Michael S. Malone
The following is from a Forbes magazine / Nightline interview, probably around August 1999 (or perhaps in early 1999).
1. Ray Kurzweil Biography: Mr. Kurzweil developed the first print-to-speech reading machine for the blind. He is the author of The Age of Intelligent Machines, which was named Most Outstanding Computer Science Book of 1990, and the recently published The Age of Spiritual Machines: When Computers Exceed Human Intelligence.
2. Kip Crosby Biography: As a writer and journalist, Mr. Crosby is the co-author of The Windows 98 Bible and a forthcoming book about Windows 2000 as well as articles on both computer and non-computer topics. In 1993 he founded the Computer History ssociation of California, an educational and historical nonprofit organization. He also lectures on the history of computing and the Internet.
Malone: Hi, I'm Mike Malone, I'm here today with Ray Kurzweil and Kip Crosby. Ray Kurzweil is the president of Kurzweil Technologies, Inc. and also the author of two classic texts on computing: The Age of Intelligent Machines, and the recently published The Age of Spiritual Machines; When Computers Exceed Human Intelligence. Kip Crosby is the author of the Windows 98 Bible, and executive director of the Computer History Association of California. Welcome gentlemen.
Kip: Thanks, Mike.
Ray: Yes, nice to be here
Mike: I'm interested in talking today, we have, you're an interesting pair of individuals. Ray, you're looking forward into the technology of the future, and Kip, you're tracking the history of computing and making sure that we don't lose. So I'm interested in receiving from each of you a little bit of grounding in computing and then also a sense of where we go from here. So why don't we begin with the state of the art in computing. Where are we at now? How powerful are our computers and where is the technology heading? Let's start with Ray.
Ray: Well, computers are still a million times simpler than the human brain but they're powerful enough to emulate some of the chaotic processes that we find in the brain. We're able to replicate some of the recognition capabilities, and we have systems that can understand continuous speech. There are systems that can translate from one language to another, there are new speech synthesizers available now that sounds like humans. I recently demonstrated a translating telephone where I spoke in English and the German gal heard me in German translated with virtually no delay. And then she replied in German and I heard her in English. We're increasingly going to be able to interact with our computers using natural speech and language technology. Computers are increasingly making intelligent decisions; about a trillion dollars of the money in the Stock Market is now managed by intelligent PIN recognition based computers.
Kip: Well, I think that what's most exciting about computing today that a grounding in the past of computing equips people to understand is that any computer that you can buy for say a thousand dollars today has in many ways the power of what would have 10 years ago been considered a million dollar super computer. And what's exciting to me, especially through, you know, the technologies that Ray is talking about like text to speech, character recognition, voice recognition is that the more capable computers become the more enthusiastically people accept the talents of computing into their lives. For example, so far as the best information I have, there are now computers in just over 50% of U.S. households. Now what does that mean? What kind of a paradigm shift is it? What are people doing with these computers besides getting on the net in gargantuan numbers? And I guess the question that I find most exciting is as more and more people have computers, use computers, they will find more and more things to do with them, and what will those things be.
Ray: You mentioned in computing compared to 10 years ago, of course, well known and often referred to as Moore's Law, but the phenomenon is actually much broader than that. In my book I put all the major computers of the last hundred years, Moore's Law which refers to integrated circuit, and we see that the exponential growth of computing goes back a hundred years, and in fact, it's part of a much broader phenomena which is inherent acceleration of any evolutionary process. And it's important to understand that the technical progress is progressing exponentially; hundreds of years ago major paradigm shifts took hundreds of years. Now we have major paradigm shifts like the web in just a few years time. And if we extrapolate out this exponential progression of computing, we see that that a millionfold disparity between the human brain and computers will disappear by the year 2020. It'll take us longer to organize those resources and complete the reverse engineering of the human brain to create computers with human level intelligence, but that will probably follow within another 10 years of that. So by 2030 we'll have machines that really do have the subtle and complexity and depth of human intelligence. And then they'll combine it with some inherent superiority?s they already have, one of which is the ability to share their memories. If I spend years learning French I can't download that to you, but computers can share their skills, their knowledge, their insights relatively instantly. So once they achieve human levels of intelligence in the domains in which humans are now superior, they'll combine it with their own superiority and memory sharing, accuracy of memory, speed, and they'll necessarily soar past human intelligence.
Mike: Well, let me ask you this. In The Age of Spiritual Machines, you speak of the metaphoric moment when computers match the range of human intelligence. Do you think that the development of computing as a whole has been characterized by several prior metaphoric moments. For example, when the web sprang full blown onto human consciousness in the space of about five years between 1990 and 95, would that qualify also as a metamorphic moment?
Ray: It's not quite at the same significance. I think machines, non-biological entities, truly challenging human intelligence, the ability to master human emotion is going to bring into question who we are. We have had a series of times where we've been sort of thrown off the pedestal, when we discovered after Copernicus that the earth wasn't the center of the universe. And discovered after Darwin that we weren't that different genetically from the other animals. But we've always prided ourselves on our intelligence, and to have that challenged will bring into question our identity. And it won't be a clear distinction as we go through the 21st century between human and machine. We'll have machines that are literally very precise copies and then extended of human brains and they'll very convincingly and compellingly claim to be human and people will believe them because probably their behavior will be just like a human. We'll have on the other hand humans who have a neural implants introduced non-surgically through technology that have extended their intelligence but still have basically a biological brain. There won't be a clear distinction. The web is significant as a communication tool. Communication in general has been enhanced by technology, it's certainly a very profound development. But the ability of non-biological intelligence to be intelligent enough to create its own next generation is I think at least is significant as the emergence of human intelligence itself generations ago.
Mike: Ray, let me ask you, before we get into the ontological questions, and I definitely want to get into those, let's set the ground here. First of all, how many neurons are in the human brain?
Ray: We have, estimates vary about an order of magnitude, but the most common estimates are about a hundred billion neurons, about a thousand connections from one neuron to the next, that's about a hundred trillion connections. Each of these connections are computing simultaneously. Of course, the brain is not one entity, it's hundreds of sort of specialized regions, it's essentially hundreds of information processing organs. They all use some variance of what we might refer to now as complexity theory or chaotic computing and that all of the connections are computing and they kind of establish a residence which is a solution to a bigger problem which might be identifying the cup on that table for example. They're very slow though ? electrochemical computing in our interneural connections only calculates about 200 calculations per second, at least 10 million times slower than electronic circuits. But 200 calculations a second times a hundred trillion-fold parallelism is about 20 million billion calculations per second, or about 20 billion mips. And we'll achieve that in a thousand dollar PC or equivalent by around 2019.
Mike: Now, Kip, architecturally though, the human brain is not like a computer, right?
Kip: Well, it depends on what kind of a computer you're talking about. The brain is a massively parallel device. In other words, as Rays says, you have many millions and billions of connections all firing at once. The brain is history's most successful parallel processor. Now, in the world of hardware computing, parallel processing has not had the success it deserves because it just happens that the software to enable parallel processing, the operating system and application software for parallel processing is inordinately difficult to write. However, we will be picking up speed on this as we build computers that are internally neural network computers, in other words, computers that work more like the human brain. We will be writing the software to enable that.
Ray: I agree. It's difficult to adapt classical AI approaches to parallel machines. But if we try to emulate the chaotic processes in the brain, this so called neural-net computing and highly simplified models of neurons, but there are new neuron models that are actually quite precise replicas of real biological neurons. And this is very amenable to parallel processing because they're inherently intended to operate in parallel. We do have good models now of neurons recently that scientists actually connected artificial electronic neurons with real neurons, and they actually accepted each other and kind of chaotically danced together and came to their own equilibrium and actually were able to do recognition combining both biological and electronic neurons indicating that our electronic neuron models are sufficiently accurate.
Mike: Am I correct in saying that there already exists a nascent massively parallel computer as complex the human brain, and that's the Internet? I mean George Dyson has suggested that perhaps the next great consciousness is actually the Internet itself, once the...there's enough computing power out there, we just need the faster and more powerful interconnection.
Ray: On one level that's kind of a loose analogy. I mean there is readily human brain power computing on the Internet. It's of course, not organized as a complexity theory neural-net, so it's not actually doing what even the brain is doing. There are proposals to harvest the unused computers on the Internet, but all the computer's doing on the Internet, probably about 99% of that computing power, is unused. Most computers aren't used at all. Those that are used are just a person might be typing at it and only using two, three percent of its computing power. People could run in the background and soak up those unused computers software which could be harvested into a massively parallel machine. So in other words, we could create a human brainpower super computer which would be highly parallel by sort of harvesting all these unused computers on the Internet. There are proposals to do that. But we'll have, I mean just conventional super computers will be even brainpowered by 2010, and of course, within 20 years that type of power will be very ubiquitous.
Mike: That's on the Internet, Kip?
Kip: Well, as a matter of fact, that kind of massively parallel harvesting of super computing power is being done already because for example, there is some networking of individual PCs to do the breaking down of results in the search for...yeah, SETI projects for alien presence in the universe, otherworldly presence of intelligence in the universe.
Mike: Where you donate the time on your PC over the Internet to help SETI decode the signals you're getting?
Kip: Right. And actually it's real easy to put together a super computer on the net in that way, at least in principle. You take people with Pentium 2, Pentium 3 level power, you hook together a couple of hundred those, you've got a super computer real easy. Now the thing that excites me about this and the next paradigm shift that we have to go through there is some work being done, I believe at MIT, I'm sorry I can't be more specific about this that will reduce the size of an individual hardware switch in a computer to the molecular level. And it would seem to me, somewhat talking through my hat, that if we can make a switch, a computing switch out of a single molecule or a pair of molecules, we will approach the processing density of the human brain because that has to be about the processing density of neuronal chemistry.
Ray: Well, actually those processes are far more dense. The human brain is actually not very efficient. I mean neurons are, compared to molecular computing, rather cumbersome devices and they use electrochemical means of processing information. I mean it's remarkable that this evolved on its own through natural selection. But molecular computing and I've talked about several different paradigms for that in the book, one of which is nanotubes which are these carbon lattices which would do computing at the molecular level. A one inch cube of nanotube circuitry would be about a million times more powerful than the human brain. There are many new very exciting hardware technologies that are beginning to work in laboratories that are at one stage or another of development, that will come onstream in the next five, 10, 15 years that will bridge the hardware gap. Now having enough computing power is a necessary but not sufficient condition to emulate human intelligence. You've got to organize those resources. And there are many different ways to do that. But one exciting paradigm is literally reverse engineering the human brain itself and we can do that by scanning it and actually creating a huge database of exactly how human brains are organized in terms of all the inter-neural connections, the patterns of neuro-transmitter concentrations and so forth. And there's a lot of progress being done on that. The scanning technologies are themselves getting more powerful because the computers are getting more powerful. Certainly within the next 20 to 25 years we will succeed in fully decoding the human brain just as we're now nearing completion of the decoding of the human genome.
Mike: Gentlemen, before we get into...I want us to get into the philosophical questions. Before we begin I want to set the ground on this. And Ray, what I'd like to do is summarize where you think we're going and then have you edit me as I go along. Okay? So essentially what you're saying is that the pace of development in computer technology will by the year 2020 produce systems of some form that will have thousand times the power of the human brain.
Ray: No. By 2020 we'll have at least in a routine personal computer type computer about equal to the human brain...
Mike: So the human brain for a thousand dollars, say.
Ray: In terms of hardware computing. In terms of actually organizing those resources so it has the flexibility and subtlety of human intelligence, I'm saying 2029. Because by that time the thousand dollar computer will be equal to a thousand...
Mike: Okay. Meanwhile brain scanning technology will reach the point that we will be able to map the synoptic structure of a given human brain and map it onto one of those computers, sometime in the years...within a few years after that. Is that correct?
Ray: Let's say by 2029.
Mike: Okay, so by 2029 we're able to map a given human brain onto the computer. At that point the brain will reside in the computer, we'll be able to upgrade it, we'll be able to add new functionality, we'll have almost infinite extension of that brain through the net. And we'll be able to copy it, duplicate it, thus in a sense creating a form of immortality.
Kip: And given that there's a sizable chance that all three of us will be around in 2029, I'd love to be around to be around on a Sunday afternoon when we all do that.
Ray: Now there are subtle but important issues. If we scan my brain and reinstantiated that information in a suitable neural computer, that new entity would have all my memories and would think that...it would sit there and claim to have been born in Queens, New York and grown up in Queens, moved to Boston and go to MIT. A few companies, walked into a scanner there and woke up in the machine here. And if you interviewed it would be very compelling and convincing unlike say today's virtual personality.
Mike: And that brings us to the great ontological questions. Entity, embodiment, and those sort of matters, not little things.
Ray: That doesn't mean that it embodies my consciousness. For one thing if we envision a non-invasive scan of my brain I'd still be here in my brain. I wouldn't even necessarily know about the fact that you had scanned my brain and reinstantiated it. Which argues that that's not really me, although that entity would think that it's me or at least act like it thinks that it's me.
Kip: Several interesting questions, one of which is what would the legal standing of this mapped entity be?
Ray: I'm not...legal standing follows what people believe. I believe that people will be convinced that these are conscious entities because unlike say today's virtual personalities which are not convincing, these entities really will be as compelling as a real human, and they'll also be extremely intelligent, but far more intelligent than humans are today. So they'll succeed in convincing us that they are conscious, that their feelings are genuine. And so they will get recognition and they'll be very human-like.
Mike: Does that make the source, the biological source of these entities essentially superfluous once the downloads been made?
Ray: Well, they will need bodies, talk about that issue, it's complicated, I mean they can get human-like nanotech engineered bodies, they can have bodies in virtual reality. We're spending an enormous amount of time in virtual reality.
Mike: Ray, one of the great questions is the matter of embodiment, can we be a human being without existing inside of a body.
Ray: I think you know, psychologically having a body is very important. We spend most of our mental energy meeting its needs and desires and protecting it and feeding it and so forth. But a body doesn't have to reside in real reality. A body can reside in virtual reality. Virtual environments will be very real and compelling and intelligent entities, both biological and unbiological. The three of us rather than talking by phone could have met in a virtual Martha's Vineyard or Mozambique game preserve or taken a virtual walk on a virtual Cancun beach, and it would be just as real and compelling as real reality.
Mike: Let me take it back though to the question of having a body, having a physical embodiment for this mapped reality and say that insofar as is it will be necessary, psychologically necessary to have these minds mapped into physical immediate bodies, I wonder whether the sophistication of the work done on the body or the embodiment will match the sophistication of the work done on the brain. Because if I'm going to map my brain into a physical body I certainly want one that's in many ways an improvement over what I have. And I want to know what will be possible along those lines. Will we...if we create mapped mental and emotional realities in engineered bodies, will we be able, for example, to create a good shot at immortality and what will the ontological repercussions of that be?
Ray: I come back to the virtual realities, bodies don't have to be physical, they can be the equivalent of physical bodies in virtual reality. But if you want to talk about physical bodies there's a whole another body of work being done in the area of nanotechnology which is at a very early stage, but as we go several decades in the 21st century, we'll be able to create physical entities including human like bodies that will be very convincing and have the properties of human bodies with suitably altered in various desirable ways. So we will be able to create human like entities that look and feel human. And we'll also be upgrading the bodies and brains of biological humans or humans of biological origin. We'll be able to, for example, have neural implants without doing the surgery by sending billions of little nanobots, nanorobots through the bloodstream which would take up permanent positions at various, at billions of key points in the brain. And we're able to communicate with the neurons and synoptics in the brain, not invasively. We already have means for electronic devices to communicate in both directions with neurons without an actual physical connection, just by being in proximity.
Mike: Ray, can you appreciate that a lot of people would find that the prospect of this very frightening, if only because if the end of generations is only a generation away, it seems as if philosophically, theologically, emotionally, we're hardly prepared for something like this.
Ray: We have a deeply ambivalent attitude towards death. We fear it, it's generally described as a bad thing and tragic and so forth. On the other hand, possibly even more frightening than death is the absence of death. And certainly human intelligence is about psychologically to live within a certain time span. And even just extending human life spans through things like bioengineering which is much more conventional than the technologies we're now talking about, already puts a great strain on human psychology. But we're going to be doing this a step at a time. And humans are remarkable in how they can adapt and the world is already very different than it was a hundred years ago, and we do adapt to these changes. And each one makes sense as we do it. We're not going to wake up in 2029 or 2049 suddenly and enter a very different type of world. We're going to get there a step at a time.
Mike: Well, Wallace Stevens in his great poem, Sunday Morning, presents a world...he basically presents Heaven as being a place without death and argues that it would be a most inhospitable place. And he says that death is the mother of beauty. Is there a danger that when we conquer death that we lose other things too? The imminent prospect of death tends to clear one's mind. Do we face the prospect of an empty, unaesthetic, cloudy immortality?
Ray: Right in the beginning of my book I talk about how death gives meaning to life, and that without death many things that we consider meaningful would become meaningless. It gives meaning and value to time. If we had too much of it we wouldn't be prepared to deal with that. But as I say, I think we're deeply ambivalent about it. I believe in life and trying to overcome death, and it's a surprisingly controversial notion...
Kip: Surprisingly. First of all, it's important to realize at a certain level this is going to be a matter of consumer choice. For example, some people, me emphatically included, will want entire new head to foot solid state bodies. That is going to be not a very accessible option for most of the world's population. So you're going to have the same stratification into digital reality then that you have now. Some people will go for solid state brains in solid state bodies, relatively few. Some people will go for upgrades to the brains and bodies they have. And for most of the world's population they're going to go slogging on with the standard wetwear and the antediluvian means of reproduction that we've already got. I don't think there will be any question of turning the entire population of the earth into a platoon of cyborgs. It will simply be too expensive...
Ray: I think you have to be careful what timeframe you're talking about. Because limitations like it's impossible to provide this type of technology for the whole world's population will be true at one point in time but if you go another 20, 30 years in the future and have another 10 orders of magnitude of improvement in price performance and technology, then greatly upgrading the minds and bodies of the entire human population would be trivial in cost in maybe 2089. So I don't think you can make definitive statements that remain true for very long. The exponential growth of technology is going to continue beyond the introduction of these types of technologies. Ultimately very powerful capabilities will be fairly trivial to achieve.
Mike: Excuse me, first of all, hurray for antediluvian production! Secondly, I'm very interested in that window. Kip, I think you're dead on this, literally dead on. What happens during that interval say from 2030 to 2060 when a certain number of people and they developed the world with the right resources have the prospect of stepping through that doorway to immortality, and the rest of the world is still looking forward to a life span of 45 years, and the imminent end of their existence? If you think there's considerable resentment now for people owning televisions and having nice cars, just imagine when immortality is at stake.
Ray: I don't think that's realistic. By even by 2030, well, it's going to greatly expand, as I think, some of the discussions we have on the social security deficit 30 years from now...
Mike: What social security's going to be like when everybody lives forever!
Ray: There's going to be extraordinary economy growth even by 2029. These technologies are going to be very inexpensive. It's not going to be a 30, 40 year gap. These technologies are going to accelerate much more quickly than people realize.
Kip: Okay, let me ask you this then, Ray, let's take the short term analog, pardon the term, to the phenomenon you're talking about. How long is it going to take to give the entire population of the world Internet access?
Ray: Well, let me put it this way. Look at how democratic a technology it is. It gets more and more affordable, less and less expensive and more and more powerful every year, as we've all pointed out. So while still not enjoyed by the entire world population, it's rapidly gaining in accessibility. And you do have the opportunity for teenagers in Bombay, India to acquire sufficient technology to start their own software company and market it on the Internet, and those sorts of things are happening. People all around the world are skipping industrialization and going to the information economy. And at least it's moving in the right direction, at an exponential rate...
Kip: Sure but what I'm asking is not that. The direction of the process is obvious. But let's say, and here again, this is off the top of my head, let's say that in terms of the population of the entire world at this point probably 10 to 15% is within some easy distance of easy Internet access. How long is it going to take to cover the rest of the world in that way?
Ray: The bigger barriers in providing the hardware for Internet access is the knowledge and education. Because if you have the computers in Internet access on are people that had no education, they really don't have the means to appreciate the content or to contribute to creating it. I do think there is a have/have not issue, and I think the primary thing is learning and access to knowledge and appreciating the value of knowledge which is what the Internet provides.
Mike: Okay. Then I see an analogy. It may be all very well for certain communities in nations in the developed world to accept the melding of man and machine in the name of the perfection of human existence. However when you get outside the most developed, the most technically sophisticated populations you're going to run into all kinds of for example, religious, ethical, personal objections to the cross breeding of man and machine in this way, and that's why I believe that no matter how far we get with this process there will always be the extension of today what is called the digital divide.
Ray: I think there is a have/have not issue, but it's going to be much more complex than that. And there's a slippery slope between where we are and where we're going. We're already putting neural implants in people's brain. I don't see a lot of religious objection, for example, for putting an implant in the brain of Parkinson's patients which suppresses the symptoms for eight or nine years. Or a cochlear implant for the deaf. And it's a small step from overcoming the handicaps of disabled persons to enhancing everyone's abilities. We do quite a bit now in terms of replacing body parts and even enhancing our brains today through technology, and it's actually not as controversial as one might expect. And this is only going to accelerate from what is today a very early stage of this type of man/machine interface.
Mike: Kip, let me ask you a question. You've been tracking the history of computing. And as you know better than just about anybody it's a history of mistakes, dead ends, processes that went bad and were revisited 20 years later and made to work. It seems to me that the game is getting much riskier. The stakes are much, much higher. When we start talking about things like immortality in the normal discourse that's pretty astonishing in itself. When we're playing with a game this big, how do we overcome the problem of human beings screwing up, riding buggy code and not wiring the machines right?
Kip: I think what answers that Mike is my old dictum that the solution to any problem involving computers is more computers. We will enter into an age where increasingly computers prevent human beings from making mistakes. I mean we're there already in a certain sense if you look at the way a spreadsheet works. A spreadsheet makes it much easier to do a financial projection without trivial errors. In the same way we will be entering on an age where computers write software, computers that design other computers will be managed by people. But I believe that there is, I'm actually very enthusiastic about a construct which combines man and machine in a way that reduces error. It would be very difficult for example to do anything today just in the medical sense, for example, it would be very difficult to do anything like a CAT-Scan or an MRI without the assistance of one heck of a lot of computer power. And I believe that that will spread and that will become true increasingly in other spheres of existence.
Ray: I think that's very true. Compare the design of the very early computers which were designed using pens and paper and then wired by hand to designing computers today where human designers specify only the very top level design parameters and then many levels of design are then carried out automatically by sophisticated software programs. Increasingly, important human decisions are going to be an amalgam of human and machine. It's already very true in the financial market. At least a trillion dollars of investment decisions or investment monies are managed entirely by programs combining genetic algorithms and neural-nets and so forth, and trillions more are a combination of human and machine and that's moving decidedly in the direction of machine intelligence.
Mike: One last question for both of you gentlemen. And that is that Ray, if your predictions are correct, and having made those predictions you've now put into the popular imagination the idea that immortality might sit out there a half a century from now. Given that and the fact, the realization that most of us living today won't make it, what do we say to everybody about the fact that they may be the last generation to die? We may be the last people to actually face death. And if that's the case, why that seems so unfair. Why us?
Ray: Well, I actually wrote a health book for those who were concerned about that question for both myself and others. And there are certainly means of extending human life. And their technology in the bioengineering area is going to make a huge difference. Over the next 10 years we will be reversing most diseases like cancer and also aging processes, we're really at a point where bioengineering is extending human life expectancy by more than a year every year. If you're in your 20s or 30s you really have a very good chance just through bioengineering alone of living a very, very long time.
Mike: It's like being the last...nobody wants to be the last soldier to die in the war.
Ray: Right, if you're in your 40s, 50s, 60s then you have to pay a little more attention to keeping yourself healthy through conventional means until a bioengineering revolution takes you to the nanotechnology artificial intelligence revolution.
Kip: Right. Exactly so. Because I mean every advance in medicine brings up this question over and over again in the same way that you can be part of the last generation that dies before the cure for cancer and heart disease, which realistically we probably are. You can have been part of the last generation to die before the development of penicillin or the last generation to die before the invention of anesthesia for surgery.
Mike: But yeah, Kip, this is different. This is the generation to die period. Before there was always the prospect of one's mortality.
Ray: There is this biotechnology revolution which is a little bit different than what we've been talking about. But that really will give the means just through overcoming disease, and it's also understanding the information processes of disease and biological systems, we'll be able to live a very long time, and take us well past the types of developments we've been talking about in this conversation.
Mike: Gentlemen, thank you very much.
Kip: Mike, thank you.
Ray: Thank you, I enjoyed it.