Here is the Washington Post link to the transcript
------------------------------------------------------------------------------------------
Transcript
Artificial Intelligence
When Humans Transcend Biology
Ray Kurzweil
Futurist, Inventor and Author
Monday, June 19, 2006; 2:00 PM
Inventor and futurist Ray Kurzweil was online Monday, June 19 at 2 p.m. ET to answer your questions about Artificial Intelligence.
In his latest book, "The Singularity Is Near," he examines the next step of the evolutionary process: the union of human and machine, in which knowledge and skills embedded in our brains will be combined with greater capacity, speed and knowledge-sharing ability.
A transcript follows .
Kurzweil is also the author of "The Age of Intelligent Machines," "The 10% Solution for a Healthy Life," "The Age of Spiritual Machines" and "Fantastic Voyage: Live Long Enough to Live Forever." As a noted inventor he is credited for his work with music synthesis, speech and character recognition, virtual reality and cybernetic art.
About this series: Beyond the Future is a weeklong series of live Web chats with noted experts and Washington Post reporters examining the kinds of technological advancements the world could see in 20, 50 or even 100 years. Related news on the subject can be found on the Science and Tech Frontier pages of washingtonpost.com.
____________________
Ray Kurzweil: Hi, this is Ray Kurzweil, look forward to our dialogue....
_______________________
Reston, Va.: Would intelligent machines represent a threat to mankind that should be banned by formal regulatory action?
Ray Kurzweil: The 3 great technology revolutions - G (genetics), N (nanotechnology) and R (robotics) have profound downsides. We have a new existential risk right now from G (genetics or biotech). The same technology that will overcome cancer and other diseases can be used by a bio terrorist to create a bioengineered biological virus. The good news, though, is that we have the technologies to overcome this. For example, RNA interference can combat new biological viruses. But we need to put a rapid response system in place. Each technological response will only work to a certain degree of sophistication so the message for society is that we need to put a higher priority on creating the defenses.
_______________________
Covington, La.: Ray,
It has been suggested that the neural system of the human brain is too complex to be able to work a sophisticated interface between biological neural processes and computer technology, and in the context of present technology, that is certainly the case. Seeing as how nanotechnology itself is still in its inception, how can we accurately predict the timescale and form of intelligent nanotechnology on the level of human neural interfacing?
-also-
If the Law of Accelerating Returns holds true and technology really does continue to increase at an exponential rate, then the rate of obsoleteness will likewise continue to increase. How will humans continue to "upgrade" technological components into their bodies when yesterday's state-of-the-art is today's old junk?
-Joseph Von Hoven
Ray Kurzweil: WRG the first question, a key issue is what the complexity of the human brain? We get a very different answer if we consider the apparent complexity of the brain versus the complexity of its design. The amount of information represented in a mature brain is thousands of trillions of bytes, which is a lot of complexity. But the design is a billion times simpler. How do we know that? The design is in the genome and with lossless data compression the genome only has 30 to 100 million bytes of information. Moreover, we're showing that we can create mathematical models and simulations of brain regions for which we have data from high resolution scans and other forms of reverse engineering. There is a simulation of 12 regions of the auditory cortex and the cerebellum, which comprises more than half the neurons in the brain. I make the case in chapter 4 of Singularity is near that we will have the models and simulations of all several hundred regions within about 20 years.
_______________________
Los Angeles, Calif.: How do you see the actual physical merging of human biology and integrated circuitry? How does the biological chemistry of thinking, imagining and dreaming translate to computer technology?
Ray Kurzweil: There are multiple scenarios. But one that I find compelling is that we will send intelligent nanobots (robots the size of blood cells with nano features) into the human body and brain through the capillaries. One application is to keep us healthy from inside. If this sounds very futuristic I would point out we're doing this already in animal experiments. One scientist cured type I diabetes in rats with blood cell sized devices with 7 nanometer pores that let insulin out in a controlled fashion and block antibodies. These technologies will be a billion times more capable in 25 years than they are now based on the doubling of power of information technology in less than a year. Ultimately these nanobots will go into our brains through the capillaries and interact with our biological neurons. The ability to do this at a small scale has already been demonstrated. One application: virtual reality involving all of the senses in full immersion highly realistic VR environments from within the nervous system.
_______________________
Ft. Lauderdale, Fla.: I was moved by your "Fantastic Voyage" book - what an excellent read! I would like to know what do you think is the most promising technological advance to occur dealing with radical life extension, since the release of that book, and why?
Ray Kurzweil: In Fantastic Voyage, we talk about 3 "bridges" to radical life extension. Bridge 1 is applying today's knowledge which is an ever changing frontier. Bridge 2 is the biotech revolution in which we are reprogramming the information processes underlying biology (and biology is essentially an information process). Our ability to do this is scaling up exponentially. For example, the amount of genetic data we have sequenced has doubled each year and the cost per base pair has come down by half each year. We are also showing the ability to reprogram these processes. Perhaps the most promising recent advance is RNAi (RNA interference) in which we can turn off genes. Turning off the fat insulin receptor gene, for example, enabled animals to eat as much as they want and remain slim and get the health benefits of being slim. There are other genes we'd like to turn off that encourage cancer or heart disease. And we have new forms of gene therapy that allow us to add new genes to adults. We can also turn on and off enzymes, the work horses of biology...
_______________________
Cupertino, Calif.: It seems likely that the first truly capable AIs will be created within the closed confines of a large organization such as Google, which is in a position to capitalize on the competitive advantages such an in-house intelligence would give them. What are your thoughts on how others can compete, once the playing field goes from slightly tilted, to overwhelmingly tilted in favor of those who control both the AI and the data it draws on?
Ray Kurzweil: It is true that Google's massive data bases are a unique asset in creating an AI. They already created a translation system that can translate from Engligh to Arabic and back using a pattern recognition system that found the translation patterns automatically from large "Rosetta stone" texts (translated texts in the 2 languages). The system was apparently equal to professional human translators yet no one on the Google team that created it spoke a word of Arabic. But the development of an AI that can span all of human intelligence (so-called "strong AI") won't occur in one lab in one step. There will be many competitive efforts and many flavors of AI as we approach that point in about a quarter century.
_______________________
Columbia, Md.: I have been following your career with great interest since my office acquired one of your first reading machines. And I have been fascinated by the SINGULARITY IS NEAR. I am also impressed with Al Gore's arguments that the global warming problem should be near the top of everyone's agenda. He claims that climate scientists say we have only about ten years, absent a concerted effort and political will to solve the problem, before the planet has reached a point where it's too late to prevent catastrophic consequences. Do you believe, given your arguments about progress acceleration, that these scientists are unduly pessimistic? Do you think that accelerated technology alone--even without political will to solve the problem--will save the planet?
Ray Kurzweil: None of the global warming discussions mention the word "nanotechnology." Yet nanotechnology will eliminate the need for fossil fuels within 20 years. If we captured 1% of 1% of the sunlight (1 part in 10,000) we could meet 100% of our energy needs without ANY fossil fuels. We can't do that today because the solar panels are too heavy, expensive, and inefficient. But there are new nanoengineered designs that are much more effective. Within five to six years, this technology will make a significant contribution. Within 20 years, it can provide all of our energy needs. The discussions talk about current trends continuing for the next century as if nothing is going to change. I think global warming is real but it has been modest thus far - 1 degree f. in 100 years. It would be concern if that continued or accelerated for a long period of time, but that's not going to happen. And it's not just environmental concern that will drive this, the $2 trillion we spend on energy is providing plenty of economic incentive. I don't see any disasters occuring in the next 10 years from this. However, I AM concerned about other environment issues. There are other reasons to want to move quickly away from fossil fuels including environmental pollution at every step and the geopolitical instability it causes.
_______________________
Escondido, Calif.: Hello Ray,
Enjoyed your book, although discussing it with others often makes them uneasy, especially the idea of living much longer and acknowledging a superior intelligence (AI) here and now.
I think people put a lot of effort into thinking about getting ready to die and 'dying well', with and without the assistance of religion, and the new thought of an extrmely long life doesn't fit in that framework.
Did you collaborate with the science fiction writer of 'Counting Heads'?
Keep up the good work.
Cheers,
Neal
Ray Kurzweil: The major religions emerged in prescientific times. There is still wisdom there but we need to consider that a major motivation was rationalizing death as a good thing. After all, we had no alternative. But death is a tragedy. That is our instinctive reaction and that reaction is correct. In my view it is not death that gives life meaning. Life gives life meaning. The creation of knowledge in all its forms (art, music, science, etc.) and relationships gives life meaning. And death is disruptive of that.
_______________________
Stanwich, Ct.: Is this Ray, or just some clever AI implemetation answering posts?
Ray Kurzweil: Well, given the quality of my replies, maybe I am just a chatterbot...
_______________________
Malvern, Pa.: I have two kids, ages 9 and 11, that I care deeply about. Preparing them for life is of great importance to me. What technological tools and/or techniques have you found or created to make learning (across any or all of the realms of educational, social, physical, mental, emotional, and personal development) more fun, more effective, more efficient? Are the same or similar as effective for adult learning?
Ray Kurzweil: The world is increasingly based on knowledge. As I said, knowledge comes in many forms: art, science, technology... We are the only species that creates knowledge that we pass down from generation to generation, and the amount of knowledge is growing exponentially. Yet it still takes passion to create knowledge of value. So children need to be herded into a scientific career despite it being a scientific age. The key is to find what you have a passion for and pursue that. Having said that, it is useful for everyone to have an appreciation for technology as it is enhancing every field of human endeavor. I was at a music conference recently as a speaker and the conference catalog read like a computer catalog, so sophisticated technology is used in every field.
_______________________
Washington, D.C.: The theological ramifications of such a fusion of man and machine immediately come to mind when I consider the possibilities represented by this next step in our potential evolution, is there any reassurance for people who are believers in the human soul and or the cosmic mind, that there will not be a negative aspect for the use of such technologies? Clearly this leads to the potential transfer of the human mind to a technological platform. What happens to a transferred mind, does the soul translate, and what happens when there's a glitch, power surge or loss?
Ray Kurzweil: I attempt to address these questions in chapter 7 of Singularity is Near. There are aspects of information entities that we take for granted in the computer realm that certainly seem perplexing in the human realm. If the hardware of our PC crashes, we don't throw all the software away, we just port it to new hardware. Yet when the hardware of our mind file crashes, the software dies with it. At least that has been the case. People a hundred years from now will think it remarkable that we went around in 2006 without backing up the most precious information we have, that contained in our minds that reflects our personality, skills, experiences...We will get to a point where we can access and back up that information. As we merge with our technology, we will become a hybrid of biological and nonbiological intelligence. The nonbiological portion, however, is growing exponentially, doubling its capability each year whereas our biological intelligence is fixed. Ultimately there will be a cross over in which most of the action will be in the nonbiological portion of our civilization. We are already a biological-technological civilization and routinely do intellectual feats that would be impossible without our machines. I agree that this raises profound philosophical issues as to where the seat of consciousness resides. Some people dismiss consciousness as not a real issue because it cannot be objectively tested. After all consciousness is a synonym for subjectivity and there is a philosophical gulf between the objectivity of science and the subjectivity of conscious experience. But I do believe it is a real issue. After all our whole moral and ethical system is based on conscous experience.
_______________________
Alexandria, Va.: Whatever happened to Asimov's Laws of Robotics? With all the doom and gloom I hear from the media (namely the recent Sci-Fi Channel's Doomsday special) on how the robots will eventually "surpass" humanity, I wonder that perhaps Asimov's rules have become obsolete? That the nature of "real" AI precludes any such built in programming as it would be too limiting to the evolution of the intelligence?
Ray Kurzweil: Asimov himself showed over and over how these 3 laws could lead to unintended conseqences. Essentially, it is not sufficient to just have these 3 laws. As I mentioned earlier, we will need to put a high priority on developing defenses from the abuse of new technology. I advocated a "Manhattan" style project to develop new defenses against bioengineered biological viruses in a recent op ed piece published in the NY Times cowritten with Bill Joy. Bill Frist recently endorsed the call. An example of what I am talking about and where we have done a good job of this is the area of software viruses. Software viruses are getting more plentiful and more sophisticated, yet we have put in place an ever evolving technology immune system that protects us. The protection is of course not perfect and we will never be able to cross this concern off our list, but it has done a good job. No one has taken even a small portion of the Internet down for even a fraction of a second over the last ten years. That's a pretty good record of reliability.
_______________________
Bowie, Md.: How have visions of the future changed in the last 50 years?
I'm not quite that old; but I remember seeing what "the future" supposedly looked like in the 60's; and the present looks a lot more like the actual 60's than their vision of the future.
(Computer and other electronic technology is the big exception.)
Ray Kurzweil: There is a lot of bad futurism and the primary problem is failing to understand the acceleration of information technology. I was at a conference in 2003 on the 50th anniversary of the discovery of the structure of DNA. So all of us speakers were asked to comment on what the next 50 years will bring. Every one of the speakers except for Bill Joy and myself used the last 50 years as a model for the next 50 years. But that is demonstrably wrong. According to my models, the paradigm shift rate (roughly the rate of technical progress) is doubling every decade, so we will see about 32 times more progress in the next 50 years than the last. Watson himself said in 50 years we'll have drugs that enable you to eat as much as you want and remain slim. I said, Jim, we've already demonstrated that in animals, it will be closer to 5 to 8 years, not 50. All of the predictions done this way were far too tame. The actual power of the information technologies in price-performance and capacity doubles every year, so we'll multply their capability by a billion in 30 years (it's actually 25 due to a second level of exponential growth). But this only applies to information technologies. However, each industry ultimately converts from a pre information era to a post information era. Biology and medicine is undergoing that transition right now. It used to be hit or miss but is now becoming an information technology. We'll even do that with energy as I discussed when we get develop nanoengineered solar panels that can capture enough sunlight.
_______________________
Cupertino, Calif.: I'm left feeling you sidestepped my question about how others can compete once they are faced with competitors (the likes of Google) that control a strong AI. Without getting into debates on what is or is not strong, or whether that is well defined, do you have any thoughts on how the world will change once certain entities have strong AI, and others do not? Even if strong AI eventually becomes widespread, what about the interim period? My question is how will the business world cope with a much wider gap between the information haves, and the information have-nots?
Ray Kurzweil: At different levels, the gap between haves and have nots is diminishing not increasing. From the user perspective, technology starts out unaffordable (only the rich can afford it) but at this stages it doesn't work. Then it becomes merely expensive and works a little bit. Then it works quite well and is very inexpensive, ultimately open source and free. Today this is about a ten year progression, but within ten years will be a five year progression and in twenty years will be 2 or 3 year progression. This applies not just to electronics and software to but to any information technology such as drugs. For example, AIDS drugs were $30K per patient per year 15 years ago and didn't work very well. Now they are $100 per patient per year in Sub Saharan Africa and work pretty well. From the developer perspective, the tools of creation are not controlled only by large organizations. Google itself is a good example of this. A couple of kids doing a project in their dorm room created this new search engine technology that is now worth $100 billion and is threatening Microsoft. But a couple of kids in an American college or in India or China or Africa can and will do the same thing as we move forward, and the pace of these disruptive changes will accelerate. There is a lot of information available to everyone, not just to Google.
_______________________
Odessa, Minn.: Do you envision a future world populated by groups of humans, cyborgs, androids, and artificial intelligences?
Ray Kurzweil: Yes.
_______________________
Herndon, Va.: Hi, Mr. Kurzweil.
I came across your name while working on some research into robotics, AI, and the future of the human condition.
You are probably familiar with the sociable robot Kismet, and with robots such as Qrio. The future of humans notwithstanding, how soon do you think we'll be seeing robots capable of learning so well that they become like the fictional Bicentennial Man?
Do you believe, like Andrew Martin or Data, that robots will really strive to be (more) human? How will global society accept these robots, when the robots learn so well that they may be indistinguishable from a biological person?
Thank you.
Ray Kurzweil: Computers and robots today still miss the power of human pattern recognition (which is the heart of human intelligence), but computers are improving in this area at a steady pace. This is my own area of interest. Robots will continue to develop to fill many economic niches. The great majority will be utilitarian and not built for their personalities, from robots like Roomba (2 million and counting) to the millions of robots that assemble products in factories (and that are becoming increasingly flexible and able to operate now in unstructured environments). We do however desire to build human like robots with which we can communicate and have relationships. These range from characters in video games (a form of virtual reality) to the kinds of experiments you mention. They won't be convincing as human equivalents until computers can pass the Turing test which I've consistently pegged at 2029. At that point we will have completed the reverse engineering of the several hundred regions that comprise the human brain (we already have pretty good models and simulations about of about two dozen regions including the cerebellum where we do our skill formation). And we'll have the hardware to implemented these principles of operation of human intelligence.
_______________________
Parkville, Md.: Alright, alright... everyone moans and whines about the downside of our intelligent machines suddenly coming to realize that we're unnecessary and wiping us out. But how about the "upside?" Surely there'd be some advantages to having the planet populated by a purely robotic species?
Ray Kurzweil: Well, we are creating increasingly intelligent machines of all kinds because of their benefits to human society. They are not a race apart but are deeply integrated into out human-machine civilization. As one example of a great many, AI algorithms search for and help design rationally designed drugs that overcome disease. A good example of this is Pfizer's Torcetrapib that very selectively turns off a single enzyme which encourages atherosclerosis. The phase 2 trials showed this drug was very effective in stopping atherosclerosis (which causes about 90% of heart attacks) and they are spending a record $1 billion on the phase 3 trials. So yes intelligent machines of enormous value in helping us to to solve problems. That is why they are being created. As I mentioned before, however, we need to keep the downsides of technology in mind both through ethical guidelines on their construction and development of technological immune systems (such as we have successfully done with software viruses).
_______________________
Silver Spring, Md.: Do you believe in free will? Obviously, technology will forever be deterministic, no matter how complicated of a deterministic system we come up with to try and make it LOOK stochastic. If human beings truely have free will, then it is clearly impossible to create TRUE artificial intelligence that models that of a human being. Many philosophers believe that the human mind is indeed also nothing but a very very complex deterministic system, and therefore AI is actually feasible (After all, no other system on earth is truly stochastic - only so complicated as to make it easier for analytical purposes to model it as such - so why should the human mind be any different?). How do you get around the problem of free will in predicting that true AI systems are actually feasible?
Ray Kurzweil: THe possibility of AI is not dependent on the issue of determinism. We know that the human brain is comprised of about 100 billion neurons and while complicated, a neuron can be described as a machine. Whether it or any other machine is fundamentally deterministic depends on your interpretation of quantum mechanics. Some interpretations say there is fundamentally a lack of predictability in events at the quantum level. Wolfram showed how a deterministic system can be very unpredictable at a fundamental level. If I describe a cellular automata, the trillionth iteration is fully determined but you cannot actually determine it without running the cellular automaton for a trillion iterations. He showed that there is no short cut. So there is no short cut to predicting biological systems other than to let them make their decisions even if at some fundamental level the processes are deterministic. There are a series of interesting philosophical issues here.
_______________________
Washington, D.C.: Please weigh in on a debate my friends and I had.
Which will come first? A replacement robotic arm that can replicate everything a human arm can do down to touch or the ability to replace a human arm with one grown from your own DNA?
Ray Kurzweil: Interesting question. Both are developing separately. There were recent findings on why we are unable to regrow our limbs unlike some other species that do this routinely. There have been interesting demonstrations of giving animal species the ability to regrow certain limbs that they generally are not able to do. So this will happen. Meanwhile replacement robotic arms exist although not presently at human levels. But they get more sophisticated with each generation, and there is less and less time between each generation. It would be a guess as to which might come first.
_______________________
Cupertino, Calif.: I think it's too late to enter another question, but as a rejoinder to what you said about a couple kids in India being able to match Google, I think Google is a moving target, and you have to factor in the training time for the AI, not just the time to put together the hardware. I'm not sure how the kids in Detroit or India are going to get access to all our email, or the entire web corpus that Google has, and match its understanding (not just storage, but understanding) of that data in a short time. But maybe as you say the acceleration in information technology will take care of this as well.
Ray Kurzweil: I do agree that Google has a lot of momentum. But there are competitors to Google that are pretty close and that will help to assure that the information provided by these vast data bases remains available. For example, people have created systems that can tell the difference bewteen a dog and a cat using the millions of pictures of dogs and cats that you can now get from google. So people can use google's data bases to create their own intelligent systems. Now if google were the only game in town, one might say that they could abuse that position, but as I say there are competitors.
_______________________
Covington, La.: Ray,
I appreciate that you answered that question, as it is one I am very interested in. The one I asked was a pretty good one too! Although the form of the brain may ultimately simple, its functioning is still on a tiny, intricate, and massively parallel scale. How can we accurately predict the timescale and form of intelligent nanobots on the level of human neural interfacing?
-JV
Ray Kurzweil: That would take a book to reply to. I did attempt to write that book. I base my predictions on models of the advance of information technology in these different areas of computation, brain modeling, and so on. We need to determine how much computation is needed to simulate all regions of the brain. What is the complexity of the models needed, and so on. I have been making reasonably accurate predictions based on these models of information technology for 25 years. My first book, The Age of intelligent Machines written about 20 years ago had hundreds of predictions about the 1990s and early 2000 years that have tracked quite well. We might wonder how it is that these information technology measures track so predictably when each project is unpredictable. We see similar results in other areas of science. For example, thermodynamics. each particle's parth is unpredictable but the overall gas made of many unpredictable parties, has very predictable properties.
_______________________
Ray Kurzweil: Thanks for the opportunity to share ideas - I thought the questions were very engaging and i wish we had more time....
