(Editor's note: transcripts don't do talks justice.
This transcript is useful for searching and reference, but we recommend watching the video rather than reading the transcript alone!
For a reader of typical speed, reading this will take 15% less time than watching the video, but you'll miss out on body language and the speaker's slides!)
OK. Hello, again. Our first speaker today, as you can see, is me. You don't have to clap. It's OK.
OK. The title of this talk is, Doug's Demo, about a live demonstration of a computing system done in 1968. But we're going to go back further than that, to get some historical context, to 1945, an article called As We May Think written by Vannevar Bush. Vannevar Bush worked on the Manhattan Project. And what he saw on that project, in terms of his management role, was a lot of information and people having a hard time just dealing with it. It was a huge endeavor involving massive numbers of people and obviously time compressed.
So he thought that something must be done about our ability to manage information if we're going to continue to do bigger and bigger projects, just as a species. And he writes this article, As We May Think. This is the header image from the version-- a sort of abridged, popularized version in Life magazine. And if we scroll down, we can see the title, As We May Think-- "A top US scientist foresees a possible future world in which man0made machines will start to think."
It's obvious that an editor wrote this title, because it doesn't really describe what happens in the article at all. What actually happens in the article is he's talking about mechanical augmentation of the human mind using this proposed machine called the memex that was never actually built in its proposed form. But spoiler, for this talk, Doug Engelbart builds the memex in the 60s, basically, as a digital computer.
So this is an artist's conception of the time of what a memex might look like. You can it's very mechanical. We're not going to talk about this, because it's just, again, an artist's conception. But I would like to ask you where your legs go when you sit at this desk.
All right. So I have a single quote from this article that I think sums up a big part of what he was getting at and the sort of original thought, or at least a synthesis of a lot of different ideas. "Wholly new forms of encyclopedias will appear, ready made with a mesh of associative trails running through them, ready to be dropped into the memex and they're amplified." Now, he's using fancy language, so let's like compress this a little bit.
First of all, "new encyclopedias will appear." That's, I guess, trivially true. "Ready made with a mesh of associative trails running through them." What he means is that the articles will be connected to each other. So you can follow the links between them in some ways. So we can just say using hypertext. The word didn't exist in the '45. But the concept did, and that's really what he's talking about.
"Ready to be dropped into the memex and there amplified." In modern terms, we would just say read via computers. Because the memex was sort of an early version of a computer-like device, a machine to help you process information. So new encyclopedias will appear using hypertext read via computers.
So on this memex, if you're reading the article about dogs, it'll say the genus of dog is canis, and then you can kind of follow that link to canis, And that's a genus, but maybe you don't know what a genus is, so you follow the link to genus. And it says that that's part of the system of biological taxonomy. So you follow the link there. So basically, in 1945, Vannevar Bush anticipated the Wikipedia rat hole.
Which is very forward thinking. And I'm sort of joking here, because recognizing that we needed to build machines to allow us to follow these links is actually a very important thing to know if you're actually going to do that, as we did with Wikipedia and other systems. And you have to compare this with shells of books. Because, in 1945, the only way you have to do reference books on a shelf. So if you want to go from to dog to canis, you have to stand up, go find the CA volume, put it on the desk, find the right page. It's a very slow process-- 1,000 times slower than clicking a link.
So that's sort of pre-history of this. Now, we can move to 1968. Not quite to the demo yet though, because I want to talk about what computing was like in 1968, so that we have context for what Doug demonstrated. This is a stack of punch cards. And if we zoom in on this, we can see sort of how these punch cards work. So I have three columns highlighted here. You can see a five of these punched out-- it's missing. A four is punched, a six is punched out. So this is the number four, five, six.
And you would punch out all of these numbers. Well, machines would help you. But all these numbers have to be punched out of the card. And then you take a giant stack of these, and you drop into a machine like this one-- specifically this part of the machine. It pulls all the cards in. That's the input. It contains the program, and the data to be run, and so on. And then you wait.
So this is offline computing. No interaction with the computer. You put a stack of input in, it does its computing, a stack of output comes out. And that computing step, if you're relatively lucky, might take your lunch break. If you are less lucky, it might take Tuesday. But it is not interactive. It is batch processing.
Of course, with computers, often, things go wrong. Which means you have to re-punch. And then it's Wednesday. So this is very, very slow, offline processing.
And now we can come to what Doug demonstrated. And you have to remember throughout all of this, that the punch card world is the world in which he's demonstrating this system. So here's Doug starting to introduce it.
And I want to mention something about this. I left in about a second before he starts talking where does a sort of nervous laugh. And he's nervous at the beginning. He's demonstrating this and in front of a live audience. He's been working on this for 18 years. There are other reasons for him to be nervous that we'll talk about later, but now it's actually listen to this.
- The research program that I'm going to describe to you is quickly characterizable by saying, if, in your office, you, as an intellectual worker, were supplied with a computer display backed up by a computer that was alive for you all day and was instantly responsive to every reaction you had, how much value could you derive from that? Well, this basically characterizes what we've been pursuing for many years in what we call the Augmented Human Intellect Research Center at Stanford Research Institute.
OK. So the Augmented Human Intellect Research Center-- that's a lofty name. And he used this phrase "a computer alive for you all day." And this is, again, in the context of punch cards-- a computer on your desk. The name of this system that he demonstrated is NLS, which is a pseudo acronym of online system in comparison with offline. And so let's start to go through the basics of what this system could. Do.
So I'm going to show you a demo of him doing copy-paste. And if I switch to the freeze frame of that-- I'm not sure how well you can see this-- his left hand is on a really weird device that we're not really going to talk about much. There's a keyboard in the middle, and his right hand is on a mouse. And if you thought about this without knowing the history, you might think, OK, it went punch cards, then keyboards to drive a computer, than mice. That's not really what happened.
NLS is such a giant leap, that all of this stuff kind of comes at once, at least in a rudimentary form. So here's Doug copying and pasting text.
- So we've see how we can start with a blank piece of paper and go to developing a file. This file is one statement with a few words in it. Let's make more statements. I'll say copy that statement, and lo and behold have another one. Copy that one-- another one.
I can even copy groups of statements. I can say, after that one, copy the group from there to there. And it does.
OK, so copy-pasting text. Now, at the top of that screen, the date is actually there. So here's a zoomed in version. It's December 9, 1968. And just to give you a sense of 1968, the White Album is 17 days old on this particular day. Star Trek the original series-- still running. It's in the middle of its third season. And Star Trek the original series is actually a really good way to understand the way that people thought about computers at the time. Because if you take pretty much any photograph of the bridge, like this one, and zoom in on the computers behind Spock, that's not a screen. That's not a monitor. It's a bunch of static lights.
And certainly, they could have taken a piece of glass, painted a fake monitor image on it, and backlit it, and they would have had a fake monitor. But that's not what computers were. Computers were lights, and if we scroll down, computers were-- it's kind of fuzzy because it's an old photograph-- but that's not a keyboard, it's a bunch of buttons-- purpose-built buttons. Each one does a specific thing. So it's not a general purpose machine. So it's static lights and static buttons, not a screen and a keyboard that can do arbitrary computing.
OK. So now, let's continue with this demo. Here's Doug showing how he can dynamically sort of filter data that he's entered.
- It would be interesting if I could ask the computer to collapse that, perhaps to show me just the first line of each of those statements. All right. Please, do that. So it did.
All right. So the data didn't change, but his view of it changed. So this is much more than just putting text in and getting text out. And this is starting to get back to what Vannevar Bush was writing about in 1945, we need to be able to organize data and see it in different ways, because we just have too much to work with.
And, of course, this example where he, earlier in the demo, typed in word, word, word-- the examples aren't that exciting. But you have to look past that to see what the computer is actually doing.
OK, so continuing through this, some hierarchical data. This example, at least, is a little bit more realistic maybe. It's a shopping list. Not exactly as difficult a problem as part of the Manhattan Project, but it does show you hierarchical structure in the data.
- All right. Produce-- I've got carrots. And I'm move under there are also, bananas. And in fact, I could move a whole a group under there, saying oranges and apples also. So I can begin categorizing things like that.
And if I looked at the numbers now, I'd find that these items fit under there as a subset. And I realize I can categorize quite extensively. I could introduce a new thing under there that was something I just invented-- a skinless banana. But I have to go there.
OK. So it's not just text, and it's not just dynamically manipulable, but there's this hierarchical structure. You might have noticed a lot of beeping in this video. And that's not just an artifact of the recording. The computer does a low beep when it's thinking. And then when it finishes it does a high beep, which is kind of endearing in a retro-futurist kind of way. And I assume that was audible to the operator and to the audience at this slide demo he's doing at this conference.
So let's now see how this view control of the ability to dynamically manipulate data intersects with the hierarchical structure that he's built up.
- So part of our view control, besides this thing we've shown you of showing numbers or not, is also whether we can show you some of these different levels or not. I can say I don't want to see two levels, or only one level deep in there. It makes it very nice for studying it. Or I can do something like, when I say, let's put produce at the top, let's open up one level below it and only show me what's below it, and another level if I wish, and then no numbers.
So very fluid interactive computing, right? And remember, everyone in the room is in a punch card world. So the competition for this at the time is paper, because punch cards, you can't do this kind of dynamic slicing of data. And someone who was in the audience, a Xerox PARC engineer, said that, Doug Engelbart was dealing lightning with both hands, which gives you a sense of how the audience was viewing this.
He's kind of-- he's not super excited up there. He's kind of mild-mannered. And to us it just looks like an old computing system. But it was such a massive leap. And at the end of his talk he got a standing ovation, which actually happened once at this conference yesterday, but usually doesn't happen very much at technical events.
And all of this is even more fascinating and sort of amazing, when you realize that Doug was largely viewed as a crackpot before this demo. He worked on this for 18 years, and people just didn't think it was either possible or worth pursuing. So it's vindication for Doug after this 18 years of working on this project. And for about the first 11 or so of those years, he was basically working alone on this project.
Now, this happens in 1968. I think, in large part, because he's been working on it for a long time. But finally, the hardware is just barely good enough, that if you have basically unlimited money you can finally do this. So let's look at the hardware a little bit.
They invented, designed, and hand-built the first mice. So here's one of them. If we zoom in a little bit, you can see the holes where they nailed the wooden pieces together. And looks like there was an accidental cut there. It had two wheels on the bottom of that directly contacted the desk and moved as the mouse slid around.
This is the weird thing his left hand was on. It's a cord keyboard. So he can push any one of those keys and get a command out of it, or push any combination up to all five at once to get different commands. This is part of what made NLS pretty difficult to learn, because it's just a lot of memorization.
So we have a custom-made mouse. There's also a custom-designed and built modem. And the reason for that is that Doug is about 48 kilometers away from the computer he's using. So the computer is actually back at Stanford, and he's at a conference, and all this is happening over copper lines-- basically, phone lines.
The computer was in SDS 940. Probably almost none of you-- maybe none of you-- know that name. I didn't know that name until I started preparing this. It's not that important historically, but it cost almost a million in today's dollars. At the time it was about 125,000. So they're not trying to sell this to people for home or even for offices, because you're not going to dedicate a $1 million computer to each person in an office.
This is Stanford Research Institute. It's not-for-profit. They're trying to figure out how computers could be used to empower individual humans. They're not trying to sell a product. It was actually later productized, sort of, unsuccessfully. So-- oh, well.
This machine could do about a quarter of a million ads for seconds. So clocked rate, something in the probably high hundreds of thousands or high hundreds of kilohertz, I would guess-- almost certainly less than a megahertz. So thousands of times slower than the phones that all of you have. So that's the hardware.
And now, I want to come to the part of the demo where you see that he actually was building the memex in particular. And it's where he demonstrates hypertext. And this is the first time that anyone in the world ever demonstrated hypertext on a computer or in any interactive system.
So what he's going to do here is first he jumps to carrots manually. And then he's going to insert a link to carrots, and click it and go back, and click it and go back a few times-- just like you would in a web browser.
- I could say jump to a name, and say go to 2A4. And it'll do it, or return. So I can jump to a location number just by giving it. Or if I wish, I can add it as text in there, and say jump to name. And just point to that, and it'll to 2A.
Carrots, right? Right, carrots. 2A4.
So he didn't make a mistake there, but he did seem to misread something. Which is great, because it made him go back and forth twice. So you really get to see click link, go back, click link, go back.
So now, we can have an encyclopedia where you have the dog entry that links to the canis genus, which links to the general concept of canis, which links to the concept of biological taxonomy. And you can click the links one-by-one, instead of pulling down each of the books in order. They didn't do that. They didn't put the encyclopedic data in there. But you can see that the computing system is basically ready for that. The pieces are in place. We know how to build that kind of system. And it took a long time to actually do it, but the team that was on this project was not big enough to make an encyclopedia.
So you might see this and ask the question, is this really hypertext, or am I looking at this historically and saying that looks kind of like a weblink? So conveniently, I can appeal directly to the ultimate authority on this question, Tim Berners-Lee, the creator of the web. He once said, Doug Engelbart, in the 1960s, made a great system, just like the World Wide Web, except that it just ran on one big computer, as the internet hadn't been invented yet. Which is a pretty good excuse for not inventing the web.
But if it's good enough for Tim Berners-Lee, it's good enough for me.
So if we step back and look at this whole history, we started with information management problems in the Manhattan Projects, and Vannevar Bush imagining even bigger information management problems in larger projects. And he said, let's build this memex thing or something like it. It wasn't really feasible-- it was a mechanical idea. But Doug Engelbart read that paper at some point between publication in 1945 and 1950, and he started working on what became NLS in 1950.
Now, 1950, Doug Engelbart is 25 years old. So he started then. He worked on it for 18 years. And when he did the demo that we've been seeing clips of, he was 43. So basically, he spent his entire adult life building the memex. And then out of that came all kinds of stuff-- the mouse, of course, is the most famous thing. But all of this interactive computing stuff, some of which went on to inspire Tim Berners-Lee to build the web, which we're still using 30 years later.
So just very quickly, because I wanted to give you more historical context and fewer clips, but here's a list of things that happened for the first time-- hypertext links, of course. We already saw that. Raster video, which we actually didn't see, because I didn't use any of my clips that had it. But here's a screen cap of a particular plot that Doug shows up. It's actually the number of staff over time on the project. And you can see the long, flat line where it's just him at the beginning.
So you could be forgiven for assuming the system is text-only. Just like you could be forgiven for assuming that we went from punch cards to a long keyboard phase and then got mice. But all of this kind of happened at once. We went straight from punch cards to Doug Engelbart demonstrating a live system with raster video and plots. And he draws them at one point. He does all kinds of graphical stuff.
Interactive data filtering, of course, happens for the first time. We saw that with view control and the hierarchical structure. The mouse, of course-- you have to mention that always with NLS. Document version control-- so managing past versions of documents happens at a certain point. I didn't even really review that for this talk. I thought it was too complicated to talk about directly on stage.
And remote screen sharing-- so Doug is remote from the computer, but other people at various points take control of the computer. And they're also not with Doug. And then there's video conferencing that happens at one point via microwave link. It's just all kinds of things that don't seem like they should have been possible.
So now, we're coming to the end of this. Why should we care about this particular thing that happened? I mean, history is just one thing after another, why do we care about this particular thing? Well, aside from the fact that it's amazing that they pulled all this off with such primitive technology, it's sort of contextualizes the tech industry. Because this is one of the events that catalyzed the spread of interactive computing. Because remember, Doug was a crackpot, and then he showed people that it actually worked. And then, suddenly, he was getting a standing ovation-- probably from a lot of those same people at that conference.
And this was all about empowering individual people. Augmenting human intellect was literally the terminology they used. And they weren't trying to sell a box to you. They were trying to figure out how it could use these machines to empower people. So it was incredibly idealistic. And also, it actually worked and spawned all this stuff that we do today-- or at least was an important stepping stone to that stuff.
So I think that this is a Silicon Valley more like what we deserve as a species to have. And if you contrast that with what we have right now, or at least the parts of it that seem to be in the popular view of the Valley, which I think is pretty accurate, that's mostly selling ads and selling user's personal information to the highest bidder. So instead of empowering individual people, we're doing a lot of selling people as commodities, selling information about people as commodities to advertisers and other companies that are doing things that aren't so great. So if you want to sort of actionable thing from this talk, if the history isn't sufficient for you, I would suggest that we can choose which one of these kinds of problems we work on. And voting with your feet, in that way, as the people who build these systems, I think is a wonderful way to impact history.
So Doug Engelbart died in 2013. He didn't live to see this year, which is the 50th anniversary of this demonstration, in which you can see a lot of stuff that looks very familiar to us today. I would like to thank SRI, which used to be called Stanford Research Institute, for giving me permission to use the clips of this demo in this talk. And I have a list here of the various Creative Commons images that I've used.
And I would encourage all of you to go and watch the demo. Google Mother of All Demos-- it's very Google-able. Find a version, hopefully that's at least 30 minutes long. So it's not just someone slicing it up, but you actually get to see the whole progression as he builds up more and more complex features of the thing and demonstrates them to an audience who's never even seen any of this stuff before. So that's Doug's Demo, and I thank you very much for listening.