David Gelernter, John Markoff, Clay Shirky [4.23.09]
Introduction by:
David Gelernter, John Markoff, Clay Shirky


The central idea we were working on was this idea of de-localized information — information for which I didn't care what computer it was stored on. It didn't depend on any particular computer. I didn't know the identities of other computers in the ensemble that I was working on. I just knew myself and the cybersphere, or sometimes we called it the tuplesphere, or just a bunch of information floating around. We used the analogy — we talked about helium balloons. We used a million ways to try and explain this idea. 


In June, 2000 Edge published David Gelernter's audacious "The Second Coming: A Manifesto", in which he wrote: "Everything is up for grabs. Everything will change. There is a magnificent sweep of intellectual landscape right in front of us". Ppublication of the manifesto led to one of the most vibrant and interesting Edge discussions, with contributions from many of the leading Edge thinkers in the area of computation. from Stewart Brand, to Freeman dyson, to W. Daniel Hillis. To reprise the introduction:

David Gelernter .....

"...prophesied the rise of the World Wide Web. He understood the idea half a decade before it happened." (John Markoff)

"...is a treasure in the world of computer science...the most articulate and thoughtful of the great living practitioners" (Jaron Lanier)

"...is one of the pioneers in getting many computers to work together and cooperate on solving a single problem, which is the future of computing." (Danny Hillis)

"...is one of the most brilliant and visionary computer scientists of our time." (Bill Joy)

Yale computer scientist David Gelernter entered the public mind one morning in January '92 when The New York Sunday Timesran his picture on the front page of the business section; it filled nearly the whole page. The text of the accompanying story occupied almost another whole page inside.

In 1991 Gelernter had published a book for technologists (an extended research paper) called Mirror Worlds, claiming in effect that one day, there would be something like the Web. As well as forecasting the Web, the book, according to the people who built these systems, also helped lay the basis for the internet programming language "Java" and Sun Microsystems' "Jini."

Gelernter's earlier work on his parallel programming language "Linda" (which allows you to distribute a computer program across a multitude of processors and thus break down problems into a multitude of parts in order to solve them more quickly) and "tuple spaces" underlies such modern-day systems as Sun's JavaSpaces, IBM's T-Spaces, a Lucent company's new "InfernoSpaces" and many other descendants worldwide.

By mid-'92 this set of ideas had taken hold and was exerting a strong influence . By 1993 the Internet was growing fast, and the Web was about to be launched. Gelernter's research group at Yale was an acknowledged world leader in network software and more important, it was known for "The Vision Thing", for the big picture.

In June '93 everything stopped for Gelernter when he was critically injured by a terrorist mailbomb. He was out of action for the rest of '93 and most of '94 as the Web took off, the Internet become an international phenomenon and his aggressive forecasts started to come true. Gelernter endured numerous surgeries through 95, and then a long recuperation period.

Now Gelernter is back. In this audacious manifesto, "The Second Coming", he writes: "Everything is up for grabs. Everything will change. There is a magnificent sweep of intellectual landscape right in front of us."

In his manifesto, Gelernter further developed ideas he had been working on since the 1980s. One such idea was that of the cyberbody as a "cloud":

17. A cyberbody can be replicated or distributed over many computers; can inhabit many computers at the same time. If the Cybersphere's computers are tiles in a paved courtyard, a cyberbody is a cloud's drifting shadow covering many tiles simultaneously.

He also inroduced his idea of the "life stream":

38. A "lifestream" organizes information not as a file cabinet does but roughly as a mind does.

39. A lifestream is a sequence of all kinds of documents — all the electronic documents, digital photos, applications, Web bookmarks, rolodex cards, email messages and every other digital information chunk in your life — arranged from oldest to youngest, constantly growing as new documents arrive, easy to browse and search, with a past, present and future, appearing on your screen as a receding parade of index cards. Documents have no names and there are no directories; you retrieve elements by content: "Fifth Avenue" yields a sub-stream of every document that mentions Fifth Avenue.

40. A stream flows because time flows, and the stream is a concrete representation of time. The "now" line divides past from future. If you have a meeting at 10AM tomorow, you put a reminder document in the future of your stream, at 10AM tomorrow. It flows steadily towards now. When now equals 10AM tomorrow, the reminder leaps over the now line and flows into the past. When you look at the future of your stream you see your plans and appointments, flowing steadily out of the future into the present, then the past.

Today, Bill Gates's name is synonymous with Microsoft Basic. A mention of Bill Joy in the press is usally accompanied by acknowledgement of his early development work on UNIX. Ted Nelson is always associated with hypertext. Jaron Lanier is often identified and credited with his pioneering work on virtual reality. But rarely are "cloud computing" and "lifestreams" (or "lifestreaming") presented in connection with, and with proper credit to, the visionary behind them.

Edge asked John Markoff, who covers technology for The New York Times, and first brought Gelernter's ideas to a wide reading public with his 1991New York Times profile, and social software seer Clay Shirky. a professor at NYU's graduate Interactive Telecommunications Program (ITP), to talk to Gelernter about his ideas. The roundtable took place in New York City on April 25, 2009.


DAVID GELERNTER is a professor of computer science at Yale and chief scientist at Mirror Worlds Technologies (New Haven). His research centers on information management, parallel programming, and artificial intelligence. The "tuple spaces" introduced in Nicholas Carriero and Gelernter's Linda system (1983) are the basis of many computer communication systems worldwide. He is the author of Mirror Worlds, and Drawiing a Life: Surviving the Unabomber.

David Gelernter's Edge Bio page

JOHN MARKOFF covers the computer industry and technology for The New York Times. He is the coauthor of Takedown: The Pursuit and Capture of America's Most Wanted Computer Outlaw (with Tsutomu Shimomura), and author of What The Dormouse Said: How the Sixties Counterculture Shaped the Personal Computer Industry.

John Markoff's Edge Bio page

CLAY SHIRKY is an adjunct professor in NYU's graduate Interactive Telecommunications Program (ITP), where he teaches courses on the interrelated effects of social and technological network topology—how our networks shape culture and vice-versa. He is the author of Here Comes Everybody.

Clay Shirky's Edge Bio page


EDGE: We're here to talk about a brief history of computation and the role that David Gelernter has played in framing a lot of the initiatives that are playing our today, absent any connection to his name. David, why don't we begin with some of the comments in your Manifesto, "The Second Coming", and also in your 2001 book Mirror Worlds. John, why don't you moderate.

MARKOFF: Why don't I start by asking David about the roots of some of the ideas. Because I think I came to your ideas through massively parallel computing before you wrote Mirror Worlds, as I remember it. Those two things are related, but parallelism at that point was a very radical notion; there were a couple of outliers and you had this computing idea that was from another world. Seymour Cray was in fashion in the big computing world. The question I wanted to start with is, what was the relationship between your work and parallelism and the ideas that came forth in Mirror Worlds? Were they related and did one lead to the other?

GELERNTER: Yes, they were related in the sense that the crux of our work on parallelism and the reason why our approach was considered radical and strange in the 1980's was the so-called Tuple Space Model — the idea being that if you had a lot of computing agents who needed to communicate, instead of sending messages to each other, essentially like e-mail, if I had information for someone, I'd just write it on a piece of datum and release it and it would just float up into the cybersphere. If I needed information, I'd look around, grab whatever I want, I would read it or, if it were a task to be done, I'd grab it so nobody else could grab it. 

The central idea we were working on was this idea of de-localized information — information for which I didn't care what computer it was stored on. It didn't depend on any particular computer. I didn't know the identities of other computers in the ensemble that I was working on. I just knew myself and the cybersphere, or sometimes we called it the tuplesphere, or just a bunch of information floating around. We used the analogy — we talked about helium balloons. We used a million ways to try and explain this idea.

EDGE: Who is we and where?

GELERNTER: I would say my research group at Yale. I had some really good graduate students. Nick Carriaro, Jerry Leichter, others — and some researchers at other universities who were interested. But the field during the 1980's said — and I was just emerging from graduate studenthood myself — "This is a very pretty idea. This is a beautiful elegant idea. It's stupid because it's impossible. It will never work. It is grossly inefficient. There is no way that you can take information, just float it out there, and expect people to search this whole vast collection, or somehow or other find what they want. And, you know, how are you going to find out what computer to put it on? How am I going to know what computer to look for it on?" 

We said from the very beginning that those questions are in a larger sense irrelevant because software develops the algorithms it needs to provide the service users want. In the final analysis, the question is not, what can software engineers build? It's the question, What do users need? If we identify our user need, the software technology will come along — in combination with hardware, obviously, and interconnect technology. 

So during the 80's, actually, we built a version of Linda that was very efficient, relying to some extent on fancy optimizations that were higher tech than we would need today, and compiler and language and stuff like that. But it had got to the point in efficient pre-processing that in the early 90's, a very large molecular dynamics code — I'm trying to think of the name, I don't even remember it, something 3-D — which was a huge consumer super-computing cycles, used to compute confirmations of large molecules and certain end-body problems and stuff like that, was ported to Linda by the developer. Gaussian was the name of the software company [actually "Lorentizian" was the name of the company, and the code was "Gaussian-3D"] that developed this huge application and I think the application for many years was a leading consumer in super computing cycles — is still a leading consumer. 

So it was ported to Linda following which super computer makers of parallel machines all needed window ports. And so all of a sudden, what I had talked about in the 1980's as something that was directly accessible to the user, a way for people to communicate, a way for information to be disbursed, as well as a way to support communication turned into a scientific computing tool.

MARKOFF: That's the technical computing direction, but then there was this other — the Mirror Worlds notion, did it grow out of that? Because it basically — you know, computing reflects reality.

GELERNTER: Exactly. As the Gaussian code, as the scientific computing, spun off in one direction, my own interests were really in the other direction. More at the user level. The Mirror Worlds idea began with the idea of delocalized information floating out there so that I could look into my computer and without knowing where to look, what file on which computer, I can sort of tune in the information I wanted the way I tuned in CNN on a TV. I don't have to know where CNN is and I don't have to know on what operating system my TV is running, or the software on my cable box — I just tune it in and I assume it's there.

That was the basis, together with the little towns and villages in New England where my wife comes from. She comes from Cape Cod, so then we were driving around in beautiful little villagelets and they really look the way they're supposed to look. A large white church in the center and they often have a pond next to the church and a mill maybe and the pond is absolutely still. Not all the time. But, you know. So you look at the pond and you see a perfect reflection of the church and stuff like that. This is not a revelation. This is the way mirrors work. 

So this is going to the cybersphere and the real world will be mirrored on the surface of software essentially. Instead of having to penetrate the real world, go places, deal with institutions in their real-world manifestations, which involves a lot of trouble and a lot of time and a lot of energy, in some cases necessary and desirable but not always, I'll be able to tune in any part of reality I want. The university I'm a student at, the company I work for, the local school my kid goes to, a hospital where I help, for instance, and so forth.

MARKOFF: What's so striking about that to me is that the world is just now catching up with that. Ideologically where you were, and what so hit me over the head at that point where Mirror Worlds came up, is that that was in a sense flying in the face of the computing ideology of the times, which were predominantly personal. You, and to a certain extent Mark Weiser at Xerox, had this notion — his notion of ubiquitous computing had many echoes of your ideas. This so much flew in the face of the computing industry and the nature of our view of computing at that point. And we still haven't caught up, basically. It's clear now with cloud computing that that's the direction the world's moving in. But only now.

GELERNTER: I think that's absolutely true. You, John, were the first guy to tell me, "When I look at the Web, it is sort of like a mirror world, although it's not quite there in that it doesn't have the real time flow, it's more static." I started thinking about it then — I think you were the first person actually to put that proposition to me. 

After that there was really a flood of a lot of people saying, "Well, this is an important book, it predicted the Web, blahblahblah." If anything, the fault of the book — well, there are plenty of faults in the book — is that it was too conservative, in one sense. It was published in '91 and said, "We're going to start seeing this ten years from now; take it from me, you're going to start seeing something like this in reality." But in '94 the Web took off. 

We still aren't at a Mirror World type situation where we're at a real time mirror of reality. But in other respects, the technology is bubbling up all over. 

SHIRKY: In some highly instrumented domains we are starting to see it, where people are instrumenting things with sensors. The other thing I think is so striking about it — John mentioned this, right — is that you were flying in the face of the dominant view of the computing industry. What you said, which is simple economics but radical to the industry is, computers will become abundant because people care about them. And when they become abundant enough, we'll stop caring about them. 

Because then we can take them for granted. In '91, no one was ready to take computers for granted yet. They were the fetish object themselves. What I got from it — to me it was a revelation, I was just moving in this direction in those times and I felt like I'd been given a new brain reading it — was the sense of, "oh right, this is going to fade into the background and the computer stops being the object you care about." That movement from object to fabric was absolutely prescient.

GELERNTER: Object to fabric is a beautiful way to put it. I think a lot of the computing industry is still stuck at the object level. I mean it's the engineering in the box that really, really matters. I started telling my students at some point in the 90's, "The picture is the system." They'd say, "Tell me about GUI design but we're going to go the graphic art department, or something. We don't have to talk about the interface here. We have more important things to do. We're talking about the real technology." As far I was concerned, I didn't care about the machine. I mean I cared about it, but the machine wasn't the point. The interconnect wasn't the point. It was a picture. I was dealing with a picture — the picture matters to me. I didn't really care how the picture was produced, but the picture from my point of view was the system. 

SHIRKY: That was not the way the computing industry wanted to think, certainly not in the 1990's. Although there were people at Sun whose work on Java was — there was a lot of prescience in the Mirror Worlds thinking — and so on JINI and JXTA.

And in more recent years, the book has been — Technology Review ran a piece two or three years ago saying, "Well, Google Earth is just taken from a figure." They ran a piece about Google Earth with figures from the Mirror Worlds book and saying "You see this is just what Mirror Worlds said about it." And Second Life and stuff like that, but for other reasons.

MARKOFF: Well, the company that Google bought to create Google Maps, I'm blanking on the name, but they were some KH-11 guys — I asked them directly because the first I saw that, I saw your ideas in their technology, and they were inspired by your book. I think they were very aware of Mirror World. It was resonating around at that point. 

GELERNTER: I didn't know that. As far as I knew, when that Tech Review article came out, a lot of people started talking about that, but I hadn't heard that. Although it was a natural thing. 

MARKOFF: To go back to Linda just one more time, you're about to potentially bail out the computer industry again because here they come with their consumer microprocessors. We're at four and eight cells now, but they're going to a hundred and they don't have any way to efficiently use those. You're going to be there again in another three years.

GELERNTER: Yes. Yes. I gave five billion university and lab talks and government lab talks and department — saying essentially the same thing again and again. And my very close colleague, Nick Carriaro, did also and others in the Linda group did, saying, "Look, it's easy to wire computers together. Computers are going to be dirt cheap. You can wire together as many as you want. You're going to have a lot of bandwidth. The question is, how to program them? What you need to do is make it easy and convenient to program. That is the roote to being able to mine the huge riches of this technology. 

They would say, "Well, no, it's really a hardware thing; we leave it to the networking guys to figure out the protocols, and we can always use messaging passing. Where essentially the design or the program is hardwired into the code; there's no dynamic assignment of tasks; if you want to spread from five computers to six computers, you have to rewrite the program — the details aren't going to matter, but it's an assembler language view of parallelism. 

People using high level rapid application development environments — maybe they're not writing in C, they're writing in Java; or maybe they're not writing in Java, they're using Perl or Python or something like that — are at the lowest conceivable level when it comes to the coordination or communication level. That was the story in the 80's and that was largely the story in the 90's. I hear these bizarre experiences — a few months ago, somebody from Austria, said, "We have an EU development project and let me tell you, Linda is really the coming thing. An exciting new development. People are just getting their heads around it." Could it possibly be that this not an old jalopy idea but actually something for the future? It's hard for me to completely figure out because it was designed to be simple. It really was not designed to be esoteric or difficult.

SHIRKY: There's also so much in the micro formats working in RDF around this idea of, once you make the data available — this is what I think is so powerful about putting the data up in the cybersphere — if it's a task it gets grabbed and acted on, but if it's not a task — if it's just a piece of data — then it can be consumed multiple times by multiple processes working completely independently of one another. So much of the microformats work is just catching up to that.

GELERNTER: That seemed to me a crucial symmetry between communication and space and communication and time, which were handled in radically different ways by the software world, but amounted to the same thing. If I communicate in space, there's some guy I synchronize with — I shoot him a letter, in effect an e-mail, an instant message or something like that. So, here's a message and I hand it to him.

But I may also want to communicate with somebody who lives three weeks from now, who occupies the building after I've gone, who's born after I die. I want to pass him the information, and if I have an environment like a tuple sphere that provides a cloud rather than a series of message-passing links, I don't have to rely on a file system or database mechanism for communication through time and a completely different messaging system for communication through space. Those are basically symmetric operations. You get a lot of power, exactly as you say. There are two cases at opposite ends; first, I'm a parallel program, I have a million tasks to do, and a thousand processors, and I want to get them all done. I want to focus lots of computers at this problem and blow it away fast.

So I put all the tasks up there and when you're free, you grab one. You may be a dynamic network of (we're saying) desktop, whatever. Somebody needs you now but then he goes to lunch. You've got free cycles, so you grab a task and get to work. We call that system, Piranha. It was something we did in the late 80's. 

At the other extreme, it might be just information. I float information out there in the tuple sphere and it's documentation for code or it's ideas, you know, within the Linda group of what we want to do. Those are two extremes. But once you have extremes, you can have a spectrum and the spectrum can be very valuable. I may have a very large computation where I float a task out there on Monday and it's not going to get done until next Monday or something like that — I have a long-running, large computation — so it's not just that the task is transient and gets grabbed by one of the parallel ensemble to be done quickly versus the information which is persistent and floats out there over the longer term. We want the intermediate points too. And by defining the spectrum, a spectrum rather than picking — 

I mean the software industry loves to say, "Look, you've got exactly seven great ways to do this." They all do it. You know, you can do it one, two, three, four. This is ridiculous. Either there's going to be one way or there should be a continuum.

MARKOFF: You know, just a point here. One community that has grasped your ideas and run with them very quickly is the cyber criminal underground. In the sense that if you look at the parallel botnets that are out there, this is sort of ethics of computing, those guys are right on top of things. If you look at particularly at Conficker, which is this worm that's floating around out there right now, it may be the most powerful computer in the world. There are at least 12 to 15 million zombie machines that are on April 1st and they take their instruction — in a an architectural sense they owe their intellectual ideas to you. But those guys haven't waited at all to pursue this archetype.

GELERNTER: The power is amazing and it figures that it would first be mobilized by anti-social lunatics.

EDGE: Could they use your disable omniscience button?

GELERNTER: They don't have to worry about omniscience. They don't suffer from that problem. 
EDGE: Can you use them on them?

GELERNTER: Up to a point

SHIRKY: You wrote that a cyberbody will contain all your life's information. And you have a tuner, your card's in it. But you also need to have a disable omniscience button.

GELERNTER: Yes. We noticed when we were building this Piranha system — we ran it on all the computers in the computer science department; a graduate student was going around handing out chocolate bars in exchange for people giving us their computer. Not everybody would agree. And I understand that. Because there were certain risks involved. 

But we noticed that there was a tremendous amount of extrapolation we could do about a machine on the third floor in this office over there. It's not just that this guy goes to lunch every day at 12:30 and comes back at 1:30; we may notice that he's often on the phone at a quarter after nine, or that there's a high probability that he's going to go to the gym on Tuesdays but if he doesn't go on Tuesdays, he'll go on Friday or something like that. Now that's valuable in principle; it makes the computer much more powerful. But it's also frightening. I wouldn't lend my computer to that ensemble if the result were that the guy who was in charge of the computation knows the entire pattern of my life. So the wall, it's got to be there. Enforcing it is difficult.

SHIRKY: Gordon Bell is in many ways building the system that requires the disable omniscience button because finally he's tipped to this view of, 'the fabric is out there, just write all the possible bits'.

MARKOFF: Well, that crossed the line, I would argue that we are all carrying around powerful computers; they're tossing bits of what we're — 

SHIRKY: Right. The reality mining stuff — you just watch where my phone happens to be, since it's a safe bet it's on my person. I can suddenly say all kinds of things. And we're far more predictable than we'd like to believe, which is one of the things that's always upsetting about getting your own life monitors. Which is, "Oh, yeah, I actually do exactly the same thing, exactly the same way all the time"

GELERNTER: Yes. We're enormously predictable and it would be easy to write a mailer that hands me mail more or less when I want to see it. That deems, 'you're not going to answer this guy until the weekend, but this thing is important and by the way I know what you're going to say so I'll type the message'. Or once you've typed the first couple of sentences, I get the gist because I've seen ten billion of your mail messages.

It's easy to do this. Just like it's easy to build a home entertainment system that says, you want to watch this channel now and you want to go to this site, download this, at such and such a time. But people are spooked by that, understandably, and that's one of the rare fields in which AI could actually do more than, in practical terms, it is doing. The AI, you know, is very applied statistics, really.

MARKOFF: Clay talked about the influence that Mirror Worlds had had on his thinking. I just wanted to say that for me it had a similar kind of impact in that I had grown up — it's this life-imitating-art kind of thing — reading basically FROM Vinges' True Names and then Snow Crash and Neuromancerby Stevenson and Gibson. And you basically gave me a prism through which I could understand that science fiction, but it all came together in one sort of mélange to understand where computing was heading kind of.

GELERNTER: Well, I certainly appreciate your saying so. That was my goal, to sort of have a synthesis of a bunch of ideas that we had been working on like the Linda tuple sphere and other things having to do with the visual interface and stuff like that. Gibson's a remarkable writer. The influence of science fiction on the evolution of computing is obviously a big topic. 

SHIRKY: Something you were saying earlier about the a-historicism of this conversation, the conversation around computing and society, struck me. The guy — and I'm blanking at his name — the guy who invented SanDisk, the USB stick, is now working on a plug-in that will transform any device into your device. It's a combination of your local key plus get stuff from the cloud and every machine you send out if your machine, every phone you pick up is your phone and so forth. And it's in the manifesto. It's in The Second Coming.

You outlined that idea of the card. But because we're such a-historical population, tribe, whatever, it seems like your sort of intellectual children and grandchildren don't even know that you've been talking about this. Do you just wake up every morning and see something in the paper and say, "Oh, yeah. I talked about that 20 years ago and here it is. But they didn't call me because they didn't even know that I was in the vanguard."

GELERNTER: You know, in a sense, there is a window of opportunity and being too early is as bad as being too late.

SHIRKY: "Too early is wrong."

GELERNTER: Exactly. I mean Babbage had brilliant ideas about computing in the nineteenth century. It didn't do him a hell of a lot of good at the time. That's one of the reasons why I have been particularly focused on the explosion of work as it seems to me on Lifestreams and Lifestreaming. Because Mirror Worlds was published in '91, that's before most of these people were born, it seems to me. But the Lifestreams work is not that old. There was this book in 1998, Machine Beauty, there was this manifesto that had a lot of the Lifestream stuff in it. 

EDGE: And could you just define it?

GELERNTER: Well, Lifestreams was already my idea that instead of keeping my information in separate pieces of digital Tupperware with some of it in this app, and some of it in that app, and some of it in the file system, and some of it in my Web brower, and some of it on my laptop, and some in my palm, and some in my cell, and blahblahblah — I didn't want to do that. I wanted every information object I owned arranged in an electronic diary or journal or narrative. Or 'Lifestream' is what I call it.

So, in principle, the first thing would be your electronic birth certificate. Then every piece of electronic stuff you either generated or received, including all of your e-mail and every draft of every paper you wrote and every photograph and snapshot and thing you scanned in and your receipts when you travel and your tax returns, blahblahblah, it would all be on this one stream. So any computer and device that I used, whether it was portable or large screen, or whatever, I would tune in this Lifestream. 

It seemed to me that, obviously there are two ways you can organize things; in space or in time. I can take the spatial route that the desktop, which was a brilliant PARC invention in 1974, and I can say, "Well, certain things will be here and certain things will be there." But usually space is not as good a cue as time. We all know when we were born more or less and what our childhoods were like and what we were doing when we were in college. We move around all the time — I move desks and I move offices and I move buildings and stuff like that — but history, time, the life that I live, my experiences are automatically arranged chronologically. So it seemed to me that data objects should be arranged that way. And gradually in the late 90's, we developed that into an actual system where everything would be indexed, which was regarded as a ridiculous move at the time. There was a thesis in the late 90's and Eric Freeman [who had written this PhD thesis] had a really hard time. He almost got thrown out of there because people said, "Well, are you talking about saving everything? Indexing everything? You've got to throw things away and you can't index everything. And you need a file hierarchy — you've got to put things in files. And you're just talking about — this is just like that nonsense about Linda ten years ago." Nobody believed it then. People hold these views very strongly and I can laugh it off, but these are real issues.

SHIRKY: Yes, they're a real concern. Yes. 

MARKOFF: So you went through this commercial period. If you were going to do that over with hindsight, are there obvious things about how to sort of introduce that idea commercially now, are there other paths you could have taken? Or was just it a question of being too early?

GELERNTER: I think there are other paths we could have taken because the company that commercialized our Yale Lifestreaming software — it wasn't all that long ago. 2000, 2001, 2002, it was sort of in that period, which was five years before people started using the word Lifestreaming generically, but it was within range and we could have done it. 

There's this big question that — everybody from technology, when he meets anybody from business, the business guy says, "Well, you've got good ideas but you don't know anything. I'm going to run this company because you don't have the vaguest idea what you're doing. You're just a professor. So go away." We, the people in the Yale Group, said, "You know what? The way this technology is going to spread is we're going to give it to students. We're going to give it to students at Yale because that's what we know. We know students at Yale and they're going to give it to their friends and their brothers and their girlfriends and their moms and dads, and that's a good way to spread technology." And the brilliant business guy said, "No, we need to go for high value enterprise." "The Enterprise". Right. 

So we sold a copy of the system to the Mississippi Welfare Department or something and it took 30 billion trips to Mississippi to get it, and we sold something to Arkansas and something to Borax, or Ajax, or something. Each sale took a huge amount of time and it was just ridiculous. I don't say I am a business genius because I am not. But I do think that businessmen are full of it and in general they love to lord it over the technology world because they've got the money. But money doesn't always make you right.

MARKOFF: Would that have been an open source strategy for Lifestreams? 

GELERNTER: Yes. I mean that's the way we wanted to go with Linda. In the case of Linda there was a company in New Haven that made scientific software and they took over the Linda effort, with our agreement, and they put money into it. We appreciated that because they brought the code to the level at which it could actually be used in production settings. So that was good. That commercialization experience — it didn't make any money, but it channeled money into a technically exciting end. 

So my students and I were primed for the idea that the commercialization route, even if it doesn't work commercially, is in many ways a better way to get funding for advanced development and research projects than government grants. In the 90's we were very lucky to get a lot of funding from one guy at the Office of Naval Research. A wonderful guy. His name is Dick Lau. I know a lot of other researchers who had one funder who really saw what it was they were doing and made a huge difference, not only in their careers, but in the university's. But when you lose that connection — this is another story — but getting federal funding is really a rat-hole. The commercial route looked very good during the 90's because of our experience with this pre-existing company in New Haven so we naively bumbled forth thinking that "We can't really lose because either we'll get the money and we'll all be billionaires or we'll get the money and any rate we'll build the software and we'll actually have it." 

SHIRKY: This gets back to how early you were. I mean the great stories of Bell Labs and the transistor and Xerox PARC and the GUI and the Dynabook and so forth. And Mirror Worlds, which is the positive externalities of speculative R&D well in an advance of commercial application, are obvious to anyone. We all make our living, in a way, on positive externalities ideas that are 50 years old. And the funding climate now seems, both on the commercial and on the industrial and on the government side, all to be about three to five year paths. A seven-year timeframe for research — that's long-term thinking right there. And I wonder, is the university the last place that's got a long enough time horizon to hold what we used to get from a lot of different sources.

EDGE: A question on your streams. Can you change your life? When your adopted child finds out she has a mother who now has a husband and new family and doesn't want her current family to find out that she has earlier experiences — will people be revising their lives online? 

GELERNTER: That was an actual software, there was an actual coding development group at Yale and then at Mirror Worlds, because it was crucial for us to be able to put things into the future. To put new things in the future was essential because, for one thing, there was no other way I could deal with e-mail. Already, in the late 90's, I was completely overwhelmed all the time. An e-mail would show up. I knew, Yes, I have to answer it. There's no way I can answer it now. So I'm going to copy it into 8 o'clock this evening. And it's going to roll around at 8 o'clock. I'll see it again. And it will keep rolling around, until I answer it. Or there are various options I can do with it. Or a memo is being circulated, I don't have time to read it, so I put it in the future and it drifts around.

So then the question arises, well, if you can put things in the future, can you go back and put things in the past? Or can you erase things in the past? Can you delete things in the past? The point was passionately argued that the idea of erasing things in the past was like dividing by zero. You could hack it up some way, but it didn't make any sense conceptually because what had happened had happened. We shouldn't support an operation to go back and change things in the past because it didn't jibe with a human sense of reality. 

On the other hand, there are practical issues. One of the important things that we did with the Stream had to do with travel. It was a small thing, but I'm travelling and I plan out a series of things. Tomorrow I'm going to be here, here's my hotel reservation and my limo, the phone calls and stuff. The next day I'm going to be there, so when I'm on the spot, it rolls around, I look at the Stream and the Stream tells me what I'm supposed to be doing now essentially. Then it all rolls into the past. So I need receipts to get tax things, reimbursements or whatever, and in cases like that I might want to say, "Well, you know, I want to amplify, I want to add documentation, I want to explain what this e-mail really meant, or what this receipt was for. But you can always copy something forward into the present to do that. 

The case you raise, John, is a fascinating case because we always said, "This is not just a small software thing. This is going to be your life. I would love to be able to inherit a Lifestream from my grandparents and see what their lives were really like and there is no reason I shouldn't leave my Lifestream to my children and my descendants. There's no reason why these shouldn't go for centuries or millennia. Or let you immerse yourself in the lives of your forbears, the lives of an earlier time. In such a radically a-historical world as ours, it's important. But on the other hand, there are cases where I want to suppress the past and it's important for me that I be able to do that. Certainly it is conceivable. It is doable.

MARKOFF: Except that there are multiplicities of Lifestreams and you may delete a portion, but your life intersected with those of others, which will not reflect that, given search. It gets very tricky.

GELERNTER: It gets very tricky very fast. That's absolutely true.

SHIRKY: In the digital world, privacy is the right to insist on incomplete information. Then if you say, "The operation doesn't support incomplete information," what you're saying is, "Privacy is all contingent on some sort of post- hoc wrapper. We'll notice but we won't search it." And those things have a tendency to get broken over time.

GELERNTER: They certainly do. We live in an environment in which the industry — and certainly computer science or the research field — is always thinking that people care more about privacy than they do. I remember, it wasn't all that long ago in the mid late 90's, when people were saying, "Well, what will be the killer app on the Web? You know, mainly it's going to be for pornography. Right? People are never going to send their credit cards over the Web." 

Now obviously it's not just that everybody does it, but that maybe .001 percent of people who do it know what a secure server is. They don't care. The point is that the convenience obviously trumps a very marginal risk in their minds. And in the larger context of privacy, I think people have been aware subliminally, or at some level, for a long time. I think satellites, which have emerged in GPS, and which people have known for a long time are always up there taking pictures of everything, are sort of a privacy leak that affects people's worldview. They don't care about privacy the way you'd really expect them to. We saw these bizarre Webcam incidents in the mid late-90's thatstarted awakening people.

SHIRKY: Now it's geosynchronism. I'm going to have to step out but it's an honor talking to you. Thanks.

GELERNTER: Thank you. It's been great to meet you.

EDGE: All right. Given all of the cybercrime and the security problems and also issues of national security — not just for this country — do we need a new Internet? Or an Internet Two? Should there be a parallel Internet where the current Internet is a bad neighborhood that you might venture into, but if you're going to do anything serious that involves personal things or your records or that requires any kind of security, do we need a different Internet?

GELERNTER: The last time this was really a big issue was when wireless becoming a big thing and people were saying, particularly in Europe, "wireless is a separate Internet and we can make our own rules and not be American in inspiration" and stuff like that. But that's not the way software works. The technology world, if you go back to old codes don't die, they just fade away. It's not that I have so much invested in Cobol that I'm never going to care about Java or Perl. It's just that there's no reason I should throw out stuff that I don't need to throw out. 

So the old environments are improved and hacked up and they sort of blend seamlessly into the new environments. I don't think we have the option of building a new structure. Of course you remember in the early 90's when it was a political issue and Gore was talking about the Information Super Highway and he never even mentioned the Internet. The Internet went online in modern form in 1982. It was up there. The Web wasn't there, but with FTP and Gophor and TelNet, there were a million things on the Internet. It was a very active site. The idea was instead, we're going to go in and build the Autobahn of the information world. But it wasn't plausible because the structure was already there. And it wasn't perfect. It had a lot of things wrong with it. But it evolved.

MARKOFF: But you're talking about the evolution of code and I think that there are some groups who have figured out what seemed to me to be clever ways to evolve the technology transparently. There are certain reasons that we need to go past IP-V4, for example, because we're running out of address space. They've been trying to make IP-V6 work for what, a decade for so and it hasn't really caught on. But then there's a group at Stanford, for example, called Clean Slate, that has figured out sort of a risk style approach to routing where you can slip something underneath today's routers and run parallel protocols, where you can evolve the protocol set, which seems to me to be a good way to approach some of the problems that are ... 

GELERNTER: Right. It reminds me of IBM ... 


GELERNTER: VM, yes, which was really clever back in the ... 

MARKOFF: Yes. Every problem in computer science can be solved by one more level of indirection. Right? Isn't that the gospel?

GELERNTER: Right. By a little more power and one more level of indirection. Yes, absolutely.

MARKOFF: But there are some problems. We didn't get into it with Clay, but this issue of anonymity — people can act in ways in cyberspace that they can't act in the physical world and we're struggling with that tension. And it hasn't been sorted out yet. 

GELERNTER: It's absolutely true. During the 90's, in the year after I published Mirror Worlds, the one part of the book that was discussed or quoted to me was the end of the book, which was an epilogue, an argument between two alter egos, one of whom said these mirror worlds are really great because I can tour the whole world without getting out of my pajamas and I can do all this stuff. And the other guy saying, "Well, you know, this is sort of an inevitable direction of technology, but you're cutting out something essential. If you cut out your direct interface with reality, don't think you're going to see everything that you would see otherwise. You're going to miss something. You won't even know what it is you're missing. That's the other side of the fact that I can be a million different, or I can be ten different people online, none of whom is really myself. 

In the final analysis, these kinds of interactions make my life easier but for what? Not so that I can spend more hours on the Web, but so that ultimately I can spend more hours with human beings. Inevitably, the early work is done by a psychologically specific type of character. They are very good programmers and they are very good engineers, designers, stuff like that, but they are not necessarily typical of the sorts of users we're going to see eventually. 

MARKOFF: And so as we get away from that point where we treat computers as a fetish but where they drop into the background and you have your Avatar but it doesn't require your presence on the Net 24 hours a day, then we could reach that where you could step back from the Web.

GELERNTER: Yeah. You know, to step back from the Web is what people desperately want to do. They don't care about the gadgetry and they don't care about the network. They care about the stuff. When I buy a TV, I care about the movies that I'm watching, not about the details of the scanning and the interleaves. But the computer industry itself, and partly as a reflection of the absolutely astonishing technological achievements of the IC- industry, has kept the focus on itself. And this is still a really controversial issue in computers. What are we looking for and should we have computers in mind, or should we have software in mind, or should we have actual users in mind? 

I think the average programmer still doesn't understand that he is not a typical user. The average programmer still thinks that insofar as people don't find his software easy to use, it's because they are childish, or ignorant, or just obtuse. But easily half the world doesn't like playing with machines. It's not something they enjoy doing. It's not something they take to.

EDGE: How would you save The New York Times?

GELERNTER: Clay's piece that recently you ran was fascinating. But I would say that The New York Times is, in fact, producing a tremendously important service aside from distribution and it's a service that gets more and more important exponentially every day and every hour, which is editing and selection. If there is more and more stuff, I give up, I can't possibly look at all the blogs I'd like to look at in principle, much less go to all the sites. Yet the information that I need is online. 

I remember several years ago, Bill Crystal telling me — of The Weekly Standard —, "You know, I can learn more about Iraq from 20 blogs than I can by flying to Baghdad." But the problem is, it's not necessarily the same 20 blogs every day. There's a large collection of blogs integrating over all the information that's there, but I just don't have the time, nor can I deputize somebody who'd have the time to look at all these sites. And the information is out there. 

We need The New York Times, we need newspapers, to provide it for us, to edit it so that it is not incoherent. But we also need them to deliver stuff in real time. We have to be able to see the flow. Arranging stuff on paper, on the page, was a hack that never made sense for me. News was always The AP Wire, you know, new things, I want to see what's new. 

MARKOFF: It's a distribution mechanism.

GELERNTER: Yes, a distribution mechanism.

EDGE: What you get from the Times — part of the editing is the placement and the presentation. So when John Markoff wrote about you in '91, it was two full pages. That's a statement. When the Times wrote about Emanuel Derman in "The Science Times" a week and a half ago, it was three-quarters of the front page of "The Science Times" and a full page besides. That means it's important, stop what you're doing.

GELERNTER: But look, John, the AP Wire had a way to do that also. I mean it was a primitive way. But there were bells. Right. Yeah. The thing chattered all day long. Important you got a bell, and more important you got two bells and so forth. We can have a much more sophisticated way to do that. 

MARKOFF: And so here's the dilemma of The New York Times. To me, so clearly, I've become a total McLuhanist on this. The average reader print reader of The New York Times spends about 35 minutes a day with us. The average online reader of The New York Times spends about 32 minutes a month. In that sort of dialectic lies our problem. 

I'm not going to go into it because I'm arguing inside the paper for a solution to this problem, but I don't think it's on the Web. But I think that tension, and that gets to the notion that the framing of the world — I mean there are lots of ways to frame the world, but you lose the ability to frame the world on the web.

EDGE: McLuhan said another really interesting personally to me, he said, "Read the paper for the ads; that's the news". The display ads.

GELERNTER: The difference between 32 minutes a day and a week or a month is a service that the Times doesn't provide but could provide over the Web, which is what I get when I sit at a sidewalk cafe and watch the people stroll past. Or the girls stroll past. Or whatever I'm watching stroll past. I'm not actively paying attention, but I like to watch the parade. I take it in. I may be reading a book or talking to somebody. But if I see somebody I know, or somebody, whatever, my attention will focus. And you'll get one of those 32 minutes. But you need to provide me with the right kind of parade so I can sit back and watch it with a tenth of my mind while I'm doing something else that's got to be attractive and interesting to look at.

MARKOFF: There's a way to do that. I actually think there's a way to do that. 

GELERNTER: So do I. It's a Lifestream. But I'm sure you've got a good way too.

MARKOFF: All right. Explain how Lifestream would work in terms of The New York Times. 

GELERNTER: Well, it's a switch so that instead of the managing editor, the city editor, or whatever, being a bigshot, there's something more like a producer of the stream in real time. So the producer of the stream has lots of feeds. A reporter is posting a new story. Another reporter is posting a new story. AP is doing stuff. Photographs are coming in. Videos are coming in. But each person looks at one thing at one time. Okay, so I as the producer want to say, "Okay, put that on the stream now. And now put this on the stream. And now put two of these on the stream."

MARKOFF: When I was in the newsroom of The San Francisco Examiner in 1987, we had something like that in the newsroom. The guy was called the wire editor. The wire editor was somebody who watched all the wires and alerted me as a reporter on a beat. Because he knew what I was interested in, he would give me the stuff I needed. We've lost that capability I keep looking at all of these efforts to build software agents and they all fail. I know it's a hard problem, but to me it's still a goal, it's a design goal.

GELERNTER: Absolutely. There's got to be a collaboration. 

EDGE: Doesn't it become CNN then? If you just gave someone the transcript to read, wouldn't they just be one of these 24-hour news stations?

MARKOFF: It is. But on the other hand, I'd like to have that. I had that in the newsroom of The San Francisco Examiner, a mid-level afternoon daily, 20 years ago and I don't have anything like that. What I've got is this deluge of information streams coming from every direction. 

GELERNTER: But what you're going to get, when you have a Lifestream with a producer, or John's system, however it looks, is, of course it's not just text but you're going to get videos and you're going to get pictures and you're going to get sound and you're going to get multimedia chunks of stuff.

I want to be entertained by everything I do today, certainly by The New York Times and so I want to see this flow of stuff. Whereas ten years ago I thought each individual thing would be like a page, now obviously it will often be a video and the video will be bouncing around and there'll be another video behind and so forth.

We need the ability to mix all this stuff together because sometimes I want to read text too. We also need the ability to march all that stuff past in real time, but in a way that allows me to reach back as far as I want when I want to read up on a topic so it's all there, part of the same system in the same interface.