From the closing panel at the 1995 Brown/MIT Vannevar Bush Symposium, featuring Doug Engelbart, Alan Kay, Ted Nelson, and Tim Berners-Lee.

For a description of the symposium, see the ACM Interactions article 50 Years After "As We May Think".

Videos of the sessions are here.


We started this discussion by looking at Bush's vision of Memex, and we just heard Alan talk about the golden age of the sixties.  We've ended up with Windows, which is kind of a bleak thought. So, I'd like to ask each of you to postulate about where we would be if Windows never existed. Would we be someplace else, and where would we be?  If Microsoft didn't exist. 

Michael Lesk:  We'd be missing nothing. Microsoft doesn't even claim that they are a highly innovative company. They take products that have already existed and sold well from other vendors such as Word Perfect and they do a commercially better job rewriting these programs and selling them as Excel and Word Perfect. We wouldn't be missing any important technology. 

Questioner:  So you think we'd be in exactly the same place. Personal computing would look the same. 

Michael Lesk:  As far as I can see. 

Alan Kay:  At PARC, one of the goals was to do NLS as a distributed system. All of the Altos had the five-finger keyboards as well as the mouse. We basically loved NLS and we'd done a few modifications which we thought even sped it up.  Part of the interaction scheme on NLS was, I believe, because the analog mouse had some drift in it, so one of the things that they did was to say what kind of a thing you were pointing at.  You'd say “move character” or “move word” or “move paragraph” and so forth.  You gave the command first and then bug bug and then command accept. We realized at Xerox PARC that you wanted to have a speedy scheme for interacting, and we thought we could go even one better by selecting the objects.  So you'd select something you'd do something to, give the command and then, in the case of “move character” you'd go “select”, “move”, “select” and it would do it with fewer keystrokes.

Now, the abortion that happened after PARC was the misunderstanding of the user interface that we did for children, which was the overlapping window interface which we made as naive as absolutely we possibly could to the point of not having any workflow ideas in it, and that was taken over uncritically out into the outside world. So we have many systems, like Lotus Notes and many mail systems, that when you say reply, it comes up with a window over the very thing you were reading as though there weren't any connection between these things. So this is an abortion to me, but its basically part of the whole field. Whereas our notion was that you start the kids off on this fairly simple, naive thing and then there would be an actual progression where you would get up to this several-commands-a-second kind of thing that you could do with NLS. 

If you have ever seen anybody use NLS, it is really marvelous because you're kind of flying along through the stuff several commands a second. There's a completely different sense of what it means to interact than you have today. I characterize what we have today as a wonderful bike with training wheels on, and nobody knows they are on so nobody is trying to take them off. I just feel like we're way way behind where we could have been if it weren't for the way commercialization turned out.

Doug Engelbart:  Well, strangely enough, I feel the same. It's part of the “easy to learn and natural to use” thing that became sort of a god to follow. The marketplace is driving it, and it’s successful, and you could market on that basis. But some of the diagrams that I didn't quite get to the other day were about, how do you ever migrate from a tricycle to a bicycle?  A bicycle is very unnatural and hard to learn compared to a tricycle, and yet in society it has superseded all the tricycles for people over five years old. So the whole idea of high-performance knowledge work is yet to come up and be in the domain. It’s still the orientation of automating what you used to do instead of moving to a whole new domain in which you are obviously going to learn quite a few new skills. As an analogy, suppose you wanted to move up to the ski slopes and be mobile on skis. Well, just visiting them for an afternoon is not going to do it. I'd love to have photographs of skateboards and skis and windsurfing and all of that to show what people can really do if they have a new way, supplied by technology, to be mobile in a different environment. None of that could be done if people insisted that it was an easy-to-learn thing.

So, moving your way around those thought vectors in concept space - I'd forgotten about that.

Alan Kay:  You said that, right? 

Doug Engelbart:  I must have, its so good. It's to externalize your thoughts in concept structures that are meaningful outside. Moving around flexibly and manipulating them and viewing them. It’s a new way to operate on a new kind of externalized medium. So, to keep doing it in a model of the old media is just a hangup. We're going to have to break that perspective and shift to the idea of high performance teams who've learned to coordinate, to get that ball down the field together in all kinds of operations. I feel like the real breakthrough for us getting someplace is going to be when we say, all right, let’s put together high-performance knowledge-work teams and let’s pick the roles they're going to play within our organizations. Even though they operate very differently from their peers out in the rest of the organization, they can interact with them and support them very effectively. So there are roles like that that would be very effective, and everyone else can sort of see, because they're interacting with these guys, what they can do. And suppose it does take 200 hours of specialized training - that's less than boot camp. 

One of those boxes on that paradigm map about deployment was really coming down and showing you that special-purpose teams are one kind of thing in the way that they can propagate, and very different from moving a group of people who have an existing set of staff and processes and methods and skills and equipment, and trying to move them all together. It's practically an impossible task to do that in any significantly large step without having casualties. They just aren't all equipped to be mobile in that space. 

So, there's a lot to go with that, and it all stems from looking at today and saying “why do we accept that?” That's the modern thing, it’s almost a religion. In any other company I'd be afraid to bring that out. Maybe I'll have to run from you too.

Ted Nelson:  I hadn't heard you say this before. So you are basically proposing some kind of information SWAT team that can move swiftly through an organization, or is going to be some sort of elite 'eizatsgroupe' in the files. This is a very exciting and interesting concept, but how would that function organizationally? 

Doug Engelbart: I mentioned the other day that you've got to have a strategy to lift organizations - you can't just lift them all at once. The strategy that I finally worked out is that improvement infrastructure - a term I mentioned, I didn't show you much below that - in that improvement infrastructure there are roles in the improvement process for high-performance teams. They would really be helping the improvement process come about, there would be roles for them to be plugged into an organization in a function that would be very supportive. This is all supposing that these are the outposts and that everybody’s going to start moving there as soon as you can, but you just can't lift the whole organization in some very radical way. So yes, they are an elite team, but you're assuming that in some number of years you'll learn how to get the whole organization there.

Tim Berners-Lee: I think that the gruesome practical point is that finding a path of a so-called vector through concept space to where you are may be one thing, but to get the world there may be a very very different path. So you have to find a path, from wherever they are, for pushing the world vectors, where each step is downhill for the person who has to make that step. Your original question was about Microsoft's existence and the dominance of DOS and later Windows. So, basically if at some point Microsoft has to make a changeover, that's going to be one of the more difficult things to put them in a position so that the next step is downhill. One of the things that can make that easier is that if instead of having just one operating system, you have two. You might ask, for example, what would have happened if the Mac hadn't been there. Maybe we'd still be using DOS without Windows. So, there's an argument which says that it’s easier. When you're in a situation where you don't have any dominance you don't have any commonality. Maybe that's what held up UNIX, because the developers didn't have any commonality. Without code that you can pass around its difficult to spread things around. So the fact that you have DOS there at least means that if you're not trying to change the operating system you can roll out all kinds of software very quickly, so that's got a great advantage. If you can do what you want with the Internet and DOS, then you can do it very quickly. It's when you want to change the Internet or change DOS that you've got yourself a big problem. So the way I hope that it will work for the Web protocols and the WWW standards which I hope will be able to evolve from their kludgy current state towards that beautiful golden dream is that in any case where you lay down a standard, you also allow there to a second and you show a path whereby is a third one comes up that is better you can move in that direction on the heels of the first, keeping it honest and you have a hook in it to be able to hook it all together later on. 

Alan Kay:  American television is kind of a counterexample. In the effort to make it really easy for everybody to watch television, I'm not sure I see any path back upwards toward enlightenment. The advertisers are quite happy and so is Microsoft, in the stage of blindness that everybody's in right now. So I don't think it’s a great idea. There's an economic law, called Gresham's Law, that says that the bad that's easily produced drives out the good because it resets the whole notion of normality. So personally, I think that if somebody invented a bicycle now, they couldn't get anybody to buy it because it would take more than five minutes to learn, and that is really pathetic. 

Ted Nelson:  As far as television is concerned, the operative phrase a few minutes ago was “comfortable step down”. 

Doug Engelbart:  I say, why not make your steps up comfortable? It's going to take stress, but then you have to minimize it. You know, the older you get, the harder it is to change. I keep getting that thrown back at me: “you're going to ask all the old guys who are ready to retire that they're supposed to shift and change?” The only answer is, look, there is a lot of potential for organizations to get a lot more competent in their knowledge work, just a lot more. And it’s going to take a lot of change. The organizations that learn how to do it are going to be successful and maybe the only ones that'll survive, including mankind. So look, we've got to get there. So what you need to do is say, look, we have to find a strategy that's as smart as possible for accommodating all the human and social characteristics that change has to cope with, as well as the dynamics and expense of changing, etc. So the major push in our bootstrapping thing is to say, yes, it looks like a very smart thing to do is to build an appropriate improvement infrastructure, and then as rapidly as possible get the improvements in these capabilities so that they can plug into the improvement structure too, so that you are improving your improvement process maybe even faster than the other. That looked like the only strategy I could think of and there's a practical way to go after it, so strange as it sounds that's what we're after.


Bob Franston, Microsoft:  I'm not going to defend Windows, but what I want to try to understand is why Windows is such a problem. If you have to change the world all at once and you can't coexist with what exists, you've a problem. The web is a great example -- independent of Windows, it adds great functional capability. There are lots of examples, Andy's roller blades, people doing AutoCAD interactive systems.  Just because people decide that it’s worth investing major amounts of time to learn new paradigms, new systems, it doesn't mean you give up the rest. Do you really feel like you have to change all the world at once, or what are the problems with getting people to try out pilots in industry or in the world at large? [I couldn't really understand his words so I'm not confident about this transcript] 

Doug Engelbart:  The only thing I can see is that you have to pilot software, there has to be some sort of conscious pursuit of that future that you can't really guarantee is there, but to say look... 

Bob Frankston:  Take Notes, for example. Trying to get people to adopt that it had similar barriers. It’s far from perfect, but it had some utility which some people eventually found. 

Alan Kay:  I don't think its worth commenting at length about it, but I think my main point is that MS-DOS got millions of people used to a normal mode of interaction that would have been barfed at in the mid '60s. But they didn't know that they could interact better with the machine, so it set a really low standard. Just like television sets a low standard. If you don't know what theater can do you're never going to find out by watching television. That's the problem. So, its not that you can't do other kinds of things. But let’s put Apple in there as well, because from the standpoint that we are talking about, the Apple interface isn't any better than MS-DOS is and Windows. They are basically part and parcel of monolithic thinking. 

For instance, if you wanted to put in a Engelbartian scheme in your application, in the Apple or Windows user interface it’s almost impossible to do, because the theory of interaction is buried so deeply in there that you can't say “okay, I want to interact in a different way”.  And yet in any reasonable object-oriented implementation of these things, as in fact there was at Xerox PARC, you could sit down and design your own interactive interface. In fact, NLS had this interesting thing where, in any part of the system, you could make up a new user interface if you wanted to experiment. You could put in an interface grammar yourself and use it to drive commands if you were experimenting. So it was what I would think of as an open system. It didn't force the user to go the way it did; it was actually extensible in some very interesting ways. You've probably never shown that in a talk because its more complex than some of the other stuff.

Doug Engelbart:  It’s interesting. When he was talking about the emergence of Smalltalk there, well what do you think we've encapsulated - well, NLS grew into Augment and now we've got a Smalltalk encapsulation package that gives you a very flexible kind of interfacing so you can go conventional if you want to, but you can shift.  But anyway, I take my hat off to Alan about providing that way in Smalltalk. 

Michael Lesk:  I get to listen to a number of researchers complain that the development groups or the users in general aren't willing to adopt their ideas. Sometimes I ask these people to look at their own desk and ask how willing are they to say “I want to be the first user on this floor of bubble memory instead of magnetic disk”, or “I’ve just abandoned the C compiler because I've just heard about Eiffel and I think its really wonderful and I want to try it”. I don't hear that very often from people, and they ought to accept that they are not different from the rest of the world. They are thinking the same way. 


Samuel Epstein,  I guess I'm young enough to not totally believe that there isn't any validity to the conspiracy theory that Ted was referring to the other day and things are pretty screwed up if you look at what was going on 20 years ago. I have to say that I've had my own experience that was described earlier today about not being able to turn in a laser-printed report and the excuse I was given was that "well, the computer wrote this and for this English class you have to write this" and I have to say that when we're looking at the big picture here I find that in my own personal case and I know there's others like me that with very few exceptions of actively searching out the people that are up on the stage here and others like you and finding this knowledge and trying to do something with it, the current educational system in this country seems to be doing a really good job of helping to create and accrete these established self interests which seem to prevent the dissemination of this information and also seem to encourage this "well if it takes more than five minutes don't waste your time to create this" environment. Do you have a recommendation of how we can restructure, improve, modify the established educational system so that access to the information you're talking about is available to people who want to do things with it. 

One of the things that Alan mentioned, he held up on a slide, the book "The Molecular Biology of the Gene". I find this very funny because if I had to name four books that did it for me, that basically set my career, the first one was "Computer Lib/Dream Machines", the next one was "The Molecular Biology of the Gene", and the other two which I have sort of lumped together were Newman and Sproull and Foley and van Dam. The last one, which I read just a few years ago, was "The Decline and Fall of the American Programmer" by Edward Yourdon. These were the things that allowed me to build up a career and maybe other people too, often at the expense of or despite their academic career. How can we integrate these things together so they don't seem to be so conflicting?

Alan Kay:  Looking back, I think that one of the paradoxes is that we made a complete mistake when we were doing the interface at PARC because we assumed that the kids would need an easy interface because we were going to try and teach them to program and stuff like that, but in fact they are the ones who are willing to put hours into getting really expert at things - shooting baskets, learning to hit baseballs, learning to ride bikes, and now on video games. I have a four-year old nephew who is really incredible and he could use NLS fantastically if it were available. He would be flying through that stuff, because his whole thing is to become part of the system he's interacting with. So if I had had that perspective I would have designed a completely different interface for the kids, one in which how you became expert was much more apparent than what I did. So I'm sorry for what I did. 

Samuel Epstein:  Let me rephrase my question a little bit... 

Lee Sproull:  Wait, wait, before you rephrase it let me point out - you're stuck with it - that Alan and Papert did do a very wise thing which was to focus on kids, because not only do they have the time to learn weird and strange new things, but they love doing it and that counts for a lot. 

Ted Nelson:  In media, as soon as you establish premises you pretty much lock in on track. For example, the time slot determines the system we call television. Once you have the time slot in commercial television, the rest is given, forget it. Similarly, once you have “the curriculum”, plus the implicit necessity of keeping the dance cards of all those teachers filled -- in other words, it’s a system for employment of teachers and so each teacher has to have an excuse for what he is doing all the time -- then basically you have the educational system as we now know it. 

As I heard your question, Sam, and by the way Sam and I go back to when he was twelve, if I heard your question correctly you said “How can we fix the educational system as we now know it”, and of course I think that that is a contradiction in terms. The point is to get rid of what is in the way, and that is the curriculum, and that is the guaranteed employment for teachers, because that essentially is what is in the way of the students. If we could have a system, for example, where you simply say to a number of fourteen-year olds... Right now we say to them "You must sit here for four years enduring the Duke of History and the Mistress of Mathematics and whatever you dislike about them because they will fix your point of view.”

You see, education is a process of ruining subjects for you and the last subject to be ruined determines your profession. Each subject is personified with the face of this local ogre, or perhaps in some cases with a wondrous person, who then represents that subject and you never find out that every subject has something in it of interest to you. Every subject can have something in it that you take into your heart. I thought I hated history, I thought I hated mathematics, because of the people I encountered and so some other way around. I would like to see a school system where you simply say to the kids at fourteen not “you have to sit here for four years and enduring and enduring and enduring and being endangered and insulted” but rather “You can get out of here as soon as you present for examination any of these any 80 of these 1000 mini-courses on this sheet. You can take more if you want and learn more before we send you out there. Have your choice.” Then I think we would see real motivation, because right now there is no way students can exercise initiative, except a) if they are male, disobeying, or b) if they are female, getting pregnant, or c) totally obeying and outdoing whatever the teacher wants, which is done by very few. Whereas if we give people a way to learn by initiative then we'd fix the problem but nobody wants to do that. 

Samuel Epstein:  That was what my question was really directed at - not at the technology but at the current system and how can we make it more conducive to bringing people online and getting the kids involved. 

Alan Kay:  Its like the cops need criminals. Its a very complex thing. Seymour and I were in Washington yesterday morning talking to two congressional committees about just this problem, and of course both of us were conscious of the irony that we were actually talking to the people who were in business to stay in business. So the interest in changing things in any big way is something that is very far from what they're trying to do. For instance, the average expenditure for children in the US is about $6500/year. If you multiply that out on a classroom you get to around 185 to 200 thousand dollars per classroom. Well, those classrooms are not receiving anywhere near $185,000 a year. From half to three quarters is soaked up by a bureaucracy, so it never gets down to the school or the kids or anything else. And I think that is an enormous problem. And basically we said exactly what you were saying which is that the problem is that there is a curriculum, and we invoked the name of Montessori whose idea was that school should always be an extended kindergarten and it’s the job of people who design the kindergarten to make what happens when you use it for your own reasons more interesting than the regular world is. That's what Montessori did, and I think that's an excellent way of thinking of designing a learning system. 

Ted Nelson:  Do you know what curriculum means in Latin - little racetrack. 

Tim Berners-Lee:  Ted, you suggest basically breaking the curriculum up into little pieces. The Montessori system allows you to do whatever you like so long as whatever Montessori school uses the same colors for the same lengths for the same blocks. So you're put within a very constrained world there.  I think one of the things I took out of your question, which is clearly a good question, is that there are four books that for you could have been your education. Given a computer, the network, and those four books, you would have been happy, you would have achieved a lot of things, and would have wasted a lot less time. 

Samuel Epstein:  And I was.

Tim Berners-Lee:  And you managed to find them. The question is how can other people find them. 

One of the things you could say is that should be a curriculum. The problem is, the reason that I feel that the curriculum won't go away... There's a neat television program in which the BBC took four poor scientists who had been denied acceptance by the academic community, their papers hadn't been published, these beautiful ideas, wonderful ideas were just outside, they were just too different and interesting, and what they did, in fact, was to show by the end of the program that they were also total junk. 

So, the problem is that if you don't have the curriculum, then it’s like throwing somebody out straightaway.  You say, “Alright guys, you're here to be free, to learn, and here's the net,” and basically they might as well do that at home. So what you're really asking is, education in the sense of leading somebody, is there a sense for having something which is established as being reasonable for people to spend their formative years studying. The problem is that it takes too long to get really good ideas into that. 

Now, looking back at things that Ted said and saying “that's really cool, that should be taught really early on, that should happen in the first grade”. It's taken a long time for people to do that because there's a filtration process. If you leave people to look at all the information, you also want to lead them and say “hey, this is really interesting”. Sometimes it seems that the situations we've got, the mechanisms we have for filtering this and reviewing it just takes generations. 

My only suggestion if we're going to change that filtering system in some cases, for example with computing we can move to a biological model. So you can leave all your programming people around and you let people do projects and hey you give them the CPU power and if they make something which demonstrably works and bubbles and enthuses people and now you've got the net and you can actually use biological methods in some cases. You can't do that for learning good history but in some cases you can use biological methods and get a faster turnaround because you can use computing and a whole lot of people out there who will participate during the hours of darkness to try these things out. 

Doug Engelbart:  You know, it just occurred to me that giving us equal time is not fair because in the same amount of time he can say so much more.

Alan Kay:  That was the problem with Butler.

Doug Engelbart:  Yeah, right, Butler. I just want to say that all these things are very relevant to look at about how you educate children in that they can learn more and differently, but the thing is that the institution of education is not likely to be changed by the children, it will be changed by the people that are there in the current state in order to change it. And we can't wait for the 20 years or so until those children are out integrated into society and have roles where they could start making changes. The change driving things about both world events and the technology are moving too fast. We’re just faced with the fact that we have to learn a better way to shift organizations with the people that are in them now. So that's really why you need a strategy for it. And its exciting to think of what the children can become and to flourish etc, but the daunting problem that I think we really have to face is how do you deal with the change of the adult world. 

Samuel Epstein:  Thank you, and also thanks for the last twenty years, guys.