some schools, is big enough and confident enough that they’ll try anything.
S: Now for a naive listener like I am, what was it about the whole idea of circuit design being done in this way that was so radically – I take it that for folks like you to get involved in that was a very new thing.
C: Yes, yes and it – okay. That’s a good question. There are a bunch of things that I haven’t – I’ve been giving history without the content of what was actually happening. What’s the story, anyway? The underlying thing was that, in fact, these courses were being taught mainly in computer science departments. Not in the electrical engineering side. They were not related so much to the details of how transistors work, but more given that you have a simple model of a transistor, how to hook them together to do interesting things. The other thing – there’s a set of – see I have a very developed knowledge of the history of electrical engineering. I’m interested in the movements that have gone through that. And to some extent there’s a thing that you can do, which is that a lot of the leaders have this in their heads, but they don’t know it. So one funny dimension of this is you can map a thing in and make it happen because it follows some previous pattern, that people kind of can sense.
There’s a thing that happened, probably happened in the early twenties, in the early days in the development of radio transmitters and receivers, there was first a focus on the devices, the vacuum tubes. The heroes had to do with people who invented the triode, or the things which were the basic devices. You know, somebody discovered that when you touch the top of a light bulb you get a shock, and out of that people gradually figured out that you can make things out of those, and you could make a controlled amplifier thing out of a piece of evacuated glass with some metal in it, with voltages on it, okay?
Now, what happened was that the development of those devices, making better such things, was the key thing for a while. And there were some very simple little systems you could build out of them. In fact, the people doing the science were the people developing the devices. Then suddenly there was a break in the field, and what happened was that making the tubes became sufficiently standardized, there were enough of them around and they were described, that the intellectual interest shifted suddenly. Because of some new things that got built that were combinations of these, it shifted to the building of systems out of them, okay? And a lot of it came through on the artifacts.
In other words instead of having pictures of vacuum tubes, the interiors of them and equations about the physics of them, suddenly books were appearing that had circuit diagrams in them. Where the tube was replaced with a little round symbol with little funny things in it, and then they were hooked together in ways – this is how you make a superheterodyne receiver, and this is how you make various kinds of transmitters, and that suddenly swept through and became the thing that was interesting.
What happened here was that the level of what you could put on a chip had increased to a point where actually large systems of interest to computer architects could be put on a chip. And by simplifying the notion of how to think about the devices, by giving some examples of systems built with those, and having a very simple, completely consistent body of examples that we present in the book, there was some threshold passed to where people – it just became intuitively obvious to the typical senior or first year grad student in computer science that you could take ideas about digital systems you want to build and very directly just map them into these silicon structures. Write little programs that describe them. And it was actually quite simple to do.
S: Now was there a tradition – these things happen so quickly it's hard to talk about traditions – but was there previously some means by which a computer architect would communicate to a designer?
C: Yes, yes, that’s – it’s sort of what’s being displaced, is the thing. Yea, there’s a whole thing being displaced by this. It has to do with the following. Traditionally in the industry, in order to design a chip, you know if you really look at a chip it’s strange. What’s so interesting is there’s art and artifacts. What do people see when they see a chip? But if you really look at one of these things, and you were to look at pictures, in fact [referring to VLSl book] where would be a picture of that chip. This whole colored picture here is this chip. And this is a bunch of student projects out at the MIT course I ran. This was in the fall – we’ll have to pick up the thread of that history again.
C: But what happens is that in the past the computer architect – like, this thing here is a LISP microprocessor, this particular one here. The computer architect would have in their head some rather abstract representation or notion of what they want to build. To build an abstract machine that will execute a particular programming language, let’s say. Interpret a language or run a simple low level machine code which you could then microcode to emulate some particular language. What the architect would do is they would draw, usually in a sort of informal notation – in fact there is no high-level hardware description language that’s in common use – they would, in any of a number of semi-formal or informal notations, they might draw a block diagram. Of what the machine looked like. And the way to think of it is, it’s much like having little pathways that you can move data around on. And there are little places you can hold data. And there’s some control mechanisms that operate under some synchronized sequencing of a clock, that say which little gates to open to let data run through which functional boxes, to go where to be stored for the next piece of time. And out of that you build up an engine. But they would have this in an abstract form. They would be thinking of bits of information, 1s and 0s, moving around on abstract pathways to registers that would hold it. And the problem is how to take that and convey it into a real technology that actually runs.
So what the architect would normally do, and some of this is because of a social stratification in the field, much like what happened in programming in the 60’s. As a little aside there was a time in the 60’s where the whole thing was to become a system analyst so that you didn’t have to code anymore. There was a – coding or actually programming was considered a lower-level thing but the people who planned what to program were system analysts, okay? What’s happened – it’s almost always been that way in hardware – has been that a lot of architects like to think the grand thoughts of the basic blocking out of the computing structure but leave it up to sort of lower-level people to do the logic design and the actual circuit design and so forth.
So what would happen would be if you wanted to design a machine often the architects would block out something, and then they would convey that description to logic designers. Who would transform the logic blocks between registers into a description, usually in terms of logic gates, and/or invert gates, nand or nor gates – simple little things which are again an abstraction. They have, you know an and gate has a bunch of inputs, if all of those inputs are high the output’s high, if any one of them is low the output is low. An or gate has a bunch of inputs and if any one of them is high, the output’s high, otherwise if they’re all low the output’s low. They’re and and or gates. And you can take and combine, you can write down – in fact the architect might even write down the equations they want a box to do, and someone would translate those into gates that are aimed at some technology. Registers would typically, be made by cross coupling some of these kinds of gates, etcetera. Or you could actually buy parts that are registers. These are chips that just are a register. Or chips that are gates, and wire those up and hook them together.
So there would be some logic designers or there’s a little bit of a difference here now. If you were going to go into designing a special chip, then in fact you’d go to a logic designer, and then they would hand it off to a circuit designer, and then to a lay-out person. If you weren’t going to build a chip where it’s all integrated, you could at the logic designer level go out to a standard technology where you could just buy nand and nor gates. You can buy those things and wire them together. And when you look at how people in CSL4 build a big computer that’s what they do. They buy the things that are little gates, and they wire them together. And they build a great big, complicated machine. But it’s all just random wires wiring the thing together.
But that’s not very interesting, because it’s not integrated as much as it could be. And in the industry, before people started doing things this way, what would happen would be the logic designers would have transformed one description into now an actual formal description in terms of gates, and flip flops, and so forth. In a way mirroring or copying the way people were doing it if you were to buy the parts off the shelf. In other words how you hook things up might not look very different. They were sort of mapping into this new technology an echo of how it had been done, at least to that level, in the old technology.
But now you couldn’t buy those gates, you had to actually go and make the stuff in integrated form. So now there was another group, of people called circuit designers, who would take the gates and transform them into little MOS circuit designs. I can show you what some of the ways – there are different symbolic representations. If you look in the book, actually all these different specialties have their own symbols. Here is a function, a little truth function, for a particular kind of gate. Here is the circuit diagram for that gate. So an architect might have done something like block out a structure. Let me show you just a block diagram sort of thing. An architect might say “I want an ALU” [Arithmetic Logic Unit] and it’s got these function blocks and each one does a thing and hook them up this way. And then in the traditional way these would have been implemented by someone taking that function, writing out logic equations, implementing them with gates. Then somebody takes those and makes the circuit for that. Then there’s a lay-out person that takes that circuit and fleshes it out in the actual lay-out that would go down here, okay?
Now when you go through all those intermediate specialties, where the people only know the adjacent one, maybe, but not the further ones, there’s a funny thing that happens. You end up with a thing that usually looks like this. I’ll show you. Here’s a picture of an early microprocessor, where what’s happened is somebody drew a block diagram, and they may have iterated so they’ve gotten some simplification, but really they’ve got little pieces that go through lots of random stuff that hook everything up. And usually the architect looking at this doesn’t really understand, in detail, what’s going on at the bottom level. The person who actually did the bottom level stuff probably doesn’t know what a computer is! I mean actually has no comprehension of what this overall thing does.
Because of that, and for some reasons that I won’t go into, to get these things to work at all usually requires fairly elaborate simulation down at the bottom level. So a whole body of lore has developed about vast computer programs you run in order to make sure these things are going to work. So that you can run a thing that’s running a representation of the bottom level so that the person that designed it at the top level can be convinced that maybe the thing will work.
Anyway, what this stuff does is it basically shows the student how to map very quickly down through that in a very direct way. So that one person can learn all of that. And the reason – it’s a funny thing because if you were to do it the way it’s done now in industry, most architects wouldn’t be interested in doing that. I mean they can’t themselves actually do enough or visualize enough – the lower-level work is very tedious the way it’s currently done. People don’t use good computational aids for it. There’s a lot of real hack work done at the lower level. And so that sort of reinforced the tradition that the architect is superior and doesn’t get involved in slave work at the bottom.
However what’s happened now is people who’ve learned to do the lower-level work, and do it cleverly, and do it in a simple and structured way, can come up with systems much more powerful using this technique. So now there’s a new group of architects who can displace the current group who don’t know how to do this. So what really seems to be happening in the universities to some extent is that there’s a feeling of a new generation of people who are going to knock off the previous leaders and all that. And they’re all polishing their guns and getting ready to go do battle. And that’s okay. There are some forces for collaboration, too, which are good.
But what, I don’t know how to express it exactly, except that one person can now very easily do things that in the older tradition would take a whole team of people. It does have the following property though. Any one of the things that the person’s doing may not be 100% optimized to the group doing that specialty in the other way. Because they may not sense what the person’s doing overall. You see?
C: Sometimes, by optimizing each of those levels, it sort of squirts out somewhere else. And you may make this – you can minimize the number of gates, for example, at the gate-level representation, and by doing that cause many more wires to be created, so that the actual lay-out is bigger. Then the lay-out person will scrunch down and make complicated geometries and pack all those wires into the smallest area. So there’s all this entropy in the current designs. I don't know how to express it exactly. Other than to say that if you were to – you could take a lay-out person, and they’ll come here and they’ll start laughing when they see any of these designs. They’ll say but look at that area there. You’re not using that area.
Well it turns out – that’s another story, too, because there’s this whole pressure on optimization that exists in the traditional industry. Because the people – there are heroes. The heroes that made the great fortunes, and made the money, and started the companies, were – especially, the people that started Intel – there was a thing that happened where the same break happened when this industry started and that was when the planer process, the basic technology, became available. I’m really jumping around now. But Fairchild had a research outfit back in the middle 60s that was right at the place where PARC is now. In that there was a whole bunch of stuff boiling up, which they couldn’t integrate, and it broke and went out. And that’s how Intel got formed. And the idea was they had a lot of people doing device physics work and so forth who’d come from the community that had developed the transistor to begin with. And had developed discrete devices. Not the devices that were basically involved in, especially the new planer MOS technology, out of which this is built.
People left there and started this company Intel. And a pattern of those got started. And what they did was they made memory parts. And later microprocessors. But right from the outset a very key part of making memory parts – memory parts look very much like these things here. The designs are simple, and regular. There’s only one cell type repeated in two dimensions. And the size of the overall thing you’re making is strongly a function of how you cleverly pack that one cell down. So there was this strong emphasis on getting things packed down. To where everybody felt it was a duty to pack things down very much.
So the lay-out experts that could cleverly get a design packed down a lot, and the company that could pack things down could, for example, get more of them on a wafer, so they could get a greater number for the number of units processed. They could sell the part at a lower price, that kind of thing. So there was, in that part of the industry, a great deal of effort to scrunch designs down. Also design was expensive. Prototyping was expensive. And if you were going to go to all that trouble you really would go to the extra trouble of packing it down.
Now a lot of these designs you’re seeing now are not necessarily meant for long runs where it matters what the yield is. There’s more of an emphasis on quickly designing something and simply getting it to work. So that’s more like using a high-level programming language, where you may trade off getting a program that occupies the least memory space or runs the quickest – you may trade that off willingly against being able to write it more quickly or get a more complex function implemented and working.
S: Now, is there a difference in the way that – from what you’re saying it sounds like there are some issues in terms of the emphasis on packing and economy in that respect – are the things being used differently? In other words is there a difference in the sense that they’re being used in larger quantities, or they’re more standardized, whereas these are more special purpose?
C: That’s right, yes.
S: There is that relationship too.
C: Yes. What we’re seeing is the following. If we come all the way up to the present, the net effect of a lot of this has been that it’s easy for people to learn how to do this, and it’s fairly easy for them to quickly design things, so that the cost of designing a chip is reduced by a very large factor. More than an order of magnitude. Also the system that I talked about – I didn’t really describe it at all, I just sort of ranted and raved about this MPC79 thing as an effort – that system makes it possible to prototype a design, also with at least an order of magnitude reduction in cost from what it normally costs in the industry. What that means is that chip design could be done for lots of things it wasn’t being done for before.
It was mainly, before, only being done for things that would sell in large quantities. Really the numbers go something like this. It may cost $500,000 to get a large complex chip design done by the traditional means. Simply making one prototype run of that chip might cost $15,000. To get your first one that you could test to see if it worked. And you might have to iterate several times and it may take a year of iteration – to iterate three or four times could take a year and $50,000 or something like that.
What we find now is that people can themselves individually design something; so one person in a few months can design something. So maybe the design cost is brought down to $10,000, $20,000. To iterate a design might only cost a few hundred dollars each time using the multiproject chip technique. So in fact what’s happened is that people can now design and prototype in this more easily and cheaply than if they were to buy off-the-shelf TTL, which is the standard building blocks like you see the people over at CSL using. I like to make that frame of reference, cause they don’t really know much about this stuff. They still think it’s just primer stuff for university students, and not very serious. But they'll find out! [Laughing] We have this neat little inter-laboratory rivalry here. I’m very proud of the fact that they scoffed at this when it started. Cause it sort of calibrates you on how good they are at envisioning the future. And it’s very motivating to have these phantom adversaries. You’d dream them up if there weren’t any real ones.
Where was I? Oh yes, the trick is that if, for example – suppose we want to add a neat function to an Alto. Right now what people would do, they might have to add a whole board full of stuff. And they’d buy the TTL parts which you plug in, which are these little gate things, and they’d wire them up with these little funny wire wrap things. And hook all these things together and they’d do logic, you know you put a clock in there and some voltage and the thing hops around, and it does logic. One of those boards, and all the parts to stuff into it, could cost you several hundred dollars. And in fact you have to buy all the parts, you have to get the man to put the thing together, so it might take you several weeks. Even if you had a design to start with, it might take you several weeks of time to get the thing, actually get enough parts and get the thing wire-wrapped and plug all the parts in and have the thing working.
What’s happened with this that’s exciting – and isn’t yet realized, but probably by the time we get a few papers on it will be – is that you could, for the same amount of money and time, prototype directly in this. Instead of buying those parts and putting it together you could go to your Alto and draw the thing on the screen and make out the mask pattern that would print it right on the silicon. Much more densely so that in fact one chip would typically hold as much as a board full of all these other kinds of parts. That’s something which means that the current way that you do electronics will be displaced. A lot of people will be doing this now. And in fact for the short run, one-of-a-kind things where you only – you may need one just to try an idea out, or you might want to make fifty to share with your friends around the second floor or whatever – for those it won’t matter whether it’s packed down all the way. You want a clean design, you want it quickly, you just want to get it working, and then you make enough of them to get out to your friends or whatever.
So that’s a very different world from the world which believes that it costs $500,000 to get that. You have to have a marketing survey, you have to set up a big program, guaranteed, you know, that when you put your ads out people are going to buy a million of them before you break even. And then you start making money. It’s a completely different world.
It turns out that there is a huge amount of this other kind of design going on. The microprocessor and the memory chip are typically surrounded by lots of these other parts that do everything else. And they’re the parts that are bought and stuffed in the board. Which now could be displaced by custom design. So there’s some really exciting implications. I think a lot of it is just economics – time and money will determine what happens. And ease with which people can learn it. Also, it’s actually in some sense a lot more fun to do this than to wire up other stuff. You know, there’s a new, simple, rather colorful thing to learn about and do. Which for a while will be exciting for people. Because it’s new, and actually easier and more fun than the other way of doing things. So there’s that.
Let’s see, why don’t I go back to 78.
Conversation with Lynn Conway re MPC79
Conversation 1, tape 2 3/12/80
C: What was interesting was that while we were working on the book, by the time we did the second two chapters – we did three chapters for the fall, and Carlo Sequin was using them. And then in the spring we released – this was, by spring I mean actually February – we released five chapters. Which had a lot of stuff. What we released was basically up through here [referring to the book]. And that was a huge panic effort. From August through February we created that stuff.
But the reason that it was really terrific was that in parallel with getting the book going there was a project going at Caltech by a bunch of students to design a machine. Using this method. Fortunately we got the book going just soon enough to get everything they were doing mapped into the exact style of design that we were going to write about. There were a lot of new ideas we were trying out, very simplified design rules, and they used those design rules. Instead of a little more complex set that would have been for a specific fabrication line.
Anyway what we did was, that design – this is the design right here – we integrated that as a design example. So basically the first three chapters give you the primitives out of which you do design, and then the next chapter tells a lot about how to describe design. There’s this thing about how to take – how to go from stick diagram to a lay-out, how to encode the lay-out, okay? And this is the starting representation for designs that people use. And they had seen some examples of those back here.
So it turns out you can go from the function, almost directly through to a stick diagram, which you then just instantiate by design rules to create the lay-out. So going from the block through to the lay-out is more transparent to students now. That’s what they’re learning in the courses.
Then, in chapter five there is this example it talks about. This machine at Caltech. And this is really slick. Because now students can analyze a design, okay, they can see all the different levels of representation of a design. And it’s colorful and exciting. And it was done by students. So that if students at MIT read that students at Caltech designed that [laugh] you see what I'm saying? Well gee, they can do better than that, right? You see they know it must be possible, I mean students did that.
Okay, so in parallel with writing the next few chapters, this thing was going on. And the key students involved were able to go over to Xerox EOS [Electro-Optical Systems] down there, and they were able to diagram – Johansen, one of the architects of it, was drawing these diagrams. So now we had a thing rolling. I was working here writing on stuff, Carver was working down there, there were other people here working on stuff, contributing. Like Dick Lyon – we had to have an interchange format. An interchange format is a common way, at the very bottom level, of describing these objects into which any design system would put itself. It’s almost like an intermediate object code. What would be a good example from programming? To some extent it’s almost like thinking of ASCII as a data representation. There’s a primitive level of code, a definition of a very primitive language, that simply is a language about geometry. About boxes, and polygons, and circles and wires, which is machine independent, that was invented. And it’s described in here. It’s called the Caltech Intermediate Form.
And this was the section – you see a thing that started to happen now was that Dick Lyon here, and Bob Sproull at Carnegie Mellon, were writing this section. As we were going. So now we have different people at different places on the network writing to get this thing to go together. A course is going to run at CMU. And the students at Caltech are doing an example which is getting mapped into here, okay, as an example. So that was, all coming along, and there was a use of the network for that.
Now we here, and Carver at Caltech, had occasionally – him for his classes there, and we here for our own designs, had gone out and made masks and a wafer fabrication. And normally that took a long time. It was a pretty complicated, ghastly thing. It was very motivating for students to see their chips come back, and that was really kind of neat. But usually that would take six months or so. I mean they might be lucky, and they might get some back in – but most of us didn’t know much about the economics of getting chips made, or we didn’t have any standard procedure for doing that.
But anyway, Bert [Sutherland] got MIT to go along with the idea that I would go back and teach a course. Basically I was pretty much afraid to do that. That was a pretty strange thing, when I got going in that. But the idea was that that gave us a tremendous push then. We wanted to have the whole book finished, basically the draft manuscript, before I went back for that course. So four more chapters got done. This was the thing [showing copy of the manuscript] this was essentially all of nine chapters. The very preliminary version of nine chapters. And by the time we got in these later ones we had collaborators all over the place. Like Chapter 7 was a whole chapter written by Chuck Seitz at Caltech, of material that fits right in with the book. And Chapter 8 had some material contributed by people at CMU. And most of these people were interacting with us over the networks. Okay?
S: Now how would something like that happen, say that chapter. Everybody would recognize the need for a chapter on that topic?
C: Well this was a very strange thing. I think there were a lot of people aware that this overall thing was going on. And it to some extent, to a slight extent, was like things that can happen here on the second floor. If an exciting project gets underway, people will see some message traffic, or they’ll see people with the artifacts of the project, and they’ll become aware that it’s some project running on the networks in that community. And we tended to, we were always enthusiastically showing everybody all this stuff. We were all very excited about it. So we would run into people who would get also excited, and if a thing is hot, you know, people want to be involved. And so people would be seeking ways to be involved. And we would be, as we noticed that people were interested in collaborating, we were thinking gee, you know, what more could be done? It’s, there’s a funny process there of, in the time available, attracting the most participation, and getting the best thing to happen from it
It turned out that a lot of it was just spontaneous. Like H.T. Kung – we knew we needed something on parallel architectures. We were trying to find somebody who would do a thing on that that would at least hint at the possibility of mapping that area in. And I think I saw a seminar notice that H.T. Kung from CMU is going to give a seminar here on some stuff he was doing which sounded vaguely like it might be applicable. And this was, I think – my god, it must have been in May or June. And I was going to MIT, you know, the next fall – well I went to this seminar. And H.T. Kung presents this stuff, and it’s fantastic. And it was all great stuff that mapped right in. In fact he was preparing a paper on it, I looked at the paper, holy mackerel, that could make a good part of Chapter 8.
Got him sold on the idea and he got instantly plugged in to what was then a running thing to a deadline. I mean a month later this whole thing was getting printed [laughing]. And so his stuff got mapped right in to Chapter 8, made Chapter 8 terrific. And a whole bunch of people were here – in fact, what are the dates on that [manuscript]. The dates are on the bottom – yea, July 24. I had to leave to go to MIT – sort of middle August or something. And we had to print like 300 of these. Some to go to MIT and some to go to other places. And so that backed us back to where the whole thing had to be in shape by the end of July. So a lot of this stuff, in the later chapters, was just coming into being in a mad, panic effort during May and June of that year. But you could draw on all of these other people, that were aware of the deadline, to create material. And it’s surprisingly consistent.
Now let’s see. 1 went back and organized and taught this course at MIT. And that did one more thing that was interesting. The course worked out pretty well, the students really learned a lot of stuff, but we had the idea of trying to see how well and how quickly we could tum projects around for the students. And that idea sort of came up – oh, really it was just a notion – gee it would be neat to do, on the way going back there. I mean that was the – in the summer that was a glimmer of something we might try to do. And what I did was, while I was back there I stayed in very close contact with people here using the network. And we were able, in parallel with getting the course going back there, to set up a set of arrangements here for actually taking files that were sent over the Arpanet, and using some ad hoc procedures, but fairly quickly getting those projects merged together, getting masks made, getting wafer fab done. Drawing on Hewlett Packard for the wafer fabrication.
We didn't know, when I went back there, that we’d be using HP for wafer fab. There were several places we could do it, and we didn’t know if we could turn it around quickly or anything, but we were at least going to try to get projects. I motivated the course by simply going in and claiming we’re going to do projects. Without knowing really exactly how we were going to pull it off. But that really got the course going. Because students believe what you tell them. You have it in the abstract, you know – there’s this stuff in here [reading] Students will do lay-out, blah blah blah – selected projects will be organized into a multiproject chip set to be implemented by commercial mask and fab firms. That was right in the course abstract.
The thing about the MIT course that was interesting was that it was a new school, none of this had been taught there before. So it was a chance to see whether the material could get mapped into a place. We had to build our own design aids there, organize a computing lab, sort of teach the material and get the supporting aids that would enable the students to describe and plot their designs and all that, and then ship the designs out here and get them implemented.
Well to make a long story short, that all worked very well. And in fact the procedures for implementation went well enough that we were able to get chips back to the students about five weeks after the things were sent out. And that was very fast for a large group like that. And it had all been done remotely and to them, back at MIT, it all seemed to be completely magic.
So this had a huge effect at MIT. I mean the faculty were all impressed and everything. Which was terrific. Because what that did was it astonished not only MIT, but the other schools around saw this weird thing happen at MIT, okay, and then what were able to do was just play that in the networks to where – see really one of the underlying motives here has been to sell books. And so a lot of it has this entrepreneurial, promotional thing. But where you use what infrastructure you have just to make things happen.
That course ended last – that was about a year ago I came back to PARC. And in the interim from then to launching MPC79, I think a lot of us just really had the idea that wouldn’t it be great to do that for a bunch of universities. But to do that we couldn’t use ad hoc procedures. We couldn’t be individually calling them on the phone and pulling files over the Arpanet and so forth. So we had to set up this system, that actually sat on the Arpanet, that would essentially rather automatically interact with all these people. And that’s how MPC79 happened. We basically build a system, turned it on this fall, and when that service was announced a lot of the other universities got involved in the thing at the same time.
We’d also run a thing for Stanford, last summer. Where – Stanford University is sort of interesting. They have a very strong group of faculty on the traditional integrated circuit design side of things. And they were in the camp of doubters about whether any of this stuff was ever going to work or anything. And what we did was we made – oh let’s see, I ought to go back even further. Doug Fairbairn and Dick Lyon, last spring, ran a course here at PARC. That was just after my MIT course, they ran a one-week course. A short course for people here on the second floor to learn how to do this. As a result of that course there’s a set of videotapes that are sufficient for an experienced digital designer to learn how to do that stuff. What we did was we used those tapes then. We made them available to a group of faculty from Stanford. In fact I don’t how the contacts were made, I can’t remember now, but they were a group of people, not in the traditional integrated circuit area, we invited over to look at these tapes to learn about this stuff. That group has since gone on to put on a really good course at Stanford. And there is a tremendous burst of activity there, but not from the traditional IC, integrated circuit area people. All of which has caused all sorts of interesting political tensions over there.
But that meant that Stanford University could get in the thing this fall also. So the universities involved were Caltech, and Stanford, and CMU, and Berkeley and MIT. And a scattering of other smaller schools that had been able to put in courses. Carver Mead, my co-author on the book, ran a course for teachers of this stuff up at University of Washington. And some faculty came from a scattering of other schools and some of those got involved in this MPC79 thing. But the neat thing was that there were enough schools running courses this fall, and they were all on the network, so that we had a big enough test of MPC79 to really make the thing worthwhile. We were able to get enough designs so that you could really dramatically demonstrate how many could go through one mask making and fabrication run. Which makes the cost per design very low.
S: Now that’s the multi-project chip part of it, right?
C: Right. We had done multiproject chips before, like at MIT we got a multiproject chip, but it was remotely entered and turned around. But what the MPC79 thing did, the key there is instead of doing it for one school, and instead of doing it by sort of ad hoc procedures, we interfaced – you can imagine a large tree of schools all over the place coming through the network and interfacing with an automated system that was gathering up and spooling their designs. Accumulating a whole set to go quickly through implementation. Okay? So the courses were geared to go along with the schedules set by this service. And the students were learning design, they did their design, and then near the end of the course they started sending preliminary versions of the designs, getting the code checked and so forth, and then there was a design cut-off. Whatever designs were in place then, we shut down that external thing and went, and quickly got them implemented.
And what that really did was allow this tremendous sharing of the capacity of mask-making and fabrication. So that if you look at the actual wafers [getting wafer] if you look at the wafers before they’re broken up, you can see how many different projects there are. Because there are [counting] one, two, three, four, five, six chip types. And then some of these chip types, like these up here, each of those has one, two, three, four, five, six, seven, eight projects in it. So there are two different kinds of wafers, or two mask sets – each of these has about forty-something projects. And that means a great economy.
S: Now in the former ways of doing things, would each of the wafers have fewer designs, or –
C: It would have a whole bunch of chips that would all be the same design. Just lots more of the same design. And it’s really strange because it’s a chicken and egg problem. In the industry there are so few designs done, and there are intervals between them – a company will only have a few designs in progress. So what they do, they’d have a whole mask set and short fabrications run devoted to one design. So then it’s very expensive to implement. It costs about $15,000 to do that. But you can have one design, or you could have maybe a hundred designs in a mask set, okay? And so in fact if you can get a hundred designs instead of one, you have one one-hundredth the cost per design to convey stuff through to get a few back so you can see if the design works.
Now once you’ve got a working design, you may want to make a mask set with only that design in it so that you get lots of them when you run wafers. So there’s some complicated trades there. But what this does is it makes the actual creation of the chips cheap enough that you could do it even for students taking university courses to implement a project. Now that’s just economics and it’s just technology, but the effect on the community is interesting.
The most interesting part, I think, is that since this is this shared culture thing, the people at each school can really visualize what’s going on at the other schoo1. It’s all visible. There’s nothing mysterious, and that really sets this whole thing I talked about a little bit in the meeting. The competition between schools, the competition between individuals, but a need for collaboration. And in fact all of the schools have to collaborate on, for example, what the design interchange format is, the standard that they represent their designs in, because all the schools have to use the same to get it off for implementation. So there’s even collaboration between schools. There’s this hierarchy of collaboration and competition that goes on.
S: Another consequence, I would imagine, would be an incredible freeing of innovation in design, right? Because you’re not bound by these constraints anymore, of the implementation.
C: That’s right. In fact if you were to go and look at how a design is done in a lot of places – in fact that’s an interesting point. Because when we did our early designs, we were just learning how to design here, we went for wafer fabrication down to the Xerox facilities in El Segundo. And because our designs looked strange, or were kind of toy-looking designs, they were in a sense censoring our designs. They, would say, “oh, we’re not sure we want to implement that.” You see what I’m saying?
S: Uh huh.
C Why, are you doing this, and so forth. And the whole thing about people censoring your work when you’re learning is really bad news. The way we looked at this is you really want to present a completely blank, unconstrained thing and let people do anything they want to. And the marketplace of ideas will sort things out. And especially people learning – not criticize their work and so forth. They have plenty of examples to draw on, they have some ideas they want to try out, let them try it out. The criticism is mostly self-criticism of how well they can create some new thing. And in fact there's been a lot of really novel stuff created. And created using novel techniques of description. A lot of different things done, that you would not have predicted would be done.
Of course the other thing that’s interesting about it is just being able to observe all that. Figuring out how to maybe tap into that. There are a lot of ways, I guess, that I rationalize how this might be of value to Xerox. If Xerox were to turn on such a capability for implementation, rather than have a centralized group that does design and implementation would simply have a printing plant where you could send your file to have your chips printed, and have anybody in the company who’s doing design able to get at it, they may be able to get a lot of really neat things designed. And that could be a way that they would do that in the future, I don’t know.
The whole thing about the networks is really kind of fascinating. There’s a set of things there that seem to make this whole thing run a lot faster than it would ordinarily. Cause it’s really moving pretty fast. I don’t know. There’s a set of things that happens in the network that tends to lead toward simplification. It’s something like this. You’re forced to do things in a somewhat more constrained and simple way if people are to interact remotely. You’re forced to have formalisms and codes for things, where ordinarily you might be able to talk and explain and mumble and observe directly, okay? And so to some extent the doing of things remotely encourages compactness and simplification.
On the other hand if there are enough people interacting remotely, you can have each person experience actually a richer environment than if they were just dealing with a few people locally. If the network is above some critical mass. So what we’ve seen is a very large community that were interested in doing this all jumping in and wanting to participate. Sort of as a group evolving the standards and discovering by trying to do things where they need the next simplification. So a lot of what’s gotten documented and created has been much like some very short-time, compact thing happening like it would normally happen. I was trying to think, how do these technologies normally evolve into place? And by evolve, I mean where all the lore and understanding of how to design and what the standards are and what’s correct procedure and what are the scripts for the different roles and all that kind of thing. That seems to normally happen over a very long period of time. Actually about, the time it takes to obsolete a technology, you know, maybe ten years or something. And by a process of the normal thing about publishing of papers and going to conferences and stuff appears in the magazines and then later the books are written.
Whereas what I think happened here that’s sort of funny is that to some extent the whole thing went the other way around. That the book and the procedures and everything all sprang out of a network community that was simply evolving it suddenly into existence. And that’s a very weird thing. It’s really weird. Like no one person is really directing it, it just sort of happens a certain way. And, you know, with the urge of everybody in that community wanting to have – wanting to be able to do this. And simply all co-operating enough to get all the things put in place that would make it possible to do it, as quickly as possible. So the book served as a focal point for that lore coming together and, you know, points would come up. Does a certain thing need to be in there, can it be simplified, and so forth. And to some extent a large group of people would make the decisions about that. You’d get opinions from people, and ideas and so forth. And Carver and I were writing a lot of the stuff down, but a lot of it was in response to this rich set of interactions that were occurring about what was working in the courses, what you needed to do.
The way I like to think about it is that each person in this large and rapidly growing community was interacting with many more people than they normally would, because they could use the message system. So the branching ratio of your contacts was very wide. Now there might be some embedded things in there, there might be little groups that interacted locally and then there would be chains through those, but really one person could decide that they might want to create a distribution list for a thing, and they could be involved in it. And actually the average time constant per interaction is very short. Because there’s a threshold you go over. I mean if you only know two people and they’re always home, you can do a lot with a telephone. But if you know a lot of people and you don’t know where they are when and so forth, you get to where you can’t do much with a telephone beyond four or five and then you spend all your time trying to find people. Whereas in the network the time constant stays the same pretty much, no matter how many people you’re in contact with it’s still only half a day, or it depends on how often a day people read messages, so it’s a day at most.
So let’s see. The other thing that’s interesting – this has been a focal point here for the creation of a lot of the artifacts. There’s this whole thing about artifacts. Namely whatever of interest was getting created tended to get put into either the book, or a thing called the implementation guidebook, by Han and Sequin. This book is about design, there’s a thing about implementation that’s been written too. That got written the summer before the MIT course cause I needed to have that information for the MIT course. So while we were finishing up and printing this, I launched Hon and Sequin and all the students in the group that summer writing another book. Because we decided yea, let’s try to do implementation for the course, and do it quickly. But there weren’t many tools back there. So I was going to need all kinds of help in getting the tools in place, and having people understand about CIF code and all this other stuff. So what we did was, we had this book put together. It had more than implementation, it had a library of design cells, commonly used things that everybody uses in their projects. Input/output pads, programmable logic array cells and some other things.
I was interested in having all this other material available to take back there. So that was a very interesting thing. Over that summer this group of people put this weird little book together. It’s been iterated since and it’s about ready to be printed as a technical report. But it’s a very interesting thing. It had color plates in it and everything, it was another one of these instant books. And I took back some of those for the MIT students. So anyway what’s happened is with the pre-prints of the book, with the implementation guide, and with a lot of the other things, we’ve used the printing equipment here as a resource where we bring documents together here and write them here because we can demand-print them and then just mail them out all over the place.
So that means that there is sort of a common source of artifacts. So there’s some consistency in everything. A lot of these things are totally arbitrary, you could do them any of a number of ways, but by having consistency all the mystery goes away. People have enough of a common base of what’s going on to visualize things. Whereas you could go to every company in the valley here that does electronics, and all of these detailed little specialties will be done differently in each one. It’s absolutely incredible.
So the demand printing and creation of artifacts has been interesting. And then of course students have a lot of fun getting these artifacts back, they love to get chips and things. Even if they never do anything with them. They go show their friends. And so there’s almost a tribal sort of thing. It’s like you could imagine this as something like how do you engineer into existence a tribe in the networks, okay? You have more than distribution lists. You’ve got to have little badges and symbols, and all sorts of things. And so now what it is is the insiders – it's almost embarrassing, you know. People talk about “The Book.” And if you don’t know what book they’re talking about you're not in the clan. And in fact what they do now, it’s even worse than that, a lot of people go around, and they talk about “Chapter 8” [laugh] Or “Chapter 5 says,” but they don't tell you which book [laughing]. It’s too much. So it’s like the Bible, you see what I mean?
S: Right. Absolutely.
C: It’s really cool. And so there’s this whole feeling among these students, of being in this sort of insider clan.
S: Membership is no small thing.
C: That’s right, that’s right. And if people use the wrong color codes then everybody will laugh [laughing]. So it’s sort of created this monster. But anyway to some extent I’ve had some chuckles imagining that that’s actually what I’ve been trying to do. Is to understand how you can create things like that, you know, what the social processes are. It’s really kind of interesting.
S: It is, it is.
C: Somehow the artifacts – it’s something like just passing these artifacts out through a group of people or giving them to visitors who find what you’re doing interesting, sort of mysteriously spreads the news that there’s some new magic going on, okay, that’s really what it does. People go away and they may not even really understand what these things are. They talk about gee, they ran this course, and the students did chips, and they don’t know quite what that means but they look at these things. All these tokens. You have no idea how many of these we’ve passed out [laughs]. I've spent days and days, just in spare time – I mean I never mass produced this particular one for the MIT course, or had anybody do that, but I’d know oh, so and so’s going to come visit so I’d madly make up three or four of these. And give them a picture. What I found fun was, for a while what I’d do is – right after the MIT course all these visitors from these other schools – I would give them this list of projects, and the implementation sequence, and a little map and a picture, and a chip. You see, the students taught them how to do this stuff and look at this, oh one of them designed a microprocessor and so forth – give them this stuff, okay?
And so there's hundreds of these things rattling around out there in the world, off the network, but the message is passed out that there’s this magic going on, okay? And people will hear about it and want to try to run a course, students at other schools will hear about it, and I was thinking it’s almost like Khomeini pictures going through the mosques, you see what I’m saying? [laughing] It represents some token of this new little clan. And then everybody wants to figure out how you get in that. And the ritual to get in it is you take a quick intense course and you do a project, as a participant in a multiproject chip effort
And so people go through this intense learning, by reading, learning by example, and learning by doing. Having to create a thing that’s theirs. They have to be brave; they have to do a thing where they create this thing and the world’s gonna see what they’ve created. And it’s printed in silicon, and it goes in and one of these blurbs pops out. In fact it was really amazing, with MPC79 the thing that was really incredible is if you go and look at some of the people here, some of these are people – in fact like at MIT some of these people, I guess it was – in fact these two guys [referring to manuscript] these are people that work at BBN, they’re computer scientists, I think they're both managers, and so they’re working in an elite research organization. They were very intensely interested in what this was all about, because at BBN they’re on the network, and they’d heard all this flack going back and forth anyway, okay? So when Jon Allen offered the course this fall at MIT – he’s the professor teaching it there now, he took it over from the one I ran – BBN sent them down to take the course. And they weren’t even sure they were going to have time to really do a project. They were working full time. But they got very caught up in it. It’s like near the end, they spent several sequences of staying up all night for two nights in a row and the whole thing, okay? [laughing] Caught up in this thing.
And in fact it was interesting. This is the fun part of it. It’s like we were here ready for the design cut-off where we have to then do our stuff, to do a lot of computing and information management, we call it, to take and pack all these things, decide what’s going to go where, convert the stuff. There was a huge computing effort that went on all one night, here, right after the design cut-off. But it was really strange, because for two or three days before that the network message rate was coming up and designs were coming in at a humongous rate. And we could sense all this activity out there. You know, imagine a hundred and twenty-four people out there and all the faculty, all going bananas. Staying up all night, doing this stuff [laughing], And all of it is an initiation experience, they’re all joining this thing.
And then it comes to here and we do our thing with it and pass it on to Micro Mask and HP. And what really comes out of that is the people that participated in it had this really bizarre experience of working under pressure and creating all this stuff. And actually with the time pressure, and having to get it done before the deadline, people actually I think are almost less inhibited. They just do what they can. And they created – we couldn’t believe how good some of the stuff was. They were, some of the things they reached too far, and they made some errors, and it won’t work, but they can iterate again. But the kinds of things that were attempted, if you really look at it in the large, were just amazing. Some of the things that people did. There’s a lot or neat stuff. I want to show you –
S: Can you characterize, somewhat non-technically, what it was about the things that was extraordinary?
C: Yea, I think so. The scope of the things, first of all. As first projects people were doing things much larger than toy designs. Like here, here is a set of designs from MIT. This design was done by Lance Glasser, and it was actually done as part of Carver Mead’s teacher’s course, late last summer. We put it in because he’s a new faculty member at MIT. There are four designs in here that came from that course, which hadn’t got implemented and were sitting unimplemented in design file form, which we wanted to bring through. This [referring to book] is the kind of design typical of designs done in previous such things. A little design, couple of cells and actually rather modest. This [referring to book, a much larger design] kind of thing students in universities are trying now. Instead of doing just one piece, they want to do the whole thing. They have a whole system they want to do where they make the whole thing. [referring to book] This is a music synthesizer – you can see the difference. There’s a level that the art has come up to, in what people are attempting. And their tools are good enough that they can, they want to design things that look like complete, full chip designs. As their first project. And the tools have come up to almost support that. And so in fact these students will go and they’ll get big pictures of these and put them as frontispieces in their resumes, you see what I mean? [laughing] And so there’s all of that sort of pressure.
But the other thing that’s interesting is if you look at some of these – let me show you. You can put anything you want in here. So they start putting their names in here. There’s this whole thing about owner artifacts. In fact some of them are really neat. Like on the LISP machine thing, this is a little map of the Boston transit system. This is their logo, okay? And –
S: How do these things come out on a chip?
c: Oh, I’ll have to show you some of those under a microscope. Look at this thing from Stanford though. This has got all sorts of bizarre stuff in it. This guy calls this CRAY 2 – CRAY is an insider thing in that CRAY 1 is a great big, complicated computer.
S: Son of Frankenstein.
c: Right. But look at this [laughing].
S: Okay, a playboy rabbit.
C: So there’s actually art appearing.
C: See, look, SWIM. I don't know what SWIM means. But anyway there’s all this weird stuff. And some of it is, you know, here’s the infamous BUFFALO chip.
S: [laughing] Right.
C: So you can start to see that people are putting art in with their stuff. And some of it’s very good stuff. I mean some of these designs are really important new things. And of a kind actually not being done in the industry. And you can see all of them are, they’re almost all at the level where, as first projects, they look like regular full size chip designs that are being done in industry. And many of these people didn’t know anything about this before they started taking the course. So this is extremely exciting to them, they have something they’re proud of now. So in fact what a lot of it is, is I think this serves as a sufficient initiation into this new clan. The people are now totally confident that they know how to operate. They now can go out and design whatever, okay, because they have done it. Whatever the initiation for it was.
S: Right. Whereas it’s the sort of thing one could read about for years and years and never get one’s feet wet.
C: That’s right. You have to do it. And do it so that you feel you’ve been successfully initiated, a successfully initiated person. I think there’s something about that I think there is this thing about forming of clubs and clans which proceeds beyond whatever the adolescent initation rites are, that’s a continual ongoing thing. People will continually join new things, and that all rattles around. And what we try to do is provide a framework which is sufficient to provide that experience to these students.
The thing is, I don't think they feel that with a lot of the other things they learn about. And so a lot of students, you know, they don’t ever have a feeling they’re participating in anything. And so the more they hear about these courses now, they get very excited about that. And of course they can turn these things with plotters into these giant artifacts to put on the walls [laughing], which is really incredible. And impress their friends and everything.
In fact the kind of thing that’s sort of interesting is, I think some of the excitement about this stems from these properties of it. The fact that artifacts are created, in an unconstrained way – much like the dimension of art and architecture. It’s an experience that’s very novel, of creating new things in a new art form. That’s part of an initiation rite, that you now join this clan. And the artifacts are visible, other people can see them. And you wonder how that could get mapped into other things. To make other experiences more exciting for students. And to give more students the feeling of being somebody. Having done something. Have been a part of some kind of adventure.
I’m sure that all of this felt like an adventure to at least half the students. I mean a real adventure. They just got caught up in it, you know, you could tell. The fact that there’s a deadline, and they’re trying to create something – there’s just something about that that gets people caught up in gee, can I really do it. Can I really make it? And then they go into this work frenzy and creative frenzy, and then if they actually pull it off, there’s this tremendous feeling of accomplishment that comes from that. I don’t know. There might be other things that could get mapped into.
This is a creative domain, not an analytical domain. The idea is you’re designing things, you’re not analyzing things. It’s art and architecture, it’s not analysis. And that’s another thing I feel kind of strongly about. That engineering has had too much of an aspect of analysis. Actually what real engineers do is they are architects. But only one out of every hundred engineers is actually like that. Most engineers are just technicians. Analyzing and taking measurements and stuff like that. But the real engineers that actually do things create new things that haven’t been seen before. A thousand years from now that [engineering] will be seen as a new art form. Of mapping functionality into things other than, you know, blank murals or musical forms. Into any kind of structure, that does stuff, okay? Airplanes, or whatever. That’s art, at some level. And you know, things work, and fly around, and the experience of creating the darn things and seeing them all flying around is really quite interesting. So that’s what these people, what these people then would find interesting is they would really like to be able to do the great new architecture for some new application, and see products made and everybody using them.
S: It’s almost like what you’re providing here is you’re making the experience of creating something familiar.
C: That’s right.
S: There’s this element of newness and risk, and yet there’s now a sense of having done it before and having some idea of what it looks like and how to proceed, and where to begin, and –
C: Exactly, that’s right. They end up with a script about all that. In fact if you view initiation rites as being some test of simply your ability to successfully follow a script or something, even when it’s risky and you’re afraid, okay? And pulling it off to where you then could repeat things of that class, even in another domain possibly. There’s that feeling –
S: a sense of having seen it before.
C: of having created something, under pressure, in a new and unfamiliar domain. And in a way, strangely, that’s caused by having the implementation, having this blank mural of silicon, defining a game for how to interact with it that’s rather exciting, okay, and challenging, and being given very simple rules of how to build the little basic things. Rules which don’t have a lot of “don't do this,” and “don't do that” to them. That mostly have, if you do this it will work, okay? And it’s colorful, okay? And you simply turn that on and it’s absolutely mind-boggling how creative people will become.
So the thing I really sense is more a matter, it’s almost, there’s something almost profound about what you can observe here. It’s that more people are more creative, and more people want to do creative things than you might think. I mean many of these students seem like very dry, analytical types who don’t seem to get inspired, and yet I see what they do when they get into this. And so it’s something like, there’s some underlying mechanism there. It’s as though when, somehow, groups of humans discover an uncharted area, you know, that there is a continent on the other side of the ocean that they didn’t know was there. Or here is this blank thing and some objects here that make marks on it. Whenever you can create artifacts of a kind that haven’t been created before, and you have any leader that can show you, initiate you into doing that, the whole horde will rush out and want to do it [laughing]. It’s filling a vacuum.
S: It’s really interesting also for the issue of what you want to give people, to get them there. I mean what you’ve provided here is an incredibly precise and right on mix of a blank piece of silicon, you know, a very open-ended definition of where you’re going, and some very simple tools. It’s the issue of finding how far you need to set up, to provide people with the framework. And how far you want to withhold directing them. And you got just the right mix there. That combined with all the issues of collaboration. You know, if you think about collaboration as an issue of feedback, of having feedback, and how you’ve magnified incredibly both the speed of that feedback and the range of it, and the simultaneity of it and everything. It’s just an amazing configuration of things.
C: The thing has to do a lot with numbers, too. I was thinking that there’s enough universities on this, and they’re all involved in the same adventure that – and they all can observe each other’s work. They don’t know what the other school’s stuff is going to look like until the game is over, but they know that something’s coming.
S: Yea, and they know that they’re following, like you said there’s no mystery to it. There’s enough standardization that everybody knows that the rules of the game are common.
C: Right. I think of it as kind of like if you look at exploration – what gets people to get in sailing ships to go explore? I mean it’s got to be insanity. But it’s the same kind of thing. There’s just this thing and there’s different countries and they all want to get in and do it. And who’s going to get ahead of who [laughing]. And the leaders get a following to go follow them on an adventure, and the whole thing, you know. And they go, and the next thing you know they’re all running around South America, carving up territory, making maps, bringing back artifacts, stoking it up all the more [laughing]. So this almost is like, you know, the exploration of the new world or something. At some threshold the thing gets magnified and, you know, there’s a whole bunch of people, who will just run at the fence. They’ll just have a go at it.
S: But that’s a real – things have not been going in that direction [laugh] until now. I mean this is a real nice, you know there’s a quality of turning things around to it, and thereby advancing, that’s really nice. The whole time you were talking about the issues of what this is a departure from, those issues of what happens when you have that kind of specialization where you’re optimizing each place, but nobody has a sense of the –
C: It’s weird, it’s really weird.
S: That as a metaphor for specialization, technical specialization in general and what its consequences are.
C: Mm hm. In fact what you could imagine now are things, like it might be possible to get much more of a synergy and an interaction now between the machine design and the software that runs on top of this. And a lot of the thoughts about what we might do with the Dynabook, or the things we at the hardware level talk about the wafer-scale integration, have to do with people now, perhaps, when they design something taking on an even richer, larger span of things. They may create not just a hardware design, but they might want to make like a voice input or voice response system. Or some other thing that’s augmenting a computer system. Where they will do the trade-offs. Not only just in the silicon design, but the functionality that might be programmed into it, okay? And so there’d be even more of a tendency to want to span a larger thing.
And you know you could think of just gradually enlarging the scope of what these projects are. And trying to constrain it by what can you do in a reasonable amount of time, and learn from scratch, but still create complicated things. The infrastructure for implementing these things has to be such that you figure out ways to take advantage of the economies made possible by the scale of the number of participants. I don’t know. We can actually keep right on exploring this and just letting the thing roll. And I think there might be other things that you could do, somehow, this way. Some of this is really like how easy it is for us to create documents around here, with pictures, and interesting stuff in them. And a lot of people like to do that. I mean we’ve all got caught up in creating, you know, word crafting with pictures. It’s fun, it’s really fun. Especially if you can print up a hundred and send them to your friends and impress everybody with all this neat stuff you can create. So there’s that property to it too.
S: Yea. It gives you an incredible appreciation for what your material resources, the consequences that they have for your creative processes.
C: Yea. In a way the Xerox copier itself is sort of a metaphor for some of this. You can do a lot with just copiers, to create and make lots of copies, and hang things up that are interesting looking. I don’t know where this really proceeds from here. I do find it interesting that so much creative output can occur, with this balance of things. These people at all these different schools, there’s just no way they would be doing this stuff and creating all this and having that experience, without this very peculiar set of circumstances that got it going that way. Understanding those and trying to generalize a little bit. Think what other areas it could get mapped to.
S: Well it’s clear that you’ve been keeping a very beautiful record of all of this.
C: I actually can reconstruct a lot of it in detail if 1 really had to. I mean what actually really happened.
S: And just in terms of the way that you’ve obviously got the artifacts nicely packaged and, you know, things like making a tape like this. I think that this is so full of lessons, that there’s no way you can get them all at one pass. Maybe there’s no way you can get them all right now. Some of them are going to emerge later on. And I think some of them have to do, it’s not just an issue of could just this sort of thing happen somewhere else. There’s all sorts of other embedded lessons within it that apply in different ways to other things. And so it’s an incredibly rich thing. One of the most amazing things I’ve ever heard.
C: It’s really a lot of fun. And there’s something very funny about the networks. Where there’s something about how to initiate tribe formation, how to get the exploration of a new territory going. It has to do with the ease with which you can configure large numbers of people to know about each other’s stuff. And I have a feeling that many more informational areas will suddenly loom as being uncharted. To where you could launch almost the same kind of adventure off into them. Letting the same kind of thing be engineered into shape. So that you could imagine us almost being able to write a guidebook for the leaders. As to how to, if some situation pops up, get the right thing to happen. And, you know, if a hundred years from now we get invaded from outer space by some weird looking creatures, we’ll know how to turn our cultural system on to find the best response quickly [laughing]. As an overall organism, you know? I mean that’s what a lot of this is, it’s really a whole group somehow exploring what can be done very quickly. They haven’t fully explored it, so this will rattle on for a long time. But almost from zero there’s been this whole group running off exploring that.
So a lot of it is the networks give you the ability to get the large branching ratio, the short time constants – in a way, to some extent with the right servers on the network, you can create the artifacts necessary to serve as tokens. And they can be commonly – you know, all you’d need is one Dover, or one place that stamps out these things. And you just keep it busy printing out artifacts of the form appropriate for that arena and mailing them back and so forth, okay? And so if you can somehow get a system like that going in any new area. I don’t know what it might be, but just different kinds of things.
Then there’s another interesting property to all this, though, is that as good stuff is being created, the one thing that we don’t have a good system for is how to actually understand and identify what is the good stuff here. There isn’t embedded in the system some mechanism for actually, off to the side or later as it matures a little bit, to be the critics of the art that’s being done. And to somehow identify something clever that’s being done and bringing it to the attention of everybody. We’ve been doing that in an ad hoc way. Ah, that's the other mechanism I did not mention. All along, anytime any of the leaders in this saw any good work being done, we would bring it to the attention of everybody. So there’s a whole thing about you too can become a hero. So the people never know but that they might be one of the ones that’s chosen, you see. And it’s dynamite [laugh] we always pick two or three examples from the last whatever it was that’s done to show off to the next stage of the game. And so they see embedded in that that it’s all – people find that very exciting too.
C: And, you know, so it’s been a lot of fun. Well, you know, I’ve gone through this just in a sort of haphazard way. And there’s some ideas on where this might go. It might be interesting for me to try to predict a little bit, to see what I think could happen, and what we’ll try to make happen and see if it would. And then the other thing I'm really fascinated with is to try to understand better the abstract processes going on here. There’s a meta-level of things that I intuitively sense as mechanisms, but I haven’t given names to. You know, are those real or just my phantom things I imagine as I observe. Or is it the other way, am I seeing a mechanism and engineering it to happen. You see what I mean?
S: Yes, yes.
C: Don't know. It’d be fun to talk about that some more.
Conversation II, March 21, 1980
Filed on Maxc<Suchman>Con2.press
(S = Suchman, C= Conway)
C: [referring to transcript of 2/12/80] It’s interesting, I’ve read this over several times and found it fascinating in many ways, in that I see reflected back here ideas that by not having articulated them were not really clear in my own head. They’re there, and I may operate on them, but it’s really interesting to see it externalized. The other thing is I’ve read your thesis proposal, and I found that very very interesting in a number of dimensions. It would be interesting to know more about where that’s all coming from. Your own background. And in general, what’s Berkeley all about, you know, the anthropology department, what’s happening in anthropology today anyway, and how does this fit in. And then there are some side things about where little bits and pieces of this [proposal] have interesting reflections on some of this MPC79 experience. Some very specific things that I could mention, things that I happened to notice that I flashed as I saw them. You know, I see a comment and then oh, here’s an example in here [MPC79] of that phenomenon you’ve articulated as a general phenomenon. Which could then reinforce your view that that is something that happens. So I have this long list of things which is just random order things that at some time it would be interesting to talk about.
S: Good, okay. That have to do with that relationship. Of the two things?
C: Yes, that’s right. Between this [proposal] and this area [MPC79] as one of probably many example areas that one could look at. And so you might find that knowing this context I could articulate or point out things that are happening here that you might find interesting to just know about as examples. Interesting, kind of important examples. There’s lots of people – not a lot, there’s not as many people doing this [design] as there are secretaries operating typewriters and word processing equipment, but there’s enough, and they’re paid a lot, you know, and there are some measures of what they accomplish that maybe aren’t present in the other area. So there’s all that. Maybe if you could just tell me a little more about where all this is coming from, then that would help a lot.
S: Okay, just one comment on the relationship between the two, which I’m very interested in. One of the most immediate ways that that has struck me is the whole issue of division of labor. The kind of demystification that went on and is going on, making a multi-level process available to a person engaged in any particular level of that process. And what you win and lose in doing that. As a metaphor for organization in general, of any kind of a project. And what, if you’re thinking in terms of computer systems and what they could offer people who are working in a very large organization, what that kind of transparency could buy you, how that might actually be reflected in different ways of organizing or dividing up tasks, that sort of thing. So that’s really interesting to me, that angle of it.
By way of the background that you were asking about, the department at Berkeley is a very traditional department for the most part. They made their name in the 30’s when [Alfred] Kroeber was around, and they’ve been resting on that, I think, ever since. So they teach quite a traditional version of anthropology, which is a very strong functionalist approach to things. In the ethnographic tradition, the tradition of going around and describing communities, it means that you have certain ways of slicing things up, functionally, in terms of economics, politics, religion. And then you’re interested in the interrelationships of those things. And basically the way you find the coherence of things is by showing how they are all mutually supportive. I find reading those kinds of ethnographies to be wonderful reading. It’s very entertaining. But you have no idea what you’re reading. You have no idea what the grounds are for what you’re reading. Because by the very nature of doing that kind of an account of something, in a way it’s a tautological explanation of things. If you say this thing has coherence because it has coherence, because everything fits together.
So I’ve been more interested in a kind of counter wing, which is really much more active in sociology than in anthropology, which is interested in stepping back and saying that all of the things that social scientists have traditionally taken to be their topics are actually things that everyone has to deal with. The whole project of making sense out of the world, of walking into any scene, whether it is very strange or very familiar, and finding the coherence of it, making sense out of it, is something which anthropology has sort of taken on as “this is what we, as anthropologists, do”, without recognizing that that’s what everybody does. All the time. And that in fact that is a major social phenomenon that’s very fascinating and that’s been largely ignored.
So then the whole thing becomes, what you’re trying to appreciate is the reflexivity of making sense out of things. That as a social scientist walking into a scene and trying to make sense out of it what you want to do is – well, l could use walking into an office as an example. If you walk into an office and you say well, there’s some sort of a system here. Things are being done in some kind of a systematic way, and I want to discover how that system is being maintained. You can approach it in two ways. The traditional way would be there’s this underlying body of rules and norms and customs that direct people in what to do. So then you can make a list of all of those rules, and you can show them at work. You can look at manuals and you can say that they explain why things are done in an orderly way.
C: These things, I take it, you look for in external artifacts. You look for a procedure manual, or you look for directives –
S: Right. You’d also look for norms, you’d ask people why you do things that way, and they’d say well that’s the rule. There’s a rule, maybe it’s an unwritten rule, but there’s a rule that you always do things that way.
C: Right. But there would be a very common articulation, rather than a guessing and a mumbling and all – so a norm would have, it would commonly happen but also people would talk about it in sort of the same way.
S: Right, you'd be looking for things that people commonly point to. So then you say well okay, there are these conceptual objects in this domain that are rules, and you can come up with a list of all of the rules. And then if you’re a very traditional social scientist then you may say well now my job is done. I’ve come up with sort of a taxonomy of the rules of this place, and that’s how it works the way it does.
Now I say that the interesting thing is the fact that people tell you that there are rules in that place. And that the reason that there are rules is because people tell you there are, and they tell each other there are rules, and they point to the rules all the time and say yea, there’s this rule here and the other person says that’s right. That it’s this ongoing maintenance of that stability that’s really interesting.
C: That if you ask them what they do they describe it somewhat in terms of the rules. When in fact those rules may be nothing more, in many cases they may not be even applicable, but in many cases they’re just constraints on what’s done so they’re not really telling you what’s done. They’re somehow putting bounds on it, or trying to frame it, but not actually telling you what it is. Interesting. That came through very interestingly in here [proposal].
S: And the fact that they tell you there are rules – the fact that everybody in a particular office will tell you there is this rule doesn’t mean that you as a social scientist should say okay, everything happens this way because of this rule. What you as a social scientist are interested in is the fact that everybody in this office tells you that there is this rule and that it’s the basis for what they do. So it’s trying to step back and see that all of the things that social scientists have been relying on as resources for explaining why things are the way they are, are the same things that everybody relies on as resources. And that what you want to do is step back and make that whole phenomenon a topic. What are those resources, how are they used, and how do they maintain this apparent stability? And that’s not saying that things are really unstable and that they only appear to be stable, but that stability doesn’t exist in the world as an “objective fact”. It’s something that, people have to make all the time.
C: Interesting. How is that kind of thinking viewed at Berkeley. What’s the feeling in that department about those sorts of studies?
S: It’s viewed very skeptically. There are a variety of criticisms, one of which is that you’re getting preoccupied with – for instance if you’re looking at a massive organization, if you’re looking at corporations – one of the critiques that I got during my Ph.D. orals was that the order of phenomena that I was talking about looking at was somehow trivial. That I was going to get lost in these details. Because I claim that the only way you can understand the stability of this huge institution is to look at the details of the everyday activities, that I’m going to get lost in these details and I’m going to lose sight of the really macro, the large-scale functional interrelationships between corporations and government, for instance. Most people talk about that at a level that requires getting way, way back from the everyday details. And I think that’s an important view. But I also think that that view without a view of the everydayness – the farther back you step from things the more of a gloss, of a generalization you get of what’s going on.
C: Yea, even if you did discover relatively quantitative relationships that would give you some ability to predict what’s going to happen, you might not be able to predict when those are going to change unpredictably, because something down in the system, down in the elements you don’t know out of which it’s composed, has triggered some new phenomenon.
S: That’s right, you don't have a good enough grasp –
C: of what it’s really built out of.
S: Exactly. So you can come up with very valuable and interesting stories about it, but they are somewhat ungrounded.
C: That’s interesting, so do people who have just the ethnographic focus, they basically are recorders, they’re like the early naturalists, is that what it’s like? They're very much focused on leaving a precise record, they aren’t really trying to make any deep interpretations necessarily.
S: Well they come up with functional interpretations. In other words Malinowski looks at the Kula ring, this trade ring in the Trobriand Islands, and he finds all of the functions that that serves, in terms of economic functions, political functions, religious functions. And you read his account and it’s fantastic. It’s this incredibly elegant fitting together of all of these diverse things into this wonderful story in which everything makes sense, and everything is contributing. And what Malinowski is responding to when he’s doing something like that is what came before him, which was the notion that you walk into a scene like the Trobriand Islands and you say my god, this is chaos. You know, because I don’t immediately see what’s going on it’s total chaos, these people are savages, they have no social order. So this is a strong reaction to that, and it says look at this order. Everything fits together perfectly. And now this is sort of a counter-reaction to that, saying that’s a little bit too nice, too clean of a picture. That if you get up closer things are a little bit more ad hoc than that, they’re a little more haphazard and that’s what gives life that nice open-ended, unpredictable quality.
C: That’s interesting. Looking for the mechanisms is always interesting. I tend to do that a lot, sort of speculate and try to find basic things out of which you can build the bigger things. Really see what’s happening.
S: Maybe we should go over some of the specific [referring to list of comments].
C: Yea, in fact let me show you something interesting. This is just an example, and I think we established enough context last time that you’ll be able to really recognize how strange this is. In here [proposal] you refer to this study by Garfinkel about the “curious absurdity” of studies of work [reading] “with respect to office settings, the observation is that while the research offers us occupational demographics, blah blah blah, the social scientists consistently treat as outside of their research domain the work itself.” It’s interesting now, what I’m thinking about there is a funny mapping, I’m not being very articulate, but it goes something like the following.
Let’s take not the social scientist studying what goes on in an organization, but let’s take the computer scientist system analyst who is trying to create computing aids and assistance things for designers to use. Any kind of designer, there are a lot of different kinds, but let’s say integrated circuit designers. There’s something of probably the same process that goes on. Now these people, the computer scientist that is trying to build an aid is going to go through an analysis phase, okay? Part of building the aid they’re trying to, based on what they’ve observed and the rules they’ve written down, create a tool that will help the person change their practice to do it better. And so to some extent they do an ad hoc sort of social science. But what I find interesting is that the style they choose to use is just like the one that the real social scientists actually use, which appears also to be ad hoc [laughs]. Okay? In that they just, as we all do, simply try to figure out what’s happening and describe it. So they ask people what they do.
Now the thing about design is that what we observed, see we were observers also. And what we observed when we first wanted – in fact our intent when the group formed was to go do that sort of thing. Back in the old naive days. The notion of designing and learning about it, talking to people, so that we could make better aids and understand what it’s about. So we were sort of amateur design aid builders, just to learn ourselves what that was about. What we observed the real design aid builders doing, because we looked at what they were doing first, to see what design aid building was about [laughing] you see how the thing goes, it’s really weird. What we observed was the following – I wrote this down.
Designing design aids, okay? These people made an assumption. They assumed that – and this is a funnier mind trick and let me get down into it. The assumption that we had made, and that certainly the other design aid builders who were already building them made, was that somebody actually knows how to design. Now [laughing] all these people are doing “it”, okay, and we’re going to watch them and then we’re going to make a tool to help them. But the key assumption there is that they know what they’re doing, that they’re actually doing a thing that’s well defined. That’s the idea of the rules, there are rules and it’s all very – And the other thing is, since the design aid designers themselves didn’t know how to design, there’s no way they could notice that the other people didn’t know [laughing]. See the fortunate thing we did was before we went to talk to the design aid designers, we struggled to understand what it was you actually tried to do, you see?
S: When you were doing design –
C: When you were doing design. And so we noticed that the design aid designers didn’t notice that no one else knew how to design because we found out that nobody did either [laughing]. I don’t mean to really be flip about it, but there’s this funny multiple set of things there. About assumptions that there is order, and that you can discover what it is and that you can in fact constructively augment that situation by something that you do. So that you in some sense have the ability to make it even more orderly. You know, rather than being part of a much different process that’s actually going on that we’re all participating in. You see? Now that would be the one that would be fun to find out what it is.
S: Right. That brings up the whole question of what does it mean to know the work itself, the actual work? It seems that the more distant one is from the actual work, the more it has these smooth contours, the more one assumes that there is this, uh –
C: That’s right.
S: And the closer one gets into it, it’s not. I think when you get an intimate knowledge of some kind of work it’s not that it loses all of its shape, but it takes on a very different kind of shape. So then as a social scientist you’re faced with, as you were – you’re faced with the issue of how much I need to know the work of the people that I’m studying. For me that’s a tremendous issue here. More of an issue than I ever dreamed it could be. Because the work that goes on here at PARC is so incredibly obscure to me.
C: Right, I understand. In fact I won’t attempt to go off and try to really explain that at all. I mean there’s so many dimensions to just the organizational structure, political structure, the personalities involved, all the past history varying amounts of which are remembered by different people, and it’s very complicated. But the thing that is interesting though, and it really struck me, that this work must be very difficult because in fact unless you can really get very close in and understand what’s actually going on, or struggle to understand it, as opposed to how the people describe it, you can’t really get at what the underlying processes are. And in fact when I talk about design one of the things I don’t mean to do is to convey that there is some, you know, somehow handed down from on high in golden tablets there is a best way to do this stuff. And so in fact what we saw is simply that, when I say that nobody knew how to design but the design aid designers didn’t notice it, all I’m really saying is the design aid designers were not concerning themselves with the content of the work. They were simply trying to somehow think of the procedures and help the procedures go better.
Now that doesn’t really do anything because the procedures aren’t what’s really being done. And making them go better doesn't really change anything. What we noticed was that, based on a totally arbitrary other way we had learned about this, was that most of what was being done at different places was being done, first, differently in different places. No common lore, okay? And if we looked at how it was being done differently in different places, we could see certain common mappings that would be simplifications in all the domains. So that you could have, number one, a simplification in each place. And number two, a more common culture. And so then it became a game to inject that different thing. Then from our perspective say now look – and this is a game we play – we’ve created these design aids, look how great they are. Everything goes better. Fantastic. We can teach this stuff easily, everybody’s using the stuff, it all works great. And in fact the funny thing now is I can realize that management, in looking at that, they also think things go by rules. There’s no way they can understand the process by which that got generated. Somehow it reinforces the notion that we ought to spend a lot of money developing design aids [laughing]. So they’ll go hire a lot of design aid designers.
And in fact the industry was really funny in that – here’s the other thing. This whole integrated circuit thing is very strange. I think at the time we started this there were throughout the whole industry more people concerned with designing design aids than there were doing design. By a very large amount. Okay, so there’s another message there. Somehow, you know, this confusion about what it is you should, how you should proceed. I don’t know. In fact it’s really hard to imagine how do you – it’s hard for me to think about what goes on in sort of the generalized office and how you can assist that. Although you might think of a lot of specialized areas.
The other thing I know I wanted to get into. There’s a notion about the rules don’t describe really what’s going on. There’s some script that people are using, in situations that are stationary, that aren’t evolving. See design has a funny quality to it in that something new is getting created, that hasn’t been around before. And so there is already a kind of a meta-level to it. There’s that new thing happening and the effects it causes. The design activity, as opposed to repetitive, more routine activities.
S: Maintenance activities.
C: Maintenance activity, yea. I mean even maintenance activity – if somebody has a certain class of machine and they maintain it, they’re always doing new and different things to it. But that’s different, somewhat different than an environment where you’re designing stuff. I can think about this design arena, in fact, almost more easily than I can about the other. I don’t know why.
S: That’s interesting. I have this – it’s almost like I’m deliberately interested in what is apparently the most non-problematic aspect of – because I have this, at this point unaccountable conviction, that in the so-called routine, maintenance activity that goes on there’s a lot more active generation happening than is apparent. It’s sort of a sense that you want to tackle what seem to be, in the case of large organizations, these absolutely bedrock, massive, immovable objects. And try and get in there, the demystification being in this case to show that they’re not that way at all. That they’re always being held together by the people that are involved in them.
C: Right. In fact you know that makes me think – to understand a mature, stable organization that’s running along, to really understand it, I would think, down at the more detailed level, to really understand it you have to understand the process of generating such organizations. You have to understand how they seed and survive. That it’s within that process, the generation of – and I’m thinking of this notion of a tribe – I think to some extent there is a tribal formation there that maybe we remember, maybe not, but some artifacts of it continue. And then as the organization becomes more complicated it spins into more tribes, and these take on partial identities from other existing things. I mean there is a personnel department, this, that and the other thing. And they give themselves names and all, but in fact the people in them may not know anybody in any other personnel department. And they have some local phenomenon that has come out of that tribal evolution.
The reason I think it’s easier for me to think about design, and maintenance – and I’m thinking about the design and maintenance of a computer system, I have that in my head – is that, and this is way down many levels, this is not the generation of organization but just an artifact – it’s that the people that create the artifact know what it’s made out of. They know the rules for putting together the thing. Now let’s suppose you make a lot of copies of those and launch them into the field. The people that maintain them, they are like the person coping with the little group that they’ve just entered. They’re coping with this machine they’ve just come in contact with.
S: It was there before them.
C: Before them. Now they have to somehow, without having any idea of how it’s created, or what it's really made out of, they have to cope with it. And interactively do the right thing. Or else they and the thing don’t last. And there’s some sort of discovery process in that, where they in their head get some rules about what it is, that’s not at all what the designer made it. So moving up a level to social systems, it’s sort of like there is a process of design, that runs by some natural process. Of design and creation of these little tribal structures. And then once they’re in place, and they’re stable if someone enters it, they interact. And in fact there’s some evolutionary process. If you’re looking at organizations that have some stability, the fact that they’re there means that they’ve survived some selection process, multi-dimensional, that kept them far enough in between hard constraints that they were able to still rattle around. A lot of frills and things there that don’t have anything to do with the selection but were just instantiated like, you know, a painting has various little frills.
But that’s the process I’m interested in. That’s why I think about the tribal, the artificial stimulation of the formation of a tribe. And then the ability to try to see can I move that this way or that way. How do I have a system where there’s gain in it, so l can inject a signal and it will amplify it and adopt it. What it really is is – I’m sort of rambling on, but I have this sort of intuitive notion in my head. I don’t really talk to people about this stuff – this sort of intuitive notion about what are the basic elements out of which you build all that. If you could even back up further, to just animals, what are the – I don’t know what the units are out of which you could really think of making it, but there’s something you could call an elementary sort of particle of a culture. Where once you inject that if in fact it starts getting passed around – it could be stuff symbolically represented, or it could be just speech that’s repeated, or it could be just some imitated thing that one animal sees another doing and imitates. Some means of, you know, randomly something happens, some other animal sees it happen and imitates it, and if it feels good it’s repeated. And you could write down the rules of how that might diffuse. And somehow out of that you might be able to build –
S: Uh huh. I have a sort of general, overarching concept that I think for me at least covers a lot of what you’re talking about. And that’s membership.
S: Now what all is constituted in that is the interesting thing.
C: Could you describe that to me? And let me map it against some things and see how it, you know –
S: Well it’s something that I carry around with me all the time and use as I look at the world. And it has to do very much with all of the things that you were talking about in terms of creating the tribe. With the artifacts, with the common, shared terms which makes the thing mutually available to all of those who participate, the initiation into it. If you think of the world as made up of – that every situation, or every setting, is constituted by members, then you can have more and less temporary, you obviously have multiple memberships, some are more important than others, but the whole notion of generating membership is connected in, I think, with that. And membership encompasses all those ways in which you know what’s going on, along with some group of people with whom you share this thing, you all know what’s going on. And I think there are tangible, discoverable ways in which you can find what that “knowing” is. That have to do with, you know, you share these artifacts, you share these experiences. You have these lines of communication.
C: Yea, that’s really interesting. In a way you could look at that as the fact that there is a common body of externalized stuff, which if you identify with that means you are a member. A way I would think about that is there would be some natural inclination to fall into those kinds of memberships anyway, because there’s in a sense survival, individual survival embedded in that. The fact that that stuff is there means that it works, in some sense. And if you can just imitate the scripts involved, and the various things involved in membership then you’re part of a larger group that probably has much more –
S: And it’s identification too, I think. Locating yourself in the world with respect to other people, which is a large part of what social life is about.
C: It’s funny, I tend to think about that in a more abstract way, without putting any subjective notions on it at all. In other words trying to think about it in the sense of what makes the process run, looking at it from the outside. Almost as though this collection of elements for any one isolated tribe or whatever, that that’s the culture and that that’s the thing that – in a sense you could unplug any of the individuals out of. And in fact the situation is only stable, and you only have something there if that whole collection is able to integrate new individuals in and be stable as people come in and out of that. That the thing is external from the people.
S: It’s public.
C: It’s public, right. And forgetting for the moment your subjective notions about membership, your internal psychological subjective notions as opposed to the whole thing.
S: Yea, I think the only sense in which you need to take into account the individual’s point of view is to understand how in any particular instance it happens. That the individual in the course of becoming a member has this process that they go through. And I think you can think about it very much as a social, a public process not an individual psychological process.
C: Right. In fact there’s a funny thing about that in that – let's suppose you could get a much better feeling, in the abstract, of some of the – could even quantify some of these things that happen in this public structure. Although that seems to in a sense dehumanize it, make it less interesting, if you can understand those well enough that you can change things. Like time constants for things happening. Then in fact you have the ability for individuals to impact processes, which would then in fact make their lives more interesting as individuals. That’s the part of it I find sort of fun about the whole thing.
Let me take an example. I’m sure there’s been a lot of stuff done, and you could go off and do all sorts of things, but just in the most abstract sense what I think is different about these networks is something like the following. That in a sense one person can be in contact with many more people than they could otherwise. In routine, day to day contact. So the span of coverage is very large. And the time constants are very short for causing something to happen. For news to reach you, or for you to get a message to somebody. So in a sense, independent of the kind of tribe or organization or anything, you could tend to make things go faster. Everything can go faster. So you could try out experiments just to have fun, or you could actually cause something to happen that you wanted to have happen, perhaps. In terms of really forming a new tribe. At a rate which is faster than the normal sort of generational rates.
I think people think gee, if I make something go four times as fast, what does it matter really. But when you’re talking years, those start to be things that change the whole quality of everything; because, you know, human life is just so long, and there’s certain phases of it and – And so the thing I find interesting is to see okay, we have this network out there and we have a known collection of tribes out there, and we might want to form a new one or we might have just the current situation. And let’s suppose that you’d like to interject some new elements into that. Get them to just rattle through it very quickly and become adopted into the sort of process of – the actual work that gets done; the way people actually do things. And you can just make that happen very fast with the networks. It’s obvious, it’s sort of intuitively obvious, and yet in fact that leads you to think okay, think of the whole network thing, and all of the memory that can be associated with it, the file servers, the actual equipment that can be repositories of information, the servers that can generate artifacts and all that. That sort of creates a very, very new and different cultural infrastructure, is the way I would think about it. That you could think of in place of the infrastructure provided by letters, the postal service, telephone, all the other things that we now use for enabling communication and transfer of messages and artifacts.
And the thing that's neat about it is you could configure it any way you want. You could try out different alternative variants of that, almost run big experiments, and see what effect these things had on the rate of propagation of new elements of culture. On the rate of awareness, imitation and selection. You could study the processes, maybe, the fundamental processes. And what I’m really saying is that something like the creation of a new tribe somehow starts from some one random new element somewhere, okay? Some one thing occurs – in fact the fact that there’s so much randomness, because we’re so in between the constraints in what we do each day, means that occasionally someone does a new thing, someone else observes it, etcetera, etcetera and off it goes. And so it’s almost as though the process is something like everything in between the constraints, but what’s actually done is a function of what has popped up that’s been selected on somehow and retained in the group as a whole.
So those are points that are interesting to think about. And to some extent, whether you see anything real in there or not, as you look at it you can imagine operating on a set of things that you’re observing and interacting with to see if you can cause things to happen differently. So what I find interesting in observing things is not to look for the rules, but to really look at what’s happening. And then I find it very interesting to think now I’m going to see if I can actually change that. And that’ll confirm – I have a deeper understanding of what’s happening if I can change it. And the thing about being able to take this whole darn electronics thing [laughing] and just, with little resources, just start sending messages and just building out.
S: So it would be interesting to try to put your finger on what those things were that enabled you to make those changes. You know, clearly in all sorts of ways the network was critical. And if you think of it again in terms of membership you have – previously each level of this design process is a domain of membership, which maintains its distinctiveness and its isolation from every other domain. As opposed to, in this new kind of tribe, membership is constituted by having a line on the entire multi-level process. That’s what defines your membership.
C: That’s right. In fact the initiation rite is to have done that. You have to have an artifact that is a product, that shows that you have done that. And if you don’t have that you’re not a member.
S: And that whole transparency is enabled by the change in the communications that you have available to you. Because I think that largely, up to this point, up to the point of computer communications, membership has been a geographical phenomenon. It’s been very much tied to physical proximity. And suddenly, with the network, that’s totally –
C: Yea. Before you could almost think of it as everything’s happening two-dimensionally. You can look at things that you observe almost like diffusion processes. The mental model you could have of little elements rippling out would be somewhere it starts, and it diffuses out. And now it’s all inter-twined.
But that really brings me back to the notion that in fact the thing that’s different about the computer communication situation is simply how you can think about very differently mapping that. In other words the fundamental entity involved in those processes is just the branching ratio and the time constants. That’s why I keep coming back to that. You know, people attribute too much magic to the computer. It’s not the computer, it’s that it radically changes the people you can communicate with and touch and talk to. Then it’s unfortunate in fact that the machines are actually anthropomorphized, and people view it as doing something for you, because they don’t visualize reaching through it and shaking hands with somebody.
S: That it’s them who’s doing it.
C: And so people in a sense then don’t get the idea of using it that way, you see? Except as, again, a tribal phenomenon that happens randomly like the CB culture. There’s still a geographic thing there, they do it only in a sort of limited span. The HAM radio people do it, but within a limited domain of message sending. They talk about HAM radios, you know.
C: Rather than everything.
S: It’s not a means to get at these other activities, it becomes the thing itself.
C: Yea, so it’s really funny. I think if you look at the effects of the postal system on science in the last century – a dramatic effect. Because that’s the funny thing about scientists as intellectual entrepreneurs, they sort of use anything to do their thing and are very excited about that. And also I think – okay membership is interesting there too, because there aren’t great densities of people that are going to be really up to what you’re doing, you know, you’re in some area where membership is so specialized that they’re scattered around, there are only a few of them. So then anything that augments your ability to communicate makes your thing go faster.
So you could imagine that the computer communications networks are like another wave of the effect of the postal system. Where, you know, it had a big impact on increasing the activity, I think in things like physics especially in the last century. And astronomy, also. People could send pictures around – a combination of the postal system and the photograph. You know, you get these weird things coming together. And that’s sort of like what we’re seeing in this stuff. Another layer of that. And all it really is is this ability to interact with anybody in the world that you want to. With a certain branching ratio and certain time constants.
[turning back to notes] Oh, that’s another thing. What I was thinking about was that if you try to imagine, you know in your head you can create an evolutionary thing about tribe formation. It’s almost like tribe formation is a natural process. That process itself is natural, okay? Any excuse will do. That’s the funny thing. So to some extent there’s the remembered thing, and how that remembrance is physically realized is hard to say, but somehow certainly within the civilized world there’s this tribe building. So you wonder how – well anyway, that’s interesting. And tribes trading with each other, trading is interesting. That whole process.
The thing I wonder is how you’re going to proceed. What your plans are and specific things you’re hoping to do. Time frames, and where you are in that [thesis] work now.
S: Well, I’m at a critical stage between planning for it and talking about it, and actually getting down to doing it. I did a project last summer up in the accounting office on the third floor of PARC here. And I discovered that that is a really interesting setting for a lot of reasons. It’s a small group of people, working together. There’s the division of tasks, but there’s also knowing each other’s work and consultation around that. And then there’s Maurice who knows the place intimately, he is very creative about the way that he does things, and he is a native theoretician, he loves to theorize about what’s going on. He was very pleased with what I did over the summer and is excited about the idea of doing more.
C: I’ve heard Maurice give a presentation, and he’s really cool. He’s really –
S: Yea, he’s tremendous. And because of his competence he has an openness. He’s totally confident of everything that he’s doing, and therefore he’s willing to show it to you.
C: So you can really discover what’s happening, at some deeper level.
S: Right, you don't have to get embroiled in interpersonal paranoia and politics. The other thing is that there was just a central edict about how every Xerox location had to put together a compendium of their procedures. And so Maurice has just put together this compendium of accounting procedures which provides this very interesting artifact. It’s like a natively generated version of what many computer scientists who are doing office modelling and procedural description are involved in.
The latest thing is I’ve been spending a lot of time talking to Rich Fikes and Austin Henderson. They’ve bought this argument of mine that while people who are doing office modelling and system design tend to say that what this work is is doing procedures, they see the centrality of procedures, they ignore the work that’s actually involved. So the issue is how can you build a system that’s going to get involved in the real work. What can that look like. So if I go up there, what I’m looking for are what are the resources here, in what forms and when do they actually appear and enter into the work, and they meanwhile are putting together a system which at this point is a way of representing what goes on, if we decide that we’re going to pick a procedure that’s done up in accounting, and there are some nice candidates, and we each go at it in our respective ways, communicating our progress, the interesting things that we’re coming up with and our troubles, and try by bringing our focus to the same thing, by bringing our respective methods to bear on the same thing, we can really start to get at where the conflicts are, where the cross fertilization can happen, and all of that.
We have approached Maurice about this and he’s very agreeable to it. So the next thing is for me to get my strategy together for more observations, and they meanwhile are working on their thing, and we’re keeping track of each other. So the notion is to start upstairs, start with one thing, and then there are all sorts of ways in which it can blow up in terms of how this procedure interfaces with these, and these with the purchasing people, so that has a lot of dimensions. And then I’ve talked to Bert about the possibility of other sites. There’s the Diablo plant, there you’ve got a manufacturing operation. There’s the Oakland Branch, you’ve got a sales and service operation. So I’m going to start upstairs and work my way out and do visits to these other places to see if I want to do a comparative thing.
C: Interesting. It seems intuitively obvious to a lot of people in SSL now that there’s a whole set of issues involved here that are actually quite fundamental.
S: I think it’s very much along the lines of what you were saying about doing design aids without knowing the work of design. I think that at this point I can’t know the work of doing computer systems design, it’s just too far away from my background, but I think I can get to know what the work is that presumably this is being done as an aid to. And act as a conduit to bring those two things together. Although I frequently have the sense that my lack of knowledge of what programming and computer science is about is – I have mixed feelings about it. I think in some ways it’s an obstacle and in some ways it’s an advantage.
C: Yea, well it’s funny, I think that over the next ten years or so you’ll more and more see a compression and simplicity of the process of once you are able to rather crisply specify what you want to do, getting that embedded in a functioning mechanism. That’s really what’s being discovered now, largely. The mapping of the software functionality down through the hardware is what’s happening in this area, and similar things could happen at the next level up in software. And so in fact what seems to be happening is that you end up with the problems all being in the area of coping with complexity and what do you want to do anyway and what really happens anyway and why are we doing all this. That’s where it all ends up being. And just getting some shared understanding of that would probably be very valuable. You know, just some insights that could help a company in just the slightest way tweak the efficiency of various things. It’s ability to cope with change.
That’s another thing I find interesting. You use the word stability, and I know what you mean by that. I tend to think of everything being highly randomized, in between some constraints which make everything appear to be stable, so it doesn’t just, the whole system isn’t as random as the things out of which it’s built, slower time constants, but in fact the interesting process about the overall system is how adaptive is it to obstacles it encounters. And how does it discover, as an overall system, how to do things better. Cope with change, be more competitive. And that's what I find really interesting. How does one army win over another one, when the game is not defined that they’re thrown into, and it’s all life and death. You know, those interesting processes of people discovering, sharing quickly, and all of that. How does that all work. And so to some extent – this thing about knowing, it’s not clear that at any level you really know what’s happening. You can just look at something happening one way and make the same sort of input output function happen differently, happen using a different mechanism and see that it’s done quicker, or takes less people, or it’s more interesting to the participants because they’re all happier, even subjective things are possible.
The other thing about this whole design thing – the thing that struck me is that once people are in what you would probably call a stable situation, know what they’re doing, and know the rules, and they all have this layering of memberships, it’s very very hard to get those people to reintegrate that whole process. It’s a whole lot easier just to start over again and seed a different tribe [laughing]. And you wonder what does that mean for organization. Now I’ve thought about that in a human sense. What’s another kind of stability you could have?
Suppose you did it this way – this is almost what happens around here in a quick thing. What are all of the alternative organizations that might work and still be humanly acceptable, okay, rather than quicker obsolescence and earlier retirement [laughs]. You know, you’ll take longer to educate, you’l1 work for ten years, and then you’ll retire, right? That just seems not to be too cool. Could you work to where the average person in their educational process, and the leaders and the people that invented the next curriculum and all of that, were in a completely different kind of situation. Where what you learned, depending on your role in all this, was how to learn quickly and fall into a new tribe, and find that exciting. And in fact the society would put a little pressure on – there would be embedded in little messages that it’s really not interesting to do things forever.
[End side 1/side 2]
C: [continuing] That whole process of tribe forming and activity and then disbanding could be something that’s very dynamic in the whole culture. Of course it is now, but over time constants that are on the order of generations, and therefore the process isn’t as obvious. Right now I know that older engineers designing in the old technology, not the new technology using the old ways of designing, but just doing the old technology stuff, can be easily taught this new stuff very very quickly. Because it’s been simplified. And in fact they find it fantastic and thrilling to learn how to do that. And so you could almost imagine if in fact everything were wired so that you were more flexible in your ability to choose the tribe you entered, and how long you were there, and then when you left it, the whole thing would be much more dynamic. So that when you generated a new tribe it didn't mean that thereby you now had to figure out what am I going to do with the five tribes it’s replaced? That somehow you get stability by getting the rate of generation right.
S: One way of attacking that, it seems to me, is what is the relationship between these five tribes that you’re displacing and people who are in your new tribe. What are the commonalities between their respective memberships and what are the differences? For some reason these people, and I think this is true in all areas where heavy specialization has occurred, are invested in certain things which make them very unamenable to the kind of change that you’re talking about. Now for some reason these people over here are invested in precisely the things that do make them amenable to change. How can you make it so that part of the definition of membership is that you’re engaged in this process of change rather than that you’ve got your turf and you’re defending it.
C: That’s right. This is fascinating. Let me go back to this example – let’s suppose we have the older engineer, who is currently designing in what we call TTL parts. Instead of doing integrated circuit design and putting a whole system on a chip, they buy smaller functional chips and wire them together. There’s a catalogue of parts about that and they’re working down here in one of the Xerox companies. Okay, they’ve got a book, in fact let me just show you this because this is an example of what everybody’s been doing [getting book] for years and years. There’s a whole set of books, Designing with TTL Integrated Circuits, these are all made by Texas Instruments, and there are books that show there’s only a few little things, a few little transistors, and all this data, and all this lore, just on and on and on it goes, okay? And in order to use those you have to have this book and you have to know about wave forms, and a lot of complicated things; and you have to wire them up right, and you put them in boards – if you look in an Alto that’s what you’ll see. You’ll see every board in there has a whole bunch of these in it, but each of those boards is really about as much as one of the new kind of integrated circuits.
Now here’s the funny thing. If you were to go to these people and say the way you’re now designing with TTL integrated circuits is very inefficient, somebody’s found a much better way to do that – not design integrated circuits, but design using TTL integrated circuits – that the way you write down manually how you hook the wires together, and the equipment you use to test the things, and all of the stuff you’re using to do your current job, okay, there’s a better way to do it, they won’t buy it. They’ll resist the hell out of it. But strangely, these people, this same person doing this, if you say I’ll show you how to design those things, how to design an integrated circuit, and oh, by the way, you can put all of what you now put on a board into one of those, they will go totally bananas to want to know how to do it. Okay? And it’s really the same thing, but it’s so differently done that they’ll join the new tribe.
And a whole new set of primitives are given to them, that to some extent – okay, the quality here is that the primitives have this property of they know that it does the same as they’re doing, the function is the same, but it looks magically different. This notice of magic, the notice of the new artifacts, is all through this stuff. Whereas if you change simply the procedures by which they do this [TTL design], none of the artifacts change a lot. It doesn’t seem magical, and they don’t want to go to all the trouble of learning that. They know all of this already, and they’re in a script type thing. And all their friends know how to do this. And in fact, once you’re in the script situation of a mature thing, part of being able to do it is that you know the rules and in fact there’s this idea of you feel funny doing it a little differently than anybody else. So that all reinforces its stability. And it’s an interactive, imitation sort of thing.
But the new thing is very different. It excites their attention, and they look at it and they know my god, that does the same thing. But then instead of being frightened by it, which they would be if it were complicated, which it is the way it is now, so it does frighten them, you show them oh, that’s simple. And they can learn it. Now I don’t know if there’s a message there or not, but I wonder, if you really look at all the things that go on in a company, all these different kinds of departments and the different processes, intellectually underneath it a lot of the stuff that’s being done is very very simple, conceptually. Fully instantiated, though, in all these little baroque instantiations of detail, it takes a long time to get your head around what you know in fact has got to be simple, you know like these planning and budgeting things. So if you were going to displace all those things somehow, how would you re-map, how would you present it to people.
S: Yea, so that instead of being too much trouble to learn –
C: It’s exciting as heck and they’re all off doing it. And see the thing about it is it’s – I think it’s not going to work at all, and you get almost no leverage by thinking somehow you’re going to apply computers to doing this traditionally organized thing, just better. You get the same reaction; it just doesn’t take. Even if you work hard to understand what’s happening there, you can’t get around these scripts people are familiar with, and it feels funny to do something differently. Why is it that, you know, you introduce a computer system, and everybody feels funny about the thing? You’re going to try to do the same thing you’ve already been doing, but now there’s this new thing you have to do, and it’s there and you don’t quite know what to do with it. It’s not in the script, or something.
Now the same thing works in this other stuff [VLSI]. It's like people laughing if you use the wrong colors. Already there are people being able to see instantly that they’re doing it like everybody else is [laughs]. And the trouble is, once that gets in there, you’re stuck with it. You better hope you chose the right colors [laughing].
S: [laughing] Uh huh, or else you’re going to have to think up a whole new concept.
C: That’s right. And so in fact a lot of thought has gone into this particular new thing, because a lot of people knew they were seeding something like that, so they really thought carefully, you know, what were the ways to keep it simple and all that. But I wonder, instead of looking at – this is just a way I would think about it – if you look at, let’s say, the maintenance function. It’s is not your design, and it’s not your sales necessarily, but it’s all your administrative kind of things – what I wonder is what really goes on there, in some generic way. That all the participants could learn to do and do very well.
See the funny thing about that is there’s also, embedded in that, a process of design. There is a design process embedded in the administrative thing, and that is that there are continually changing external factors and forces, government regulations, all kinds of things, forces for a continual tweaking of the whole organization, okay? So that there’s always going to be change in the tax laws, in the forms, every time you deal with new vendors you’ve got forms problems – there’s all this stuff that’s continually changing. And probably a lot of the baroque stuff you see is a remembrance of some minimal amount of that that’s still hung in the system somehow. And if you could just get some idea of the generic process across that. What are you really doing in essence? And have the people turn more into designers. Turn more into understanding how to create the new stuff that copes with the change. And run it. And have that be viewed as a process that they are managing. I don’t know, does that make any sense?
S: Absolutely. And I think that what that requires in some sense is the kind of transparency that you’re making available. The kind of grasp of – you have your domain, but your domain is not this little piece of some larger thing, your domain is some kind of entity. It’s some kind of coherent activity in the world.
C: Yea; it’s almost like you are, at work, a member of a particular tribe that has a function, rather than you are having a specialized function. Now the tribe can structure itself in a lot of ways, but generally everybody is working toward something. I mean now we’re migrating from here to here because it’s springtime and we, you know, and there’s little detail in that and individual roles, but the whole group is doing this thing as part of a larger process and understands that, visualizes that.
S: But in order to maintain your design, your creativity in your domain, I think that the domain has to be parsed or given to you in a chunk that facilitates, that promotes that way of managing it. Which is why my sense is that using computers to come up with better versions of current ways of modelling the office, better flow charts, is not going to get at that.
Something interesting was done at Citibank, at least from the accounts that I’ve heard of it. It’s a procedure for issuing letters of credit, which is apparently a fairly complicated procedure and requires a lot of changing and updating. They had that procedure all chunked out so that this person did this, passed it on, passed it on, passed it on. So that the client comes in at this end, it goes through all these people, and the letter of credit comes out the other end. They reorganized that using workstations so that you have a relationship of person to client. And each person, their workstation enables them to compose the letter of credit, print it out, maintain the customer’s records, and as a worker you have your workstation, you have your territorial domain that has all of the material resources that you need, and you have your clients. And you handle their letters of credit, their business, all the way through the generation process. So that the membership is that you are all engaged in the same work, with different clients. Rather than that you’re each doing a different piece of this larger process.
C: Beautiful. Now that means that there’s a force, if anybody discovers a simplification that will help the whole group, it will take in the whole group.
S: Sure. And it will take in a way that doesn’t mean that if I change my little part then oh my god, everybody else is out of kilter. But in a way that here it is, I'm going to do my thing this way, if you’d like to do your thing that way too you can. So that the interrelationships aren’t such that everybody is constantly pulling the rug out from under everybody else which is the way, now I think it’s wired in such a way that any change has these repercussions whereby everybody else is – if you can wire it differently so that changes can happen more autonomously or something.
C: That’s cool. That sounds exactly right.
S: The only accounts I’ve heard of this, of course, are by Citibank. But it sure sounds great. It sounds like it worked really well. And they claim nobody lost their job.
C: In fact, there’s so many things that run through that. If you’re able to get a person handling the whole thing, and groups of them, then they can share common discoveries and you can really reinforce the notion of tribal membership. They as a whole group can take pride in this body of stuff that they were able individually to do all of. And the other thing that’s neat about that is you have a force, then, for simplification running. You really do. Because it’s to everybody’s advantage to be able to take on more if in fact the organization as a whole – obviously there’s a force for competition too.
The other thing I wonder is how do you reward people for injecting discoveries that then prove to be accepted. That would be another thing to look at. Because if there’s some sort of reward system – like right now the thing that runs in this thing in the network, there’s no money in it – is that it’s, the way people become discovered and identified. Is to put up some service. So in fact I guess what a lot of this is, it really ends up how you organize and structure the work. So that then you are able to put the machine infrastructure in place that then lets you run that new thing faster than it would otherwise go. So in fact the machines may help one person do the whole thing because they allow communication over a wide range and all that. They amplify your ability to deal with a lot of people. Rather than having things go through a pipeline. Okay, you might argue that a lot of the specialization comes from essentially the need for pipelining complicated things. So that’s an artifact of –
S: It’s an assembly line.
C: It’s an assembly line model, I found that very interesting in here [proposal]. Because another thing, there’s a political dimension to this thing in engineering that I think potentially is quite – you know, politics is part of all of this. And so what are ways that you could cause restructuring. Make it happen more easily. Let’s look at what could happen in engineering.
I found it completely dynamite, some talks I’ve given at schools, I’ll talk about things like the real artists in this stuff are in the back rooms in the print shops and don’t get their names on all the work. And it’s dynamite, I mean the students they all think I’ll work for one of those places [laughs]. And so the idea that you could take on the whole thing, and put your name on the work – That’s what I found interesting about this, about the assembly line. Is that right now the large organization views and uses a lot of engineers in exactly that way. You build the final thing on an assembly line; the work flows through. Instead of thinking of this as a social hierarchy – I mean, it’s made into a social hierarchy because that’s the way the organization casts it so that it isn’t viewed as an assembly line, but in fact the people involved [laughing] in fact there’s this stuff going in one end and it goes from station to station and pops out the other end. And those people are putting in their time, they’re chained on the line, and they’re fed and clothed and housed and they die, and you put some more in there and it just keeps running, you know?
And in fact to some extent, I think, engineering in this country isn’t viewed as art and architecture. I mean it isn’t viewed as that serious an intellectual thing, when in fact at some levels it’s really quite interesting. And if in fact engineers in a company were much freer to operate in the network in that company, to join a project, to be part of events that way, okay? To enter a new tribe of design that’s formed, or whatever, very freely and sort of independent of a lot of normally constraining factors, you might be able to get the work done more efficiently if they could do more of it individually and were more creative.
And you might be able to cause it to happen by this political notion, you see. By oh, at IBM the engineers work in assembly line mode. And they aren’t really engineers, and etcetera. I don’t know. I don’t know how you’d do it with administrative people. That’s probably a thing that’s worth thinking about some more. Is this thing about, even intuitively, just roughly, how new does a thing have to be before people will pop over into that? What do you have to do so that’s acceptable?
It’s strange though, you look at all the people in personnel, in accounting, administrative work – and I suppose this is probably applicable to a very very wide range of people in what is commonly called secretarial work – you really look at what a lot of them do and there’s some common underlying thing there. Of course all of that is political and economic dynamite too, in a variety of dimensions. Yet on the other hand, from Xerox’s point of view as a corporation, getting a handle on how to cast all this stuff would also be dynamite. I mean they don’t really have to understand it, as long as the procedures could be developed and the organizational mechanism so that it worked. That’s sort of like this stuff. They don’t have to know how the new design aids came in, that it wasn’t a design aid problem it was getting a new design process underway and getting the tribal movement.
[going back to notes] It’s interesting to have a more leisurely and contemplative conversation about this stuff, to think about what some of the more abstract issues are. Oh, another thing I was curious about – I’m very interested in quickly understanding where anthropology is. And so the question is how could I discover that very quickly. Are there any current controversies in anthropology, and can they be articulated? And where is it done, who’s really doing it, at what schools, and that sort of thing.
S: Hm, I’m not really good on that, because I guess it seems to me that that’s such a central preoccupation that nobody ever gets down to the substance of things. So I’ve ignored it to some extent. You know, the hot debates never seem to be worth debating quite so hotly.
C: Right, no, I understand. But I find it interesting to know what they are. Because those are kind of like people questioning the rules, and you may discover what the process is if you hear their talk about it.
[phone call interruption]
The question is really maybe not so much what are the controversies, but what do you see it is that anthropology’s doing. I’s the reflection of the same problem, of understanding what the tribe does that you belong to. You know while I was talking on the phone, I was thinking hm, that’s a funny question I asked. Because how could I possibly explain what computer science or electrical engineering are doing now. And then I think well, the reason I can ask you the question and maybe expect answers is that anthropology is a smaller domain. But I don't know –
s: Well, I guess I’ve been a sort of maverick or black sheep anthropologist, because I’ve made a point of not being in touch. And I think that’s a crazy thing to do. I think it would be very good for me to have in my pocket a sense of that. A sense of what my tribe, which I’m supposedly a member of, is up to.
C: Right. Then maybe you could contrast your views with that general group if they are quite different. I tend to be a maverick also; and in fact the way this whole thing got started, and Carver Mead, my collaborator from Caltech on this stuff, and a lot of the people involved are also sort of mavericks, in the sense that to some extent you can end up with an intuitive notion in your head that hey, what they’re all doing is kind of off track. And I can see a way I want to go, and I don’t want to get too deeply into that, or I will get my mind polluted with things that –that’s really what a lot of it is.
S: Given that I have a limited amount of time and energy.
C: That’s right. But then the thing is that if in fact you want your work to be retained, once you’re in it enough and develop it enough and feel good about it, have something to say that might be workable, you need to understand that if for no other reason than to be able to map the work either into it or around it. And so to some extent I’m just curious about, there’s the notion of anthropology and I wonder what that’s about. I don’t really know about that. I have intuitive notions about sociology that goes something like this. That there are these vast numbers of people that call themselves sociologists, and there’s just a huge amount of ad hocery, hackery going on. And you know, I’ve been listening, and I don’t hear anything interesting and [laughing] you know, and everybody does it anyway, sort of like a lot of psychology.
S: Well my feeling, I guess, is that a vast amount of both sociology and anthropology I actively dislike. But there’s a point to knowing about it even if you feel that way.
C: That’s right. I was thinking, you were talking about some stuff that Kroeber had published on trade cycles. and stuff. Is that –
S: Malinowski, yea.
C: Okay. The thing about that is that knowing about how to find stuff, just as a source of ideas, to give you ideas about experiments to do, or studies to do, or stunts to pull off. That’s what’s interesting about it. Okay now, you’ve gone off to get some ideas, now how can I go map that up against some stuff that somebody’s done that’s descriptive and see if any of it maps or something.
The other thing that I was thinking, in line with this sort of characterizing anthropology, is that this is minority stuff here. There aren’t many anthropologists around. There aren’t very many sources of information on that, so I’m very curious about what that’s about. Now on the other hand, for you there are probably a lot of people here that you can talk to so you can gradually absorb the tribal, cult-like aspects of computer science.
In fact, you know while we’re thinking about it, let’s view you, and sociologists, and some psychologists, and a few computer scientists, and some physicists and a bunch of other people. Here we have this collection of specialists [laughing] but if you study it closely, we’re all sort of poking around the same kind of stuff. So what I’m trying to say is could we discover what’s in common. Could we understand enough about what each other’s doing so that somehow when a new problem presents itself, we as a group can sense a new way of thinking about working more of the problem. I’m almost thinking it would be really neat if more people could be scientists in the sense of a couple of hundred years ago. When things seemed less specialized and simpler because there hadn’t been so much lore accumulated and the thing hadn’t all instantiated yet. To where people had a wider range of interests and were speculating more about phenomena they encountered. There is in fact this whole world of phenomena out there that not many people are looking at. I mean it’s almost like this tremendous zoo of stuff going on. Which really requires an understanding of some social science stuff, some computer science stuff, some other economic things, and it would – I don’t know if I’m making any sense, but it’s almost like we could run the same lesson on us that we would then want to run on the administrative people. It’s just a rough notion.
For example this cognitive science stuff fascinates me. You know here are a couple of little instances, and I see an example, and an experiment is done. It’s sort of like what I wish they would do is they would write an instant book about that so I could read it [laughing] and understand what it was. Do you see what I’m saying?
S: Yea, I do.
C: And you know, try as hard as possible to get the key ideas in a simplified form and put them in an instant book. And not care if anybody’s going to laugh at it. Because I would read it for sure, you see? And a bunch of others would. And if we could each do that. I mean you might find it interesting, a lot of this stuff [VLSI] – there’s very little content there. I mean it’s very simple stuff. In fact all this is, to a large extent, is just wires. I mean they’re just wires that conduct electricity. It’s just like, you know, streets and sidewalks and places that people can walk. They’re just wires that can conduct electricity, and switches. Where two kinds of wires cross is a switch. And out of that, by this interesting process of taking certain little arrangements of the switches, you can make a thing at a higher level which you can then hook those together so you can keep your mind around what it is you’re doing. And after only three levels of that you can just make any kind of engine you want [laughing].
I was thinking, what’s the role of research. That’s a point that might be interesting to think about. Talking about a tribe in terms of a research organization, it’s something like discovering the primitives out of which you could build the simplest spanning sort of hierarchical structure that would let you most easily explain and teach to people that are going to run enterprises – you know, we have to think of something that sounds cool – how to create information, how to share it, how to manage and operate enterprises. You know, what are the sort of primitives out of which they could see how to build their stuff and do things. So they could build their little tribal culture. What would be the artifacts that clarified that sort of stuff.
You could think of maybe research being the place that had the procedure in it to discover such things for new domains which involved people and machines and all that sort of thing. That’s really the problem that presents itself, is that the replacement of all of the procedures and tribal structure and aids and tools and everything starts with a new area and this simple structuring of the problem into that. And so it’s interesting then to think about research here in this company, in fact it probably should involve all these different kinds of people communicating with each other, trying to explain each other’s stuff, so that they can somehow work together in a collaborative way to discover what those primitives are and run experiments with them to see if they work and then go off and try to start a tribe [laughing].
S: [laughing] You have such small ambitions.
C: See the thing about Xerox that’s really wild is it isn’t going to be that long, pretty soon it’s going to be just all wired. Everywhere you go you could pop up images and text on the screens and print out stuff. So you have a way of leveraging.
Oh, the numbers are important. That’s another thing I really didn’t get into. We in this thing [MPC79], we don’t have to convince anybody to join the tribe. If you can just get a taker one out of every ten, but if the branching ratio is high on your efforts you get enough people involved and then if the tribe is sufficiently different and dynamic, you can actually then just get it to instantiate out. So getting something to instantiate out as an alternative way of structuring work would be interesting. You could almost imagine simply running it as sort of a social revolution that sweeps the company [laughing]. Why not? I mean that’s probably the way it will happen, actually.
Okay let’s hypothesize that what will happen as a result of these kinds of things going into place in all the big companies is the intended effect will not take at all. But in various different companies, in different ways depending on how they’re seeded, different social revolutions will occur. Depending upon which of those result in the most resilient, adaptive, powerful, competitive overall organism, that will be selected on, on average. So it’s something like that.
Now if you take that as a way of thinking about it, what you could actually do is simply go about to cause all of that. Now I know that sounds crazy, but I almost think that’s what happens.
S: To cause it in the sense of?
C: That there’s a process of individual people who might aspire to play leader, okay? If they sense the possibility of something like that they will simply try and make it happen. That’s sort of the natural tribe formation. So that somehow the role of leadership, and how that’s so embedded in the notion of tribes and all, is there are some kinds of people there that are always scanning, looking for the opportunity. And you can think about it that in sort of an entrepreneurial business way, but it’s almost like that’s the mechanism that gets the leader to take the risk to get the following, and if it works that’s selected on. And so there’s some mechanism there.
I just have this funny feeling about all these networks, that there are opportunities latent in that that will be discovered by opportunistic potential leaders.
S: Are you saying that you think that will be the case regardless or in spite of the way the systems are introduced? That by virtue of having them there, and having those people have access to them –
C: That’s right. And in fact the thing that’s interesting – there are some peculiar technical subtleties to the different systems that could either inhibit that tendency or make it more likely to happen, that has to do with things like why the Arpanet, and why the network we use here on the second floor and out through Xerox, is so different from the common experiences people have using computer timesharing terminals or working on IBM computer systems or something. There is an opportunity for the equipment to seem very transparent to the user here in a way that isn’t present on these other systems. There’s a machine independence on these networks, independent of the kind of computer you have you can talk with anybody else on the network. You somehow can see through the equipment.
Now many people who aren’t familiar with that environment, when they come and use the equipment, still perceive that they’re dealing with a machine and they don’t see through it. And so it depends on how many of these kinds of networks are out there. And of course those little revolutions could be minor. They could be minor ones within a personnel department, or a marketing arm of a firm. And so maybe typically they get formed and then ripple through that kind of organization and throughout the whole economy, so maybe you don’t really notice what’s happening. So what would be the most startling – and it’s only a potentiality, it’s not like you would predict it would really happen – the most startling thing would be if it actually swept like a company [laughs]. But then, you see, you can create fantasies like that in order to motivate you to actually have fun trying to poke around and figure out what’s going on because you might be able to do something like that.
That’s why I think it would be very fascinating – I wish I could talk John and his group into creating an instant book or something, where I could really understand about that. You know why, because by externalizing it so it isn’t just within their group, if they externalize their ideas so that you and I and some others could read it, then we could in fact join their tribe. And to some extent vice versa. We could share in their discussions in a deeper way. Map them up against our different experiences. So that’s why, you know –
S: Well, yea, it’s the issue of what’s the format for a book like that. You know, what’s the best level at which to present what you’re doing, for that purpose. Back to that whole question of how much you have to know the work to be able to connect with it in some way.
C: That’s right. It’s hard, and yet I’m not thinking of this as a very soft science area, we’re all going to sort of brainstorm or something, okay? I’m thinking that there are clearly very important lessons and discoveries in their work, which obviously any of us could understand if we simply knew the context and framework and terminology so that they could be fully understood. And I’m confident that those are important lessons for us to know about.
S: Yea, it would be amazing if you could come up, apropos of the framework that you provided in your project, if you could come up with a framework or format that would guide people in knowing what the relevant things about their particular discipline were for you.
C: And maybe people like that should have the confidence to feel hey, you know, we’re good enough why don’t we simply define what this is and see if it takes. See if it’s usable. See if it instantiates out.
S: Or even if you approach it within the bounds of say this research center. That you want a common template of some kind for collaboration. And everybody could use that as the framework for knowing what of their stuff – sort of what we’re playing around with doing here in terms of comparing the transcript and the proposal. You know, the specialization itself, and what are those common issues.
C: And there are points where, in fact if we start to get deeper into this you start to see there are places where in fact we’ll get into some particular technical things about computer systems that ripple back up to have a very important effect at the level of social interactions of people on the other ends of them. And yet they’re very strange lower-level subtleties that the people who thought about the alternatives may have had no idea about the effect it’s going to have at this higher level.
In fact another interesting thing is if you look at these wafers here, and we think about this thing we’re calling the wafer scale framework or wafer scale integration, if we have different subsystems – like this is nine copies of one kind of chip and here’s nine of another kind [showing on wafer] actually there’s four on there – if these were all wired together so they could talk to each other, what would happen is – see the problem with designing things at that next level is suppose a team has designed this [one chip] and a team has designed that [another on the same wafer]. To some extent the way those electrically and logically interact, at all levels on up to the level of the intended user level interaction of the things, is constrained, supported, hindered or helped by the framework of interaction that’s imposed. You come up with a scheme for wiring them together. And for the rules of the game of being allowed to put your design into that, okay?
For example right now, if you look at the design of one of these systems, if somebody designs a thing that interfaces, in a sense the wiring and the conductivity and the logic and all of that are all mapped deeply down into the processor here. Maybe not as much as it used to be, but still a lot. And what we think we’re going to need to do, just to make it possible for more people to independently design, is have much higher-level protocols of interaction defined and carried by the groups that work on voice input systems to processors. I mean, there will be people working on that class of thing, and they’ll somehow have some very high-level notion of how that interacts. They can press the data down to some amount so that the interaction is at a user level protocol. So that they don’t all design a different thing through that’s intimately hooked into this main processor. And in a sense, that’s a metaphor for, there are some interactions to support some overall activity where we need to understand each other’s work, but in fact probably don’t need to know a whole lot about it. But it’s sort of what is important to know. And right now the trouble is all these things are so separate that we don't even know what the lessons are.
S: And everybody has the sense that there are two choices. Either you can really know the thing, and that means being an absolute member of it and going through ten years of school and all of that, or you can have a “popular” version of it, which is by definition nothing that you should in any way rely on or actually try and use. And so it’s finding something that is less than the thing itself, and yet is closely enough taken from the thing that it really bears some direct relationship and can be relied on.
C: It’s sort of like designing an overall system where the people that design the central processor, or the people that design the voice input or the graphics output thing, they don’t need to know exactly what goes on. Within each other’s thing. But as an overall system a game has to be defined of what it’s going to do, where you can work to define the interfaces and the kind of messages that should be communicable across those interfaces so that you can rely on those messages. And somehow, you’ve learned or adapted to support the overall activity.
I mean I think these subtribes can be like individuals plugged together so that somehow, we know what – there has to be a sense of potentially what might be interesting to the other, or useful. In some interactive way finding a framework for that might be very interesting. The thing that really came through in the discussion with you and in reading this [proposal], and in fact this thing that I had reflected back on about designing the design aids, was okay – there are thousands and thousands of people out there, writing programs and trying to support procedures, that those thoughts could become known to, and they’re intelligent enough to internalize it and reflect oh my god, you know, and they come up a couple of levels and they realize that they may just waste their lives doing some stuff that if they thought a little higher up about they might have decided just to abandon the whole thing [laughing] or whatever, I don’t know. But they have no way of getting the message
It’s almost like you need to be able to broadcast essentially a question out on the network and get an answer. I mean that’s really what you’re after. This is a close enough club in that any of us who are in a situation like what I’m in at MIT, the reason I can always play it cool – there were probably two or three times during the class there where somebody asked a really embarrassing question, I just didn’t know the answer to. And I would always finesse that the following way. I’d say that's a really good question [laughing] and I don’t have time to get into it right now but let me get with you tomorrow and we’ll go over it, and I think it might be generally interesting in class, so I’ll try and prepare a little bit on it. Well I can always go home, see, and broadcast over the terminal, and I can get all the possible answers. And that works really well.
Oh, the other thing – I don't know if you've read this. It's a book that's been around a long time but it’s sort of interesting. This book by Simon, The Sciences of the Artificial.
S: Yea, I’m reading it at the moment.
C: Alright. In fact I had read ‘The Architecture of Complexity,’ he had published that separately. But what I found very interesting here was ‘The Science of Design’.
S: Yes, there was some stuff that really seemed appropriate to what we are engaged in.
C: Exactly. And I read this, in fact I just read that for the first time a couple of weeks ago. And it’s really interesting. And yet the thing that’s interesting here is that in a funny way the way he’s written it, or actually thinks about it, it has a sort of individual or egocentric view of the process. It doesn’t take the next view of how the group discover the thing. Instead of designing a design process, it’s sort of how to create essentially a thing which generates design processes. How do you design the design process generator? That being a research function or something. How do you create that? Or maybe even capture it.
S: And how do you do it in a genuinely multi-disciplinary way.
C: Right. I saw this very interesting book review, this thing here A View of Scientific Thought. This is so cool. This is a book of which only 200 copies were sold in 1935, but which Kuhn quoted in his Structure of Scientific Revolutions, and somebody’s just translated it and everybody’s all excited about discovering all these things in it. And what this guy did was he wrote a whole book about where “it” came from, where “it” was one “scientific fact”. Where “it” came from. I thought you might want to read this book review. And that’s like, one scientific fact is like one little compound element of a cultural medium. Where “it” came from. The problem being – and this is sort of history of science stuff, which is a different sort of framework – which is after something has evolved into place, people backtrack what the history of it was, and they have set out to discover it, you see what I’m saying?
C: And the whole thing is they don’t realize that it’s been created rather than discovered. And so you have this strange view about science, that it’s all there and we’re discovering “it” – equations, you know, you just look between the cobwebs and equations will pop out, laws of nature[laughing]. And you know what’s happened in computer science that’s even more bizarre? People don’t even see that as a science of the artificial. It’s as though you’re discovering the computers that are there, rather than that’s all completely artificial. In fact that’s where a lot of this conflict between SSL and CSL has come in. Is that people there are much more into optimization analysis of structures that are sort of a static thing that already is there. And a lot of what’s happened here is the actual conscious synthesis of new things, where you start out with a thing and it gradually gets more instantiated. Then that will get out somewhere and finally CSL people will “discover” it out there and they’ll start utilizing it [laughing]. And then they’ll think they invented it, once they’ve got an equation that describes it.
I mean I know that’s outrageous, but there’s a thing about this whole process that’s really interesting. That’s why this was cool [MPC79] because as I realize in talking about it, it’s just talking through the process, you know, and in fact there’s a clear recognition of the fact that the actual final – you know, here is the design methodology. It’s essentially a culturally evolved thing, with many participants, that evolved in a short time because of the networks. With the conscious notion that it’s not a process of discovery, it’s a process of actual free-running evolution and some selection principles.
S: Right. And maybe it’s that condensed time span that makes it a little bit more available in those terms. You don’t keep losing it behind you as you’re going along in the way that you usually do.
C: That's right. The fact that it was in an artificial infrastructure, means that you could keep it separate from all the journal publications and all that. So that there was a totally independent, artificial infrastructure of communication here. And artifacts. None of them the normal ones. And I started just reflecting. It’s incredible, if you look at all these things that engineers get [periodicals], these things on electronics and all, and you look in here and you see all this stuff and all this representation, and people see the new products – everybody has a sense of new products being created, but somehow there’s this pervasive thing that comes through that it doesn’t have the quality of creation like it did in the early days. There are no names associated with anything – people, I mean. Sure, they’re on the articles, but it’s not like there’s any remembrance. It also casts it in the – everybody’s observers [reading] “U.S. losing ground in the lab,” everybody’s an observer.
[End side 2, Tape 2 side 1]
The thing about this that’s interesting is if you’re an observer – let’s treat all of this as we’re just observing nature basically. I mean this isn’t any different than seeing bugs and stuff like that. So really what I found interesting here, with electronics in particular, was that – Carver Mead knows a lot of people at Intel, has done consulting there, and we’d published all these things printed on the Dover, you know, those early versions of the textbook that we were scattering around like crazy. In this underground thing that was going on. What I would find very interesting is after a rather long time delay, like about a year ago, before the textbook was actually published by Addison-Wesley, so there’d been like a thousand – god, a thousand pre-publications printed. But what happened was about every second or third Electronics that I would open, I would see art that I knew I had created. In other words what I could see embedded in there was the exact – I could recognize enough of the thing – somebody has seen a picture in the book out there, and it’s gotten integrated into mainline industry in some way, you see? When somebody does their article, they’re using this other way of looking at the thing.
So what I can do now when I look at this is, I realize where all this comes from is the process of just flow of artifacts and imitation and it just mumble mumble and it gets – and then what happens is everybody thinks somebody else is doing “it” [laughing]. And what you read is, this is so terrific, there was this Scientific American about microelectronics back in 77. And there’s all these articles in here. And if you read them there will be a description, by a professor at a university, about what this is about. Invariably, independent of where the person is placed, even the people in the companies, they all talk about it from the observer’s viewpoint. They are describing what they believe “they” do. And not a one of them is self-identified in a sense as doing “it”. And everybody we ever came in contact with in industry – this is how we discovered that no one was doing “it” – no “one”, you see – is no one talked about it like they did it! That’s the most astonishing thing.
And in fact it’s so interesting that even till now people that read that book [Introduction to VLSI), a lot of them, they take a long time to catch on that the book is meant that you’re supposed to do it [laughs]. They still think of it as it’s telling you how “they” do it. Really strange messages in all of that. Especially the one about where does it come from, and who does it. No one feels like they’re a participant. So your notion of membership is very interesting.
In fact I guess that reflects back to really the beginning of the conversation. I can see that there’s some fundamental distinction made there – it’s really interesting to come back to this who is doing “it”. In this field, an important industrial field, no one seemed to identify with doing it. In fact almost none of the people involved had a clear picture of every step, so they couldn’t carry the concept through to this [chip]. And to this day I think a lot of people think people somehow on the line, the manufacturing line, wire all these things [MPCs] up. That it’s done under a microscope. But it isn’t, you know, it’s the pattern made by blah, blah, blah and you could draw it on your screen, and it would get printed like the document there. People just don’t see how simple it is, because they’ve never been a participant in the whole thing.
So they’re in fact all identifying with a different sub-specialty. Insufficient to let them identify with actually the thing that they’re involved in. And it’s sad in a way. I know there were people down in E1 Segundo that I talked to that – one lady down there worked on lay-out. And she had spent a couple of years, she would get a certain kind of description in and she would transform the circuit diagram into a layout, and it would pass on to people who digitize that to go out for mask-making and so forth. And she had never, in all that time, seen any of the results of that work. That’s your assembly line. No feedback of any kind, none whatsoever. You know, reviews, and oh, that looks good and maybe you violate a design rule once in a while, but no feedback. In fact the work is checked by design rule checkers.
Now maybe there might be a mechanism for breaking down stratification. Giving the person involved in carrying out an entire task the satisfaction of interacting with the environment in a much richer way. Maybe that might frighten some people, I don’t know. But if they could be given a script that lets them do it, and they get the feeling of really accomplishing things, maybe that – because I remember very much, it was a fascinating encounter, was how excited they would be at being able to do more of it. Because they’d never actually seen any of the stuff that it went into. Isn’t that weird?
Now of course think about administrative people, and the movement of paperwork. What does that all mean, anyway? How meaningless the work is?
S: Right. What does that make you a member of? I mean it makes you a member of this slice of people who do this process. And like we were saying before, there’s not much that you can tell each other about ways of doing it differently. You can commiserate, which is what I think membership in those kinds of jobs largely comes down to. You commiserate. The guys on the assembly line share that common experience. And then you bring social scientists in to make people feel better. And it all gets totally shuttled off to the level of your subjective appraisal of your situation, being satisfied or dissatisfied. Not with what you’re actually doing.
C: Yea, and maybe that's something that could really be altered to where people were much more satisfied with their work, and the organization was more adaptive, responsive, efficient, free running.
S: Yea. You might have to make some fairly radical changes that – I think inevitably those people who are getting the most out of the present way of doing things would resist. But organizationally speaking the likelihood is, I think, that you’d be better off.
C: There will always be a small percentage that have an extreme vested interest in the status quo. And they’ll be the losers, but that’s okay. The really good ones will see it as an opportunity to go even further. So – who knows.
It would be interesting, really to keep up the dialogue. And I would actually like to think if there were a way I could be drawn more into the interactions that go on in the North Pod. Just to understand a little more what’s happening there, the framework there. And think of ways of having a somewhat more integrated activity.
S: Well, this then gets awfully “meta” [laughs], but it seems to me that these transcripts – you know, people in the North Pod are interested in them, they’re starting to disseminate, and everything that you and I say is in them.
C: Oh [laughs] Right. If that is away of starting a larger dialogue – if people had an inclination and an interest in seeing that maybe if different people collaborated, and could explain their work to each other, at least to the extent that they could send messages about it, that then you as a group could really take on more interesting problems.
S: Well, and the ability to know what of your work would be interesting to whom.
C: Maybe that would be something to focus on the next time. Thinking if there were any way that we could try to more quickly zero in on what would be natural interfaces. And what do we need to explain to each other about our stuff in order to bridge those. It’s how do we discover the protocols for sending messages to each other. How do we set a framework for that?
And then there is the other thing about when a body of stuff that’s developing is just at that critical point where there’s stuff that you’re excited about trying to start getting it written down, in order to firm it up, and to create it – it’s sort of the reverse process, of letting the writing of the book on it drive the evolution of it. And pull in collaborators who send messages to each other to make it appear. You can almost get a purer form of thing crated, I think, if you get that process going.
So maybe there’s two things here. If the different areas could seed themselves in purer form, because they’re trying very hard to get wide critique of a rapidly evolving body of knowledge, okay, then maybe they’re then in a better position to share the little bits of all that with each other. Right now there are these large camps of people, like Electronics I mean you look at this [journal] and there’s just no way that this could be communicated to a sociologist [laughs].
S: [laughs] Right. And it probably comes out every week or something.
C: And yet purer forms of those, that work to get the interfaces right, might be able to communicate and collaborate. That’s probably worth thinking about.