The Coming Merger of Human and Machine
Posted by Jeriaska on August 1st, 2007
Ray Kurzweil delivering his keynote presentation at Transvision 2007
Ray Kurzweil is an inventor, entrepreneur, author, and futurist. Called “the restless genius” by the Wall Street Journal and “the ultimate thinking machine” by Forbes. Sun Microsystems Chief Scientist Bill Joy, whose own discussions of the promise and peril of technology have attracted worldwide attention, writes in his now famous Wired magazine cover story that “I can date the onset of my unease to the day I met Ray Kurzweil, the deservedly famous inventor of the first reading machine for the blind and many other amazing things.”
On July 26, 2007, he presented at the Transvision conference the keynote lecture entitled “The Coming Merger of Human and Machine,” in which he outlined the foreseeable implications of accelerating technological progress. At the reception following the event taking place at the Field Museum in Chicago, Illinois, the technology trends researcher, inventor, and writer was presented the HG Wells Award for Outstanding Transhumanist Contributions.
Ray Kurzweil delivering his keynote presentation at Transvision 2007
The following transcript of Ray Kurzweil’s July 26, 2007 Transvision presentation “The Coming Merger of Human and Machine” has not been approved by the author.
“We would systematically go through 10,000 compounds to find something, but invariably these drugs, which describe the majority of the market today had all sorts of side effects. The new paradigm is called rational drug design, it’s a term that’s been around for a long time but we now actually have the tools to do it. It involves actually understanding, modeling, simulating, and reprogramming biology as the set of information processes that it fundamentally is.”
“There is not a single system in our bodies that is not being redesigned, and the power of these technologies is going to grow a billion-fold over the next 25 years.”
The Coming Merger of Human and Machine
So, I speak to a lot of different groups, and lots of different industries at a lot of different technical levels, and I try to get people up to a threshold of at least understanding accelerating change and what that means, so I’ve been looking forward to this talk today because that’s a common assumption and we can try to talk about some implications.
So, I was talking to Philip Rosedale at breakfast about a key question that I’ve been wondering about, which is why some people readily grasp very quickly the notion of accelerating change and its implications and some people are very resistant to the idea. And it’s not a question of technical level or intelligence. There are some brilliant people in computer science who just don’t get it, or kind of get it, but then they really resist appreciating and understanding the implications. So, one could hypothesize that the idea was attacking some of their coping mechanisms or their basic fundamental philosophies.
Well, a child quickly learns two very powerful words. One is ‘no’ and the other is ‘why.’ And people when confronted by the subject of accelerating change ask the question ‘Why?’ quite a bit. You know, a child asks, ‘Why is the sky blue?’ and the answer is that there are particles in the air that change its color. ‘Why?’ So, you can keep asking why and keep the conversation going quite intelligently. So, why do people resist the implications of accelerating change? Why does it attack their fundamental beliefs? One of those fundamental beliefs is rationalizing death. Probably the bulk of literature deals with this fundamental human fear, and one way of dealing with it is to rationalize it as a good thing. But in talking about radical life extension you are attacking that coping mechanism.
But the key implication that I see from technological innovation is exponential growth. We are doubling the price performance and bandwidth of information technologies in many different areas every year. Linear growth and exponential growth look the same for a short period of time, but if you go out 20, 30, 40 steps, it’s radically different. So, a lot of my predictions are considered radical. They end up being conservative by design, very often. But the implications of exponential growth are really profound. You know, when I was at MIT, a computer took up not the size of this room, that would be an overstatement, but easily two of these sections. It cost $1 million and was the MIT computer. It weighed many tons and all of us thousands of students were supposed to share it. That computer is thousands of times less powerful than your cell phone today.
Think about the things that you can do today. You can put a computer in the brain to replace a portion of the brain destroyed by Parkinson’s. And the latest generation of this FDA approved neural implant can download new software to this pea-sized computer inside the brain from outside the patient. Someday we might have nanobots in our bloodstream that can keep us healthy from inside. To many people that sounds very futuristic, but there are currently four major conferences on Bio MEMS, dozens of experiments in animals. At MIT there is a device that scouts out cancer cells, burrows inside and destroys the cell. We’re gaining means of reprogramming our biology.
Biology did not used to be an information technology. We really did not understand biology as a set of information processes. It was really hit or miss. We would just happen to find something and automated that drug development process in what is called drug discovery. So we would systematically go through 10,000 compounds to find something, but invariably these drugs, which describe the majority of the market today had all sorts of side effects. The new paradigm is called rational drug design, it’s a term that’s been around for a long time but we now actually have the tools to do it. It involves actually understanding, modeling, simulating, and reprogramming biology as the set of information processes that it fundamentally is.
We have outdated software running in our bodies. A gene is a linear sequence of data–it interacts in 3-dimensional ways, but it’s a software program. And they evolved when times were very different. It was in the interest of the species to hold on to every calorie. The fat insulin receptor gene says to do that. That was the innovation that allowed our ancestors to roam around. That evolved in an environment of scarcity. But now we have an epidemic of obesity. There is a new technology, RNA interference, that can identify a gene and turn it off. At the Joslin Diabetes Center they turned off the fat insulin receptor gene in mice. These animals ate ravenously and remained slim. They got the health benefits of a caloric restriction diet, living 20% longer, and there are five pharmaceutical companies rushing to bring fat insulin receptor inhibition drugs to the human market.
It reminds me of just how few people really appreciate the implications of exponential growth. A few years ago I was at this conference that TIME Magazine organized on the 50th anniversary of the discovery of the structure of DNA. And every speaker there, except for Bill Joy and myself, used the last 50 years as a model for the next 50 years. James D. Watson himself said that in 50 years we will have drugs that will allow us to eat as much as we want and remain slim. We already have already demonstrated that in mammals. We will see that in one decade, most of that will be taken up by the regulatory process. It will be one decade, not five. All the predictions there were overly conservative.
It is very clear that the rate of progress is accelerating. As Bill Joy mentioned, we didn’t notice that there was change at all, and then it became an axiom that change is constant. Now, we should take into consideration that change is not constant, but accelerating. Let me read you a passage from my book The Singularity is Near.
[Photographs a page from his book with the Kurzweil-National Federation of the Blind Reader]
The Kurzweil-National Federation of the Blind Reader, a digital camera equipped with character recognition software and text-to-speech conversion technology
“The AI winter is long since over. We are well into the spring of narrow AI. Most of the examples above were research projects just ten to fifteen years ago. If all the AI systems in the world suddenly stopped functioning, our economic infrastructure would grind to a halt. Your bank would cease doing business. Most transportation would be crippled. Most communications would fail. This was not the case a decade ago. Of course, our AI systems are not smart enough—yet—to organize such a conspiracy. If you understand something in only one way, then you don’t really understand it at all. This is because, if something goes wrong, you get stuck with a thought that just sits in your mind with nowhere to go. The secret of what anything means to us depends on how we’ve connected it to all the other things we know.”
That’s a good idea of why I came up with these models of technology evolution. I decided I wanted to be an inventor when I was five. I quickly realized that the key to being successful was timing. Another key was actually not to be dismayed by failure, and not thinking of experimentation as failure. Nobody remembers the thousands of filaments that Edison tried in inventing the light bulb. He easily could have given up, and was strongly encouraged to give up after trying 500 different filaments, none of which came close to working. Part of my advice to entrepreneurs and inventors is that you define success. And if you commit to a project, see it through. Persistence generally pays off in the end.
But another key is timing. Most technologies fail if the timing is executed incorrectly. Realizing this, I became an ardent student of technology trends. Four years ago I was having a conversation with the head of the National Federation of the Blind, who had worked with me on the first reading machine, which was about as big as that podium. He asked me when a handheld device would be feasible. And I said that according to our models the digital camera and pocket computer technology would be available in four years, the spring of 2006. So, we started in 2002 and introduced the product in the summer of 2006. There’s now a thousand guys and gals going around taking pictures of labels in their clothing, menus. Other companies have noticed that this is feasible and they have started working on it themselves.
So, inventing is like surfing. You have to catch the wave. Anticipating timing is very key. That is why I started tracking technology trends. We do not yet have the inventions of 2015 or 2025, but we can predict them and talk about them at conferences like this one. Commenting on the passage I just read, a lot of people ask, “Whatever happened to artificial intelligence?” It reminds me of people who go into the rainforest and ask where are all the species that are supposed to be here, when there are 25 species of ants within 50 feet of them. But you don’t see them. They’re hidden in the infrastructure. And the same thing is true of AI.
There’s hundreds of examples of AI, narrow AI, but programs performing functions that people used to perform, performing them at generally better levels detecting credit card fraud, diagnosing electrocardiograms, landing and flying airplanes, intelligent weapons systems… you could list hundreds of different examples. If all these programs were to stop, our economic infrastructure would grind to a halt. That was not the case even 25 years ago. Then, these were research projects. The narrowness of these applications is gradually getting less narrow. And I will talk about how exponential growth in both computer hardware and understanding in how the brain produces intelligence, the human software, is growing exponentially. It is not that far away when we will understand the full range of human intelligence.
So let me quickly go through these graphs. I mentioned the paradigm shift is doubling every decade. These are logarithmic graphs. As you go up the graph, each level is ten times greater than the one below it. A straight line represents exponential growth. Telephones in fifty years were adopted by a quarter of the human population. About half the population of the world now has a cell phone. These early communication technologies–telegraph, radio, television–took decades to be adopted. We see technologies today take just a few years’ time. And if you think back five years ago, most people did not use search engines, there were no social networks and blogs. It seems like ancient history. Things like Second Life emerged very recently. So, the pace of change continues to accelerate.
A billion years ago, not much happened in a million years. A thousand years ago, not much happened in a hundred years. And the cutting edge of our evolution is now our technological evolution. The key is: if you have a process that can store information and replicate itself, you have the makings of an evolutionary process. Somehow we have a universe, and Jim Garner can talk about these issues, that is able to encode information in an atomic structures. How did we get that? We have a standard model of physics that has all sorts of delicate constants and physical laws that allows for the evolution of information in atomic structures. If some other constants were just a little bit different, there would be no atoms and no stars. Maybe there was an intelligent designer who designed our universe at a science fair and we’ll see what kind of grade he gets. It depends on how we deal with some of these existential risks.
That evolutionary process created DNA, where evolutionary information could be encoded in the information backbone of biology. And that evolutionary process led to brains that could hold information. Brains allowed for the control of an opposable appendage. Chimpanzees’ thumbs look similar but the pivot point is down one inch. Evolution didn’t finish the job. If you watch chimps, they are very clumsy. They don’t have a power grip, they don’t have fine motor coordination.
The fifth epoch of human evolution has just started. That is the merger of biology with technology. It involves creating technology which has all the secrets of biology. Reverse engineering biology, whether it’s through biotechnology or our brains, is really a great revolution. It is true that specific paradigms do run out of steam. Moore’s Law was not the first but the fifth paradigm for describing exponential growth in the price performance of computer power, and we will go onto the sixth paradigm, which is 3-dimensional molecular computing. When a particular paradigm runs out of steam, it creates research pressure to make the next. And that continues the exponential logarithmic growth. There are limits to the physical capacity supporting computation, but they’re not very limiting. One cubic inch of nanotube circuitry would be 100 million times more powerful than the most conservative estimate of the amount of computation we need to simulate all several hundred regions of the human brain. And nanotube circuitry is not the only limit either.
There has been one hundred twenty years of smooth, predictable exponential growth in computing, though the 20th century had all kinds of history–two world wars, the Cold War, the great depression, stock market booms and busts. You don’t see any evidence of all that human history on this graph. Critics say you can’t predict the future, and that’s true of specific projects. But the overall impact is very predictable. You can use these curves and project them out well into the 21st century. It’s a double logarithmic growth curve. It took three years for the price performance of computers to double in 1900, two years in 1950, and one year today. In some cases I’ve been rather conservative. The Singularity is Near, which came out two years ago, predicted that in 2015 supercomputers would hit the threshold of the most conservative estimate of the amount of computation needed for simulating the brain: 10^16 cps. There are now two projects going on in IBM and Japan that are estimated to hit that threshold in 2010.
You see this in example after example of information technology. The progression is very smooth. As technologies are smaller they’re better and faster, because the electrons have less distance to travel. While most of the economy is not yet an information economy, we’re getting there. The economy is not shrinking. Growth rates have been increasing as information technology has accounted for a greater percentage of the economy. We’re seeing smooth growth in biotechnology. It cost $10 to sequence each base pair in 1990 and it’s a fraction of a penny today. These smooth exponential growth curves are not a planned project. This is just the result of this self-organizing chaotic process involving thousands of companies and tens of thousands of people. There is not a single system in our bodies that is not being redesigned, and the power of these technologies is going to grow a billion-fold over the next 25 years.
If you talk to Intel scientists, the cross-over point in terms of 3-dimensional circuits is going to be the teen years. This will keep exponential growth going well into the 21st century. The key issue really is the software. Some people say that software is stuck in the mud. That’s really not the case. Software is getting more sophisticated and more complex. I invite critics to use a computer from 15 years ago and try to get some work done. We forget just how much things have progressed. 15 years ago there was hardly any AI in our economic infrastructure. Reverse engineering the human brain will expand the AI toolkit. We’re not currently learning anything about machine vision from the human brain and we don’t really have good models yet of how the visual cortex works, but I think we’re going to get there soon. We had the same experience a number of years ago in speech recognition. We are doubling the spacial resolution of brain scanning every year, and we are showing that we can turn this information into working models and simulations.
It turns out the brain is neither very simple nor is it so vastly complex that you need to model the position of every ion channel in order to understand it. There is a manageable amount of complexity. In fact, we can already measure that. The genome holds the design of the human body and the brain. Without compression and including all the so-called junk DNA, all of it is 800 million bytes. About half of it codes for the body and half for the brain. But it’s actually replete with redundancies. The ALU sequence is repeated 800,000 times. The pancreatic islet cells follow a very simple algorithm just a little bit more complicated than a thermostat. That’s the salient detail: you don’t have to get lost in all the tremendous amount of complexity.
Brain regions are a little more intricate than that, and a similar principle applies. There is about an equal amount of complexity in the brain and in the body. About 25 million bytes describes the design of the human brain. If you measure the amount of data in the brain, it’s billions of times greater than that. That’s because the genome generates itself fractall, creating more and more information as it evolves into a real brain. For example, take the cerebellum: there’s only a few genes, only tens of thousands of bytes of design information in the genome that describes the wiring of the cerebellum. But if you actually measured all the data in that wiring, it’s trillions of bytes. I’ll summarize what the genome says. There are four different neuron types in the cerebellum. They’re organized something like this. Now repeat that pattern 10 billion times, and it’s a random variation within certain constraints. There is a great deal of randomness within its initial configuration, and then a child grows up to learn how to walk and talk and catch a fly ball. The cerebellum gets filled up with meaningful information.
I make the case in chapter 4 of the book that we will have models and simulations of all several hundred regions of the brain within twenty years. And that may be a conservative estimate. My view and the consensus view of the AI community are getting closer together, and it’s not because I am changing my position. My position for human level artificial intelligence has been 2029. I think that’s slightly conservative. It coincides with this project of reverse engineering the human brain. The consensus view when we had The Age of Spiritual Machines conference at Stanford was that it would take several centuries. And at AI 50, which was last year, the consensus among the AI leaders was fifty years. So, I’m saying 25 years. I know that some people at this conference think it’s sooner than that.
Reverse engineering the human brain will be a very important project because it will give us greater insight into ourselves. All of this is really fueling economic growth. The non-information technologies are shrinking. And you see this all across the world. The World Bank is reporting that poverty has been cut by half in Asia over the past ten years because of information technologies, and it will be cut by 90% over the next decade. The adoption of e-commerce is growing exponentially. It’s now well over $1 trillion. Were the internet a country it would soon have the largest gross national product in the world, but it is not limited by national boundaries. People sometimes say, “Are we going to allow transhumanism and artificial intelligence to occur?” Well, I don’t recall when we voted that there would be an internet. It happened bit by bit, node by node.
There has been a vast movement for the democratization of the tools of creativity. Without any government being involved, we have an economy that transcends national boundaries. Tiny steps are made on thousands of fronts that move us forward. The same is going to be true of radical life extension. You don’t see the boom and bust of the dot com era, because that was a Wall Street phenomenon. Wall Street got very excited about the internet, believing it would turn every business model on its head, and that was accurate, but they lost patience in 2000. Meanwhile, the internet was growing rapidly, undaunted by either the boom or the bust. If you look at the big picture, we went from 16,000 college students in 1970 to 6 million today. We spend ten times as much in constant dollars per capita on K through 12 to keep up with the rising skill level.
Just a completely random aside, I really don’t like a lot of the terms that we use, but we’re stuck with them. ‘Artificial intelligence’ implies that it’s fake intelligence. It does not have the full range of human intelligence, but my thesis is that will end when we really understand how the brain performs intelligence. Or take ‘virtual reality.’ If I say I was virtually on time, that means I was not on time. Again, that is currently deserved. Second Life is not as real as real reality, but you can form a community with people around the world based on shared interests. You can change who you are, which is a big advantage of virtual reality. I give about a quarter of my presentations virtually, and it’s not video conferencing. I appear three-dimensional at the venue, and life sized, and I can establish eye contact, which is a key thing in virtual reality. This will be ubiquitous technology in less than a decade.
‘Transhuman’ means transcending human limitations. It’s leaving out the word ‘limitations.’ But how do we define what is human? Is it based on our limitations? That is the kind of idea that fuels fundamentalism. In my view, humans are of the species that goes beyond our limitations. We thought that we were descended from the gods, and really we were descended from the worms. But really, there is something unique about our species. We are the species that seeks to go beyond our limitations. Human life expectancy was 25 a thousand years ago.
Early the next decade computers will disappear. They won’t be these physical objects, they will make their way into our bodies and brains. We will put them in our eyeglasses to create a virtual screen that hovers in the air and detects your head motions and keeps the image constant as you move your head. Augmented reality will be like a search engine that doesn’t wait to be asked. It will be following your conversation and help you out as soon as you get stuck.
We will be multiplying the power of these technologies almost a billionfold and computers will pass the Turing test by 2029. I’m currently making a movie called The Singularity is Near: A True Story About the Future, where I interview twenty big thinkers like Marvin Minsky, Eric Drexler, and Richard Clark about their ideas about the future. There will also be a B line narrative story that illustrates some of the ideas. The story of Ramona starts at the 2001 TED conference: she’s my female alter-ego. She goes into the future and becomes more and more lifelike, human, and independent, kind of like the Pinocchio story. She hires Al Durscwitz to represent her and press for her legal rights as a person. He’s going to play himself. She gets coaching from Tony Robbins to become more human. It’s a story that illustrates the ideas. I bet on the Long Now website that by 2029 computers will pass the Turing test. And it will take the human powers of pattern recognition, which when combined with computer intelligence will be a powerful combination.
I’ve asked a number of Parkinson’s patients who have neural implants if they feel it is part of who they are. And all of them have said yes. Same thing for deaf persons with a cochlear implant. And these are the very early stages. There will be thousands of technologies like this. Some will be quite conservative and you will be hard pressed to find people who don’t use them. We have people in our society today, like the Amish, who try to eschew technology. But most people use quite a bit of technology. There will not be one thing but thousands of technologies that span our experience and intelligence in many different ways. Life expectancy was 37 in 1800. That’s not that long ago. There was no knowledge of the germ theory of disease or use of sanitation. So life was short, brutish, disaster prone, as Thomas Hobbes described it. Bill McKibben wrote a book Enough, which has lines in it that sound like we have enough advanced technology already. But I had a dialog with him in my book and I suggested that if he was really serious about stopping global warming then we should go full steam ahead with nanotechnology, because we are close to creating nanoengineered solar panels that will be cheaper than fossil fuels. That’s not that far away. And we have 10,000 times more sunlight than we need. We only need to capture one part in 10,000 for all our global energy needs. But for that we need nanotechnology, because the silicon-based panels are too expensive, too heavy, too hard to integrate with buildings. Bill Joy heard this and said, “Go for it.”
That’s one reason why these technologies will move ahead. Only technology, particularly GNR (genetics, nanotechnology, and robotics technology) has the scale to overcome poverty, environmental effects, and disease. This progression is going to go into high gear. It’s going to go much more rapidly when we get to the mature phase of this biotechnology revolution. According to our models, you will be adding more than a year every year to your remaining life expectancy within fifteen years. It’s not a guarantee of immortality, and Aubrey de Grey makes this point very eloquently, because he is really the chief theorist of this second bridge–the first bridge being the two hundred supplements I take a day. And the goal of that is really just to keep me healthy for another ten or fifteen years, until we get the full flowering of the biotechnology revolution. That will involve the reprogramming of our biology and lead to the third bridge–nanotechnology.
Critics ask, “Do you really want Windows running inside your body?” The insinuation is that software is too crash prone to rely on. Well, we do know how to build reliable systems: 911 call centers, intensive care units. The internet, which is decentralized, is highly stable, despite the attempts of all these software viruses. That was the solution that biology found, by building a highly decentralized system. There is no Chief Executive Officer neuron. Neurons are dying and decaying all the time, but the overall ability to self-organize information remains intact. The internet is actually a very good example of that.
As for promise versus peril, we need to put a higher priority on defending ourselves. Bill Joy and I wrote this op-ed piece called “A Recipe for Destruction” criticizing the U.S. government for posting the 1918 flu genome on the web. That’s something we shouldn’t be doing. This is actually more dangerous than putting the designs for an atom bomb on the web. Another issue, and we call for that in this article, is a rapid response system. The ability to take a biological virus and modify it to be more deadly, stealthy, and communicable is with us today.
The Op-Ed piece “Recipe for Destruction” by Ray Kurzweil and Bill Joy was printed in the New York Times on October 17, 2005
After 9-11, we know that people can be purposefully destructive. The good news is that we have the technologies to protect us. We can put a rapid response system in place. A key step is genetic sequencing. It took us five years to sequence HIV, we could sequence SARS in 21 days, we can sequence a virus now in one or two days. If a natural virus emerges, like a mutated bird flu, or an unnatural one, like a bioterrorist weapon, we could sequence it in a day or two and create an RNA interference medication very rapidly if we set up a system to do that. A good model for a rapid response system is a software virus protection system. If you put out a software virus tomorrow, it will be detected , reverse engineered, an antidote will be coded which will be spread virally on the internet. We need to put a system like that in place for biological viruses.
As I mentioned, we are unique after all, in spite of the ways we thought we were unique and weren’t. We are the only species that creates knowledge that expands exponentially. We pass that from generation to generation, and it includes all of our music, art, science and technology. People say death gives life meaning. In my view, it is life and our ability to create and share knowledge that gives life meaning. That makes death a tragedy. We should embrace the ability to go beyond our limitations. Thank you very much.
Related article: 7.25.07 The Power of Intelligence: Eliezer Yudkowsky describes how improved cognitive technologies will have a dramatic impact on human conception and control over the physical universe.