Read the transcript of a public conversation with best-selling author George Dyson on the subject of technology, time, and apocalyptic visions and predictions. Moderated by Mark C. Taylor, co-director of the IRCPL and Chair of the Department of Religion at Columbia University.
Mark Taylor: Good evening, and thank you all for coming. My name is Mark Taylor, I am the co-director at the Institute for Religion, Culture and Public Life, which is sponsoring tonight’s event. It’s part of a series we’re doing this semester called Apocalypse Now: End Times in the Contemporary Imaginary. We come at this from a variety of perspectives, from a cultural critique to our next event which will be on October 30th with Karen Thompson Walker, who is a novelist. Her book The Age of Miracles will be the subject of that discussion. Tonight we are delighted to have George Dyson with us, his work has long fascinated me, partly because of the broad range of fields and issues that he covers. As a lay person in this discussion, it is increasingly difficult for me to be sure where science fiction ends and science begins, and I’m not sure anyone can draw that line clearly in terms of apocalypse. George will talk about that tonight.
We have Cyber terrorism and genetic algorithms running wild, we have high speed trading, which we will talk about, we have various kinds of codes that intersect – t genetic, computer, and otherwise, flash crashes, and the like. One of the novelists who captures a lot of this is Michael Crichton, and he prefaces his book Prey, which as some of you may know, is about a series of nanobots that escape into the wild and start reproducing, by a quotation from, J. Doyne Farmer. Farmer himself is an interesting person, he’s a professor at the Santa Fe Institute, where some of the ideas we are talking about tonight come from, and he started the Prediction Company, which is a company that does economic market modeling using some of the theories of emerging complex data systems, which come out of the Santa Fe Institute.
Farmer writes, “Within fifty to a hundred years a new class of organisms is likely to emerge. These organisms will be artificial in the sense that they will originally be designed by humans. However, they will reproduce, and will evolve into something other than their initial form; they will be alive under any reasonable definition of the word. These organisms will evolve in a fundamentally different manner the pace will be extremely rapid. The impact on humanity and the biosphere could be enormous, larger than the industrial revolution, nuclear weapons, or environmental pollution. We must take steps now to shape the emergence of artificial organisms.”
The second epigram to Prey, is from Eric Drexler, who has a book also related to some of issues tonight, Engines of Creation: The Coming Era of Nanotechnology. Drexler writes, and this is 1992, “There are many people, including myself, who are quite queasy about the consequences of this technology for the future.”
George Dyson is hard to categorize in terms of the work he has done. He is concerned with Historio-Technology in many ways, and his writing covers a broad range. His book Darwin Among the Machines: The Evolution of Global Intelligence is one of the important formative books in some of the things we will be discussing tonight. Most recently he wrote a very important book called Turing’s Cathedral, but his writing is only one part of his life. He dropped out of high school at the age of 16 and headed to Canada, where he still lives. He is a naturalist and outdoor person, and for 20 years, he has been designing Kayaks and exploring the water ways near where he lives.
We have a lot to talk about tonight, and will be covering a broad range of areas. George and I are going to engage in conversation for about 45 minutes to an hour, and then we will open it up to all of you, to ask whatever questions you have. George, again, we are delighted you could be here and are giving your time. Maybe we could begin with where I ended the introduction, and that’s with kayaks. What’s the deal with kayaks?
George Dyson: Kayaks are very interesting optimization problems. Kayaks are the answer to the question of how you can get the maximum amount of boat out of the minimum amount of material. And if you ask that question, you get the kayak. That fascinated me from the age of 12, and it never stopped interesting me, I just kept looking at that question.
M.T.: You also mentioned earlier this afternoon while we were talking that the corner of the world where you decided to settle down is an interesting area in terms of shaping your thoughts and relationship to Russia.
G.D.: Right, in the historical sense, when I became interested in kayaks I found out very quickly that the most advanced kayaks were built by the Aleuts, by the people who live in Aleutian Islands, who have been building kayaks for at least 10,000 years, a long time. Then one day September 1749 the Russians show up. And instead of the normal colonial model where the Russians show up and the new world comes in and replaces the indigenous technology with a European technology; the Russians did the opposite thing, they adopted the Aleut technology. And that fascinated me. I think it’s why the Russians are the people keeping the Space Station alive, because they know how to fix plumbing, and do things.
M.T.: We’ll come back to the Russians and the Former Soviet Union in a few minutes, but I guess it’s sort of an irresistible question to ask, you obviously had an unusual childhood, growing up where you did and with whom you did. Can you just give us some sense of what it was like to grow up in Princeton at the time with all those people around?
G.D: The first thing I will say is that I saw my childhood disappear on my way over here, I was walking, and about 3 blocks from here and there was this gang of people taking card catalogues – old oak card catalogues out of one of your buildings and just destroying them with sledgehammers and just putting the remains in the dumpster. I grew up surrounded by oak and card catalogues and that old world is gone.
As a child I lived in this place called the Institute for Advanced Study, which is completely independent of Princeton University, which people always mistake it for. It was created as a paradise for scholars of two classes; people who were invited for one year, and people who were invited for life. There was nothing in between, in fact, Benoit Mandelbrot talked about that as being a very “strange distribution of people.” So you had lifers and one-years, and nothing in between, no students. It was a very odd world, but children have this way of adapting to whatever world they grow up in.
M.T.: So what was playgroup like?
G.D: Play group was… in fact, people talk about these important schools, at the Institute there was a school of mathematics and a school of classical history and a school of medieval history, now there’s a school of string theory, but the most important school was Cross Roads Nursery School, which was the school for the children who lived there. I learned a lot at Cross Roads Nursery School.
M.T.: Have you kept in touch with any of those….
G.D.: Yes, in fact, Dominique is here in the room, quite the girl next door. It is a very important place and a lot happened there, some of which is known and some of which is unknown.
M.T.: So Turing was obviously important at the institute but also with your recent book, and as a student of philosophy and religion, I think images matter. Why did you choose the image of the cathedral for the title of your book?
G.D.: It has several levels of meaning, most importantly people understand the cathedral as something that is built by large numbers of individual craftsmen to honor something larger than any craftsman. Cathedrals are built, in many cases, over hundreds of years and this computational world we live in really is a cathedral. It is built by large numbers of anonymous craftsmen, who usually don’t get the credit, but the end result is this cathedral and in the case of computation, we really owe the fundamental idea to Turing.
Now more specifically, when Alan Turing began to think about what we now call artificial intelligence, he forced a lot of the opposition to say well, well only God can create intelligence, you’re sort of messing where you shouldn’t go, and his answer to that was a wonderful statement that, in creating intelligent machines you are no more creating intelligence than we are creating souls in the creation of children. We are only creating mansions for the souls that only He can create. Then in 2005 I went to visit Google when they were just going public, and the engineers showed me what they were doing, and I just walked out of there into the Palo Alto evening saying, “my God, this is not Turing’s mansion, this is Turing’s Cathedral.” It’s a double entendre; it refers partly to Turing, and partly to Google itself.
M.T.: Most of us know just the very basics about Turing, tell us a little more about how the work that he has done continues to shape the world you are now studying and investigating.
G.D.: Turing, like many famous people is hugely misunderstood. He is remembered for a couple of things, which he is unfairly associated with. It is the same as Darwin, who was not really a Darwinist. People can have something named after them that can be something that they didn’t really believe themselves. So, Alan Turing landed in New York as a 24 year old graduate student who came to Princeton University to study under Alonso Church. At the age of 23 he had written a paper, where he outlined… all of the computers we use today are effectively implementation of this idea of what we now call a universal Turing machine. It is a single device that can do anything that any other device can do because you feed it the description of the device you want. That’s why you buy one IPhone- and then you download Apps- that’s a Turing Machine. But even by the time he landed in Princeton, that was old stuff to him. He was on to thinking very deeply about non-deterministic computation. What we now call non-Turing computation – the idea that you cannot predict what the computer is going to do next, and that’s a much more interesting question.
M.T.: So let’s go back to Google. There was some kind of suggestion of an “aha” moment there, when you came out of that, so what did you see?
G.D.: When Turing landed in Princeton, in 1937 or 36, he started thinking what’s next, and he came up with the idea of the Oracle machine. The Oracle Machine is a Turing Machine that proceeds deterministically, step by step by step, and then in gets to a point, and it makes a leap – of what we would call intuition. A non-deterministic jump and then it says, “ok, here we are” and keeps going step by step and makes these leaps, which is really how our minds work.
M.T.: You use the word intuitively, tell us more about that.
G.D.: Right. He was interested in how you define what intuition is; a question of what is the difference between ingenuity, which is just being clever, step by step, and intuition, which is taking this leap. So this oracle machine gets to these points, and then it asks a different kind of computer, maybe an analogue computer that is, non-deterministic computer, and takes that leap.
And then if you look at what Google does, all Google is, is this enormous collection of strict deterministic Turing machines, those are the perfect servers that work. But then now they are using the human beings as the leaps of intuition. When a human, who is logged into Google, and is given a bunch of search results, and makes a click, that is that intuitionistic/nondeterministic leap. You put them all together and you get the oracle machine Alan Turing was talking about.
M.T.: Alright, let’s talk about the notion of emergence, and emergent phenomenon for a moment, and how they work in these systems.
G.D.: Emergence has a sort of fuzzy definition of being a behavior as a whole that cannot be explained at the level of the individual part. You can’t understand a book at the level of individual words, you have to put it all together and then a picture emerges. It seems to be, maybe not a universal law of nature, but it is the way all societies work most of the things we know that are large systems work in emerging ways.
M.T.: Right, so one of the questions about emergence is; is emergence is itself lawful? Might emergence be nondeterministic?
G.D.: Yes, that’s the sense, the first word borders on chaos, and can be mathematically defined. (inaudible) Even if you have a planetary system where the path of each object can be almost perfectly defined it quickly becomes so complex that you cannot predict the outcome of it.
M.T.: Alright, so let’s come back to Google, you have all of these Turing machines monitoring, and the nondeterministic moment is the intervention of the human. Some argue, as you well know, that these emergent complex systems are isomorphic across mediums. That one way to explain consciousness itself is as emergent phenomenon. For now, the nondeterministic moment or the emergent moment that you saw at Google was introduced by the intervention of the human. Do you think it is possible to create these machines where you have emergent phenomenon that approximate what we have in our consciousness?
G.D. Yes, I believe the answer is yes, but, there’s a lot of argument about that. The first most important thing is that Turing himself, in his earlier result he showed that even in a strictly deterministic Turing machine, which was really this great result that brought him to America, was showing that there is no systematic way to predict from looking at a string of codes, what that code is going to do, except by running it. That was a really profound result. That really explains why the digital universe will always be interesting. No matter how much we study it, it is always going to produce mysterious behavior, and that is an important fact.
M.T.: You said a few minutes ago that Darwin was not a Darwinist, I want to come back to Darwin, I want to do that by going to some of what you suggested in Darwin Among the Machines. In Darwin Among the Machines you discuss your father’s short work, The Origins of Life, while your father was, of course, one of the most important mathematical physicists of the 20th century, this book is remarkably pressing. You write of your father’s work, “in 20 years Dyson developed the toy mathematical model that allows populations of several thousand molecular units to make the transition from disorder to order with reasonable probability. These self-sustaining haphazardly producing autocatalytic systems provide energy, and information gradients, hospitable to the development of replication, perhaps first of the parasite infecting the metabolism of primitive precursors of modern cells. Once the metabolism is infected by replication, as Darwin showed us, natural selection will do the rest. Turing and (inaudible) both died working on biology. You had written that they would be fascinated by the way that biological code and digital code are now intertwined. Can you explain to me the interplay between biological systems and information systems?
G.D.: Yes, I wrote that, I think 18 years ago, when the interplay was mainly one way. We were starting to read genetic code, at great expense, it cost hundreds of millions of dollars to read a human genome, and we weren’t writing it. And now, we are not only reading it, for pennies, but we are fluently writing the other way. And not only writing nucleotype sequences to DNA, but writing to proteins, which is the interesting part. When you talk about the boundary between reality and science fiction, when Michael Crichton was writing his book, I mean, now, computers are directly writing to proteins. They can write to a hard disk or they can write to your Dropbox account or they can write to a protein. That’s a world that was unimaginable not long ago. For better or worse. It’s a sobering thought, we are in a way being pushed out of the loop.
When the people at Columbia were thinking about this sixty or seventy years ago, they were trying to find the logical language of the brain. And it turned out the other way, that the genetic language was actually digital, and the language of our brain is non-digital, so it’s reversed. But the language of computers is digital, and the language of genetics is digital, so they are speaking directly to each other in a way that we are sort of left out.
M.T.: So, say more, because, earlier this afternoon you were telling me about it, and I had not known about the work here at La Mont at Columbia, sixty or seventy years ago, launched in this whole trajectory we are here talking about tonight.
G.D.: It was during WWII, when there was this big life or death war going on where everybody was doing their part. There were people like Vannevar Bush in the office of scientific research who were trying to put everybody to work, and a group at Columbia was formed called the statistical research group, who were given real hard technical military questions like, for a given fighter plane, is it better to have four twenty mm guns or two eighty mm guns. The particular problem was the reliability of munitions. If you had a million machine guns and aircraft shells, how can you predict when you are going to get a failure. So they actually invented something that, through sequential analysis, did a whole lot of fundamental science, solving these real problems for the war. This statistical research group, that Julian Bigelow was involved in, who became (inaudible) engineer. So in a way, the whole problem of digital computing was how to develop reliable sequences of events was very much an analogue of developing reliable sequences of munitions during the war. Another member in that group was Milton Freedman, who was an economist, who for better or worse, gave us a lot of our economic theory. So the group here at Columbia definitely was very important.
M.T.: So the two strands here, that intersect yet again, we’ll come back to the biological phenomena, are the military and economic sides of this. You recently spent some time at Los Alamos –
G.D.: Actually, Sandia.
M.T.: Right. One of the things you suggested this afternoon was that you really think that the most interesting work is going on in analogue and not digital.
M.T.: So can you tell us a little more about that, about what you mean by that, about what kind of research you saw, and the relationship between analogue and digital, and why you think the analogue computation is going to be increasingly important.
G.D.: The short answer to the question is that nature could have used digital computation but used analogue and it works very well in nature. Why are we not trying that with our machines? And, I’ve been preaching this for a while, and I got a call from the people at Sandia, which is sort of an offshoot of Los Alamos who built the first nuclear weapon. But it was way up on the top of a mountain, and wasn’t a good place to build lots of them, so the military needed a place to actually produce them, and that became the Sandia National Laboratory. They have a large computing facility because they need to build their own chips to go in the weapons, which is idle a lot of the time. They have a very creative group looking at new forms of (inaudible), which was new to me, because, they are not secret. They invited me to give them a talk about the history, because all the people from the old days- the van Neuheims and the Julian Bigelows, and even Alan Turing, would be shocked to come back to the world today, sixty years later, to find that we are exactly copying this very crude machine they built in 1946. All we are doing is making faster copies of it- not different. Why not try something different?
M.T.: You were explaining that the digital and the analogue are not antithetical, but they are interrelated so that they digital can generate the analogue.
G.D.: Right, they can work together, so effectively that we have digital coding in biology. Digital coding is very good for errant direction, which is essentially what we do as organisms. We are analogue organisms and once a generation, we reproduce ourselves through digital code, so that they errors are corrected, and we go back into the analogue cycle. And computing is very much the same, analogue computing is very robust in a different kind of way. And the military, for obvious reasons, is interested in robust machines, machines that don’t fail. In fact, coming through the airport, JFK was a huge monitor that was just saying buffer overflow error #1022 or something. It had probably been saying that for about eight hours. Analogue systems don’t fail that way.
M.T.: Earlier you had said that there was something sobering about humans being left out of the loop. So let’s come back to this notion of self-replicating machines. Part of what goes on here, is the capacity to now create programs that humans themselves couldn’t program.
G.D.: Right, well, that’s true already.
M.T.: Right, so, project out- is there a foreseeable area where the machine takes over, or in Crichton’s words, where the “bots” are released into the wild.
G.D.: I believe they were released into the wild in 1946, I mean they’ve totally taken over. Walk out on the street and half the people are following directions they are getting from this chip that the carry around, telling them to turn left down the next street. And most people are happy about it. It is not necessarily a bad thing. If you ask me, “when is it going to happen?” Well, it already happened.
M.T.: Okay, so let’s take one of the places where it is pretty clear that it has already happened, money markets and the issues that go along with that.
M.T.: I want to approach the issues here in terms of the problem of speed. So before we get too technical here, let us talk about the whole issue of speed and acceleration, and the implication of it that you see.
G.D.: Well, I think the person who can make the most powerful statement to that is Julian Bigelow, who was here in the 1940’s, and built the first computer. One of the last things he said, and he gave almost no interviews the rest of his life, but when he did, he said, “no time is there. Sequence is different than time, and the computer I built ran on sequence, not time” That’s a very fundamental thing, the difference between the digital universe and our universe.
M.T.: So, say more about sequence, because that might not be common sense to us, we mostly think of time as sequence. You are distinguishing senquentiality from temporality.
G.D.: We tend to equate them because we have clocks that tick, and an hour later it is 2:00. But in the world of computation, it is based on a sequence of events, not a time. We say that our chips have a clock speed, which gives us the illusion that they are running on the same kind of clock we are, but they are not. One thing simply happens and then the thing happens, and our chips are getting faster and faster and faster. A computer now is billions of cycles per second. That world is moving in a completely different time than our world.
M.T.: Alright, so this might what you were getting at with another quotation from John Brockman’s blog, The Edge, you write, “in 1945 we actually did create a new universe, this is a universe of numbers, with life of their own, that we only see in terms of what those numbers can do for us; can they record this interview, can they record our music, can they order books on Amazon? If you cross there in the other direction, you cross into the universe of self-reproducing digital code. When I last checked it was growing at 5 trillion bits per second. And that’s not just a metaphor for something else, it actually is a physical reality.” Then you go on to say that this alternative universe that we’ve created operates by laws very different from our laws. Talk a little bit about that alternative universe, and the difference of its laws from ours.
G.D.: If you imagine you are in that universe, looking at our universe, we would appear to be slowing down, people would be just like the end of a phonograph record, just coming to a stop. It’s a pretty strange world. And the problems are, in the situations like we just mentioned, where we put our real money, our real, Wall St. money, that used to be here at Wall St., into Wall St, except it’s not on Wall St. anymore. It’s over there in Hoboken in a big bunch of server firms, that’s where the money is. And it is going at that ever faster cycle, to the point that, as I understand it, the only legal speed limit, barrier, is that legally, if you move your trading servers, if you buy the lot across the street, and put them 500 feet away, you still have to connect with a 1,000 foot cable, so that no one can get closer than 1000 feet at the speed of light. That is the only rule. So it is out of our hands.
M.T.: In Chicago they say you can’t participate in these markets, because of the time play so they realize you suggest, that banks will try to build their physical universe still matters. So they build their computers as close to their server forms as they can. So let’s look at this in the form of a concrete instance, that many of us have read before, and that is the flash crash. You contributed to what I think is a very fine documentary, Even Money. It is distributed only as an app I think. You should get the app and look at it. It is a documentary that is a second by second analysis of the day the market plummeted and went right back up. So say something about what is going on here. There is a universe that is operating at a super-human level.
G.D.: So the SEC, they were gathering market data, big data on intervals of seconds, or tenths of seconds, and they had a big investigation, they said “oh it looks ok to us, it was caused by this computer bid and that was it- move on, nothing to see here, folks.” Dutch public television, VPRO, sent a documentary crew to look into this further, and they found an analyst in Chicago, who had been very interested in collecting microsecond data. He was independently wealthy and had his own servers, and he had the data. That is what is on this app that you can download for free from ITunes, or whatever, And the data is there, you can look. It happened at a timescale that most people just don’t look at.
M.T.: So basically, they are trading at fractions of fractions of seconds?
G.D.: Yeah, literally, I mean, it’s under 25 micro seconds, millionths of a second, to execute a trade, so the trading can only be done by algorithms, it can’t be done by people. So again, this world that Michael Crichton could only imagine in science fiction, is here. They don’t mess around with words, they actually call them “Black Boxes.” And these black boxes are codes that live and die by whether or not, at the end of the day, can pull money out of the market.
M.T.: Seventy percent of all trades on the New York Stock exchange are now done by program trading, and I think about nine or eleven percent are done at these hyper-speeds. So, talk a little about how all of this is related to dark pools. Maybe we ought to define that.
G.D.: Well, a black box is a machine, a black pool is a black box full of money.
M.T.: In other words, once you have these algorithms running at speed, executing trades, you lose your edge, so what they try to do is hide the money, as it were.
G.D.: Right. There is no time for humans to be at all involved, by the time you check back in to your office in Manhattan, it is too late. So the codes are left to trade autonomously, and it’s a very fertile system, because of course, it is selecting for better code, and selecting competitively for better code. You couldn’t design a more interesting laboratory to design these interesting codes.
M.T.: You also indicate that, in the course of working on the phenomenon of flash crashes, that there are improprieties, if not illegalities going on throughout all of this.
G.D.: Yes, that is the very difficult legal question, where does the behavior cross the line, what is fair, and what becomes unfair? What is illegal, clearly illegal, is insider trading. If you figure out a way to start gaming the system, where your codes have access to information that at the speed of light has not yet reached the open market, to me, that sounds like insider trading. But it hasn’t been enforced. I think they are ahead of us in Europe, trying to put delays on this so you don’t have these black areas.
M.T.: You probably saw, yesterday on the front page of the New York Times there was an article about the SEC trying to figure out how to put brakes on all of this, the problem is that the people at the SEC have no idea how it works, so they had to hire a firm that specializes in speed trading in order to help them develop a system to check the system that trades at these speeds.
Let’s shift gears somewhat. As if we haven’t been apocalyptic enough, talk a little bit more about some of your concerns, not necessarily related to the digital worlds we have been talking about, but to the role of nuclear power.
G.D.: Well, nuclear weapons. I was very happy to walk into the building and see that the poster for this series had the good old nuclear, real apocalypse on it, which is not the new apocalypse, but, I believe it is the image of the French Mururoa hydrogen bomb test and that is the real apocalypse. That is the apocalypse that is still with us, although we seem to be strangely ignoring it. I mean, we are fanatically upset about Iran enriching a little uranium, which could make, at most, a few, small nuclear weapons, while we still have an enormous stockpile of thermo-nuclear weapons- as does Russia, as do China and France, Britain and Israel, which are designed to destroy large civilian populations, and that is a real threat that is really with us. And is still the most important threat that we face.
M.T.: You have expressed concern about things that we aren’t doing, that we were doing somewhat more of during the 1950’s and early 60’s, in terms of developing systems of possible response in the case of nuclear attack.
G.D.: Well, we had a civil defense network, which you can argue was just sort of public relations made to make our side feel like they were doing something, that the situation wasn’t as disastrous as it was. Saying you would be ok as long as you go in a fallout shelter, which may or may not have been true, and it was also a bluff to make the other side feel that their preemptive nuclear attack would not destroy us. But there is a value in civil defense, and it is odd that we have homeland security, but not civil defense, I think it should be the other way around. We should have civil defense and eliminate homeland security. The person who I think said it best was Marvin Mitzki, the day they started homeland security. The day it was announced, he said “what we need is a department of homeland arithmetic.” And it’s true.
M.T.: As part of that, you’ve also suggested that there’s no back-up system that we have for the internet. In some ways, the internet was created as a back-up system.
But we have no back-up system for that back-up system, and we have become completely dependent. So, whether a cyber-attack or another kind of attack that wipes out the grid or the internet, the problem is, what do we do the day after?
G.D.: We are paralyzed. This is actually Dan Hillson’s, idea, the internet was set up as a back-up in case everything else went out, now everything is on the internet, and if it does fail, for any reason, we have no plan to slowly reboot. As irritating as Windows can be, at least it can reboot, it has a safe mode to boot up to the most basic essentials. In every school there should be some people taught how to get the system back up, learn to use it as a text only system, so we can slowly spread the information and get the system back up.
Now what happens when it fails is that the next day, everybody will be trying to check Google and Facebook at the same time. It’s like when you have a power failure, and everybody turns on everything in the house. “Oh the lights in the living room are not working, let’s try the washing machine, and see if the garage lights are on.” So all of the lights in the whole grid are on, and as soon as they start the generators, the fuses blow.
M.T.: Isn’t the point of greatest vulnerability the grid?
M.T.: I mean, it is ok to have a plan B if the internet goes down, and get text only and everything, but if there is a cyber-attack, or anything that takes down the grid, which may not as difficult as it appears, or at least, take down significant portions of it, I mean, when the lights go out-
G.D.: Right. I was a kid during the big Manhattan blackout. You turned on your transistor radio, which had some battery in it, and they told you to turn off everything in your house. This was the first thing to do. And we have no such information. The first thing everybody will be trying to do is to log on to what’s going on, checking on CNN.
M.T.: If you think about it, taking it back to financial markets, the issue then is not what is money, but where is money.
G.D.: And where is it right now?
M.T.: No one knows where the money is.
G.D.: And the question is, what is now?
M.T.: Money has become current, that is, electric. Imagine a global power failure where the money disappears, and the whole financial system collapses, and there is no back up.
G.D.: One of the books you sent me was about Nixon abolishing the Gold Standard which now seems like- “wait, we had a Gold Standard? When did we have a Gold Standard?”
M.T.: 1973 was when it concluded.
M.T.: And what happened was, in 1973 two years after Nixon suspended the Gold Standard Initiative, in 73, we went to floating exchange rates, and the same year that we went to floating exchange rate was the year that Reuters installed the electronic global currency trading system, and when we had that situation, we entered the world of virtual economy, which we have now, in which currency becomes liquid and virtual. But the virtual still depends on the material, insofar that it depends on the electricity. And again, I don’t know if you saw in the Times a few weeks ago, a two-part series on the energy consumed and required by these server farms-
G.D.: Except his numbers were completely wrong.
M.T.: Is it worse?
G.D.: No, it’s not that bad. It’s 30 gigawatts, it’s nowhere near that
M.T.: Oh, so I don’t need to worry about that?
G.D.: No, don’t believe everything you hear.
M.T.: It is interesting to me that from what you’ve said, the thing that really keeps you awake at night is the nuclear side of all of this, rather than the cyber side. Not that they aren’t related, but that’s really what you are worried about.
G.D.: Yeah, well, that’s still a very real problem that is with us, and we don’t worry about it as we should. Russia is a very unstable place right now, and China is clearly an unstable place.
M.T.: So Mitt Romney is right?
G.D.: Maybe, I mean, the first thing I will say is that we are too worried about terrorism, and not worried enough about the large stockpiles of big weapons. There is an enormous difference between a crude, homemade weapon and the things we developed in the 1950’s.
M.T.: Also, and I know this is a particularly vexed issue for you or perhaps your family, but our last session in this series is going to on about climate change. That doesn’t worry you as much as some of these issues?
G.D.: No, but, of course I live in the Pacific Northwest, where a little global warming never hurt. But, it is a real problem, but the problem is to do the right thing as the solution, not to do the wrong thing that does more harm than good.
M.T.: Good, well, we have covered a lot of territory, and many of you have questions or comments, so let’s open it up now. Just direct your questions or comments to George, so let’s start over here.
Male voice 1: Thank you, I’ve read a little bit about quantum computing, and you didn’t mention it you talked about analogue computing, and I was wondering if you could talk a little about how the two compare, and what quantum computing really is?
G.D.: The problem with quantum computing is that it isn’t really here yet. Analogue computing is alive and well, and available off the shelf in large quantities very cheaply. Quantum computing is a very interesting possibility. It is a very different way of doing computation. Personally, and this is just opinion, I believe in quantum computing, but that it has enormous input-output problems. So I think the practical applications of it are a very very very long way away. But, I hope to be proven wrong. I’m like that person saying “airplanes will never fly, because they don’t have feathers” Alan Turing got his big boost in computation during World War II breaking codes. So, similarly, it’s the spy agencies that are pushing very hard into quantum computing, and I think those will be the first applications, factoring large numbers, but as for practical applications, the jury’s out. In 50 years there may be lots of quantum computing or it may be a dead end, I don’t know. There are large numbers of very good people working on it – I’m not well informed.
Male Voice 1: Can you say a little more about what the difference between analogue and quantum computing actually is?
G.D.: Yes. In normal computing, digital computing, you have gates, and they are either on or they are off, there is no in between, in analogue computing, they can be in between, neither on nor off, and that becomes very interesting, that is how brains work, you have states that are continuous. In quantum computing you have something even more interesting, very puzzling, you have states where a gate exists in multiple states at the same time. So you can be listening to this lecture and listening to another lecture across the campus at the same time, and that is eternally perplexing, but that sort of is the way the real world works. Things actually exist in multiple states, and then they coalesce into the reality that we live in.
Quantum computing is a way at sort of cheating at that. And some people really believe in it, and some people think it is impossible. I’m of the wait and see opinion. I wouldn’t invest in it, but if I was the National Security Agency, I would certainly be working hard at it, because it might be very important.
Male Voice 2: A new question, you were saying that the revolution is already happening in technology, but I think something that is on people’s minds about that revolution about the agency of those machines and when the agency and intentionality(inaudible)
G.D.: Yes, to me that question about intentionality is an unanswerable question, because it’s the paradox in artificial intelligence, if it’s complicated enough to be intelligent it is too complicated to understand, and if it is simple enough to be understandable, it’s not complicated enough to be intelligent. By definition, real artificial intelligence will not be understandable to us, because if it is understandable, we will say it is not intelligent. So there is this paradox that it only becomes interesting beyond what we can understand.
This is why we are here. It is almost a religious question; do you believe it has intentionality or not? I went to a very interesting game theory meeting where someone had not gotten the instructions and came from Edinburg and gave a beautiful mathematical proof using, it was a van Neumann conference, using classical van Neumann game theory not proving the existence of god, but proving that if there is a god, no matter what value function you choose the payoff is higher if god does not reveal herself. And I think the same is true for AI. You can make the same mathematical argument for AI. The payoff will always be higher if it does not reveal itself. But that does not prove the existence of AI, it just answers the question of why it may not be visible to us. But that’s really philosophy, not science.
Female Voice 1: Thank you George, for all the comments, I was curious about your thinking, and if you could give more comments on big data, which is such a buzzword today. Yet the efficacy of computation is only in part how the computation itself is being done another important part of the equation is the information or the data that is being used in the computation, at least for some applications. So I was wondering if you could talk about the intersection of those two thoughts.
G.D.: The second half of the question was the intersection of big data and computation.
M.T.: Maybe you should start by defining big data.
G.D.: Right. Just last week I was at a conference on big data, and I didn’t know what big data was, so I invented my own definition, and they loved it. In my understanding, big data is what happens when the human decision cost to throw something away- the cost of our deciding to throw something away, is bigger than the machine cost to store it. Then you get big data. Nothing is thrown away.
I think this is a very important transition that we have made. We can probably pinpoint the moment we crossed into big data, it is probably in the last two or three years that data storage became so cheap that it is too expensive to throw anything away. It is cheaper to store it all. It costs you more to decide what emails to delete than it does for gmail to store them all for you. So your question is how does this world of limitless data intersect with the world of also unlimited computation. The two things are growing together. The only way we can survive this deluge of big data is that we have all of this processing power, searching power, to find the email you want among the 10 million you never threw away. Miraculously, we are surviving, but it is an absolutely unprecedented change. That’s where it really is an apocalyptic change, from a world where paper costs money, and disk space costs money, to a world where it’s-
M.T.: Yes, but that comes back to the issue of the server farm. There is no paper, but there are the machines on which it has to be stored, there are the minerals required for the production of those machines, and there is the energy that is required- maybe it is not what the Times said, but it is still significant, and part of it is clean, but much of it is not. And as I understand it, only a small portion of the total storage capacity is used at any given time, because, what they are doing is having to back up in case there is a glitch or something, so that everyone does not get angry when something goes down. So there will be physical limitations to the amount of data that can be stored, in that sense, right?
G.D.: Yes, I mean, eventually the world becomes full of chips and nothing else, I mean, that is the world we are heading for. If you graph the growth of all this stuff, it is like an infection that is taking over the entire world, and it will have to reach a limit at some point. What that limit is, and what we are doing with cloud computing, and these buzzword things is using it vastly more efficiently. It is remarkable how quickly we are developing more efficient ways, but we are using more and more of it. Where that goes, I don’t know. The human brain uses something like 15% of the energy budget of your biological system. We’re almost at that point already if you look at the percent of the energy budget that goes into running this computational world.
Male Voice 3: Could you explain the use of 3D computing in the manufacturing process?
G.D.: In the current manufacturing process? Or do you mean 3D printing?
Male Voice 3: (Inaudible)
G.D.: Computing, yes, well the interesting thing is that most of our computers are built entirely two dimensionally. The circuits are drawn and printed on a flat plane with a certain number of layers, but the important thing to compute is the address matrix of how you address things which is still a two dimensional space. This is one of Julian Bigelow’s pet peeves, why are we still doing it like that? Why don’t we make it fully three dimensional? And there are people working on that.
It is mostly a memory problem, if we go to three dimensional memory you vastly increase your storage space, but I think that it is inevitable that we will get to that, even though we have become so good at the two-dimensional, and it works so well and is inexpensive, and we keep making it smaller so that these factories are down now to billionths of a meter scale, so they haven’t really needed to go 3-D because they keep cramming more and more into two-dimensional space. I think it is inevitable that we will have that transition to fully three-dimensional chips, and that will be a whole new ball game. You might know more about that than I do.
M.T.: You had a question?
G.D.: Yes, many of the problems are the same old problems that have been with us from the beginning, I mean, the problem of large three-dimensional hydrodynamics which was why we got these computers in the first place; to design nuclear weapons. It is still a real problem, that’s what they spend most of their time doing at Sandia and these other labs. The other problem is cryptography, breaking supposedly unbreakable codes. If you look at who is driving the building of super computers it’s the spy agencies or the weapons designers.
These are the two forces, and now we have this third front, which is in a way new, is the big data people. Facebook is keeping track of everybody all at once and that becomes a huge mapping problem that is also driving the growth of these computers, and then there is the left field, inventing, back to quantum computing, inventing what you would do with quantum computing if you had it. So thinking in advanced “well if I had quantum computing, what would I do with it?” Often those are the things that get done and then forgotten and then later when we do get quantum computing, people have already found some ways to use it.
M.T.: There was a question over here.
Male Voice 5: Just a couple of comments really, following on the agency and intentionality thing, (inaudible) another thing, there is a blur between agency and intentionality, (inaudible) where does it start. And the other comment (inaudible) when we are talking about dark pools, isn’t it reasonable to talk about intentionality (inaudible)
G.D.: Yes, the intent is to make money at the end of the day, or at the end of the next millisecond. So there is very clear intent. My hero is Samuel Butler, who believed molecules were alive, said you could never divide between life and non-life, you have to consider that every molecule has a life of its own. You put them together and you get more advanced life. I think it is the same when you have a whole field of these little agents, who each have their intent, and then, what is the intent of the whole thing? It is an interesting question. You get all of this interesting behavior.
Male Voice 6: Do you foresee a point, with any of these topics, where society says “enough” and if so, does society even have the ability to effect it?
G.D.: That’s a very good, very important, and very difficult question, because historically, usually when governments have stepped in and tried to prevent something, it has had tremendous unintended consequences. The high-speed trading is a good example of it. The fact is, you have people arguing both sides of it “well if you try to slow this down, it will make the system more unstable, we should just let the algorithms get better and better.” I’m not so sure, because then, who is taking the money out? We do have speed limits on our roads, and for very good reason, we have regulations. We were very clear, an unfathomably long time ago- in 1975, when we could see this recombinant DNA coming down the pike, we had big series in congress with all the right people speaking to set some limits on recombinant DNA research, and I think that worked. It did slow things down at an important time. In Europe they are having very strong, unified government hearings on the question of digital privacy. How much should private companies be allowed to do with the data they collect from individuals, and place some limits on that. I’m not a complete libertarian, I don’t believe people should do whatever they want.
Male Voice 7: So, your statement before that as the servers in New Jersey are able to make more and more decisions, (Inaudible) to what degree do you think, people need an agency that essentially starts to record things at their highest speeds, so that at least retrospectively, they can go back and see the (inaudible) that are going on?
G.D.: Yes I think you are absolutely right that we have the Securities and Exchange Commission, and it is inexcusable if they are not collecting the full data in full real time and preserving it, because you can end up with the question “where was the money” when someone lost it. I think now they have been shaken up a bit, they probably are collecting better data. How you regulate it, and how you define what the clock really is, these are real problems that are going to end up in big law suits. It is amazing how quickly this world has changed.
Male Voice 6: You said earlier that the industry had to hire a group of people who were experts in this in order to do it, so the ability for these agencies to police is only as good as their relationships with those communities that have this expertise? So, from what you have been exposed to, how open are those communities to giving out those secrets?
G.D.: Oh, not at all! By definition these things thrive under a certain amount of secrecy. I think there is an easy solution, but, for some reason, it is unpopular, which is just simply to charge a very small incremental – I mean, it can be a hundredth of a penny for these fast trades. So then if you are trading a million times a second it adds up to real money, and you pay it, it’s like a tax.
It’s a very strange system. It puts a certain friction in it, which is probably good. It is not good to trade that quickly.
Male Voice 7: There are a number of services now where many people can learn to code, (inaudible) What do you think should be the literacy of lay-people to understand computer code, and understanding really how computers work? People rely on these machines without really understanding what they are or how they work. So what is your opinion of what everybody ought to know.
G.D.: Well, I think people should not use machines that they don’t understand, and that’s how I came to this,
M.T.: You mean we can’t drive?
G.D.: Well, no, I believe in taking apart carburators. But I’m a completely hopeless romantic relic of the past. That’s how I became interested in computing. “Hey I’m using this machine, and I don’t know how it works, and the way I learn how something works is by understanding how it began. Now whether you can teach that to everybody, I don’t know, but I think it should be a fundamental part of modern education. I think it is criminal that high schools no longer have shop classes, where you learn something about machine tools, because the things we use are built by machine tools. We should understand that, at least, have the opportunity to understand that that if we want to, and I think teaching basic what computers are and how they work is very important. Otherwise they become a sort of mystery that you tend to be afraid of. So I think that is important, but I think I am in the minority, most people just want to know how to use it, and not really understand it.
M.T.: Your last question was a good one, you want to wrap it up for us?
Male Voice 7: Following up on the question of government regulation, what is the future role of the state, what role will these emerging technologies have in, on one hand, bolstering the state, and (inaudible)
G.D.: I think the role of the state is very important, and that only time will tell. And I think it is up to us as citizens to try and support a more sensible state. The danger of having an equivalent of 9/11 or something where there is a disaster, and we end up with the computer/ homeland security department where you have to remove all of your toothpaste to get on a plane. What we have now is an auto-immune disease where we are suffering an inflammation of being immune to our own system. You don’t want total regulation, Alan Touring proved that you cannot ever totally regulate a digital system, but, you also don’t want zero regulation. There is a middle ground, what it is, I don’t know, but governments are being dragged into it whether they like it or not, because so much of our lives depend on it(technology) now.
M.T.: In some ways this is a good point to come to a conclusion on, which is; civil responsibility in understanding these issues in technology. One of the things I admire about your work is your ability to explain these kinds of complicated processes in ways which laypeople can understand, and appreciate the significance of their implications in the lives that we are trying to live. So for that, and your presence tonight, thank you very much.
G.D.: Just one closing moral argument is that this whole thing began effectively with a deal with the devil. Van Neumann wanted a computer, and said “I will build the devil’s worst weapon, if you give me the computer.” And we felt that we’ve done great – we haven’t had a thermo-nuclear war. The devil just disappeared. But my theory is that we have to be careful, because the devil could come back and say, “Well, you owe me. I don’t want the thermo-nuclear weapons, I want the computer, and that will be my weapon.” So we have to be careful and make sure that the computer itself is not used as a weapon of the totalitarian state.
M.T.: And do we remember? The price of the first apple computer?
M.T.: Thank you all.
G.D.: Thank you.