Ray Kurzweil has received the National Medal of Technology, was inducted into the National Inventors Hall of Fame, and has been honored by three U.S. presidents. For the February 21 keynote presentation of the 2008 Game Developers Conference in San Francisco, he gave a talk entitled “The Next 20 Years of Gaming” where he discussed the foreseeable ramifications of the accelerating price performance growth of information technologies such as those found in the videogame industry.
The following transcript of Ray Kurzweil’s keynote speech at the 2008 Game Developers Conference entitled “The Next 20 Years of Gaming” has not been approved by the speaker. Slides are from the Singularity Summit at Stanford presentation “The Singularity: A Hard or Soft Take-off?” Corrections to the transcript have been provided by Richard Karpinski.
The Next 20 Years of Gaming
Let me try to provide some perspective of the next several decades. To do that we need to look at the last several decades. I have been studying technology trends for thirty years. I did that because my basic interest is being an inventor. I decided I was going to become an inventor when I was five. I’m not exactly sure why that is. I’m trying to reconstruct that now–maybe a few more years of analysis will reveal why I made that decision–but I remember other kids talking about how they were going to become a nurse, a fireman… and I knew what I wanted to be.I quickly caught on that the key to being successful as an inventor was timing. I became an ardent student of technology trends. Being an engineer, I gathered a lot of data, and this has taken on a life of its own. I have a team now of ten people that helps me gather data in many different fields, and we make mathematical models of how technology evolves. With these models now, we can project in advance the technologies of five years from now, ten years from now, twenty years from now. We don’t have the computer circuits of 2020 today, but we can imagine what they will be capable of doing.
You might say, “Now, wait a second, the future is unpredictable.” That’s the common wisdom. That is true, for specific projects. Is Microsoft’s acquisition of Yahoo going to be good or bad? That’s hard to predict. Will Google stock be higher or lower than it is today three years from now? That’s hard to predict. What’s going to be the number one game a year from now? That’s hard to predict. What will the next wireless standard be? That’s hard to predict. But if you ask me, “What will the cost of sequencing a base pair of DNA in 2012?” or the spacial resolution of a 3D volume in brain scanning in 2014, I can give you a figure, and it is very likely to be correct. I have been not just looking back at past data, but making these forward predictions for 25 years.
I wrote my first book in the 1980s. The Age of Intelligent Machines had predictions about the 1990′s and early 2000′s, which have tracked very well. For instance, I saw the ARPANET doubling every year. It went from 10,000 nodes to 20,000 nodes, and felt for various reasons that this would continue doubling every year, multiplied by a thousand in ten years. This would be going to 150 million in the mid-1990′s, and I predicted a World Wide Web connecting hundreds of millions of people around the world. That seemed ridiculous when with great effort we could sign up another thousand scientists on the ARPANET, but it came right on schedule.
I saw the chess supercomputers doubling in power every year. That put the cross-over at 1998. I predicted that in the early ’80s when the average player could beat the best chess machine. Kasparov said in ’93 that he had played the best chess machines and that they were predictable, brittle, and pathetic. That was a reasonable perspective in ’93, but they soared past him in ’97. There are many other examples of this.
This is a very democratizing technology. I predicted in the ’80s that the Soviet Union would be swept away by this decentralized electronic communication, which at that time was FAX machines and teletype machines. In that coup against Gorbachev in ’91, the photo-op was Yeltsin bravely standing on top of a tank, but tanks had nothing to do with it. It was this network of communications that swept away totalitarian control. The old paradigm of the powers grabbing the central TV and radio station was swept away.
It is also democratizing the tools of creativity. Creating a new massively multiplayer game played around the world can be done on a thousand-dollar laptop. Kids in their dorm room can create a high definition movie with their notebook computer and a $500 HD camera. A couple of kids at Stanford using their thousand-dollar laptops created a little bit of software that revolutionized web search and extends our knowledge. The company they started is now worth $160 billion.
The tools of creativity have been democratized. Ultimately the tools of production, when we have full-scale nanotechnology that can create physical products from massively parallel information processes, will also be democratized.
This is quite a revolutionary phenomenon, so when we talk about just how predictable this phenomenon is, it is ultimately explosive. Doubling every year means multiplying by a thousand in ten years, a million in twenty years, a billion in thirty years–the rate of exponential growth is slowly growing, so it’s actually twenty-five years.
When I came to MIT in 1965, it was so advanced that it actually had its own computer, unlike most of the other schools I considered. It took up half a building, cost $11 million, and was shared by college students and professors who needed influence to get close to it. The computer in your cell phone today is a million times smaller, a million times cheaper, and a thousand times more powerful. That’s a billionfold increase in the price performance that we have seen in computation in the last forty years.
As influential as information technology is today, and you can certainly feel this in the game industry, we will see another billionfold increase in capability for the same cost in the next 25 years because of the slow rate of increase in the rate of exponential growth. We are also shrinking the size of these technologies. Today you can put a pea-size computer inside your brain to replace biological neurons destroyed by a disease. They actually interface quite well with your other neurons, and the newest generation allows you to download new software to the computer inside your brain from outside the patient.
Take what we can do today, apply this billionfold increase in price performance capability, a hundred thousandfold shrinking, and what you will get are blood cell sized devices that are even more capable and can go inside our bodies and keep us healthy, expand our intelligence and provide full-immersion virtual reality from within the nervous system. I’m getting a little bit ahead of myself, but the point I want to make is–it’s predictable. I’m going to show you a few examples.
We have hundreds of these that show this exquisitely smooth exponential progression. This goes back 110 years, to the 1890 census, despite all the history that we saw in the 20th century .The other factor is it’s not just computers and communication devices. It also affects everything that we care about, like for example our health, and biology and medicine–that did not used to be an information technology. We would try to find something that has some benefit. Invariably these drugs have lots of side effects. 99% of the drugs on the market today are done that way. We did not even have the software that biology is based on.
Health and medicine is now, unlike just a few years ago, an information technology. We have the software that life runs on, as of only four years ago, and we have the tools to actually reprogram it. I’ll give you an example. There is one gene called the fat insulin receptor gene. That gene basically says, “Hold onto every calorie because the next hunting season may not work out so well.” That was a good idea ten thousand years ago. You worked all day to get calories, there were no refrigerators, so you stored them in the fat cells of your body. In fact, it’s an ancient gene that allows animals to roam around–plants don’t have a fat insulin receptor gene.
That is now the factor underlying obesity. We live in an era of abundance, where obesity underlies the epidemic of diabetes and heart disease. What would happen if we turned that gene off? Well, we have a technology to turn genes off called RNA interference. This was tried at the Joslin Diabetes Center near where I work, and these animals ate ravenously and remained slim. It wasn’t a fake slimness; they got all the health benefits of being slim. They didn’t get diabetes or heart disease. They lived 20% longer and got the benefits of caloric restriction while doing the opposite, and there are four pharmaceutical companies rushing to bring patents to fat insulin gene receptor inhibitors to the human market. This is actually what we want in terms of a diet drug. All the diet drugs on the market today work by inhibiting appetite. It’s kind of like a birth control pill that inhibits interest in sex. So there are several patents on insulin gene receptor inhibitors in the development pipeline.
This brings up another point. Exponential expansion of information technology is accelerating the pace of change. A few years ago I was at a conference that Time Magazine organized on the 50th anniversary of the discovery of the structure of DNA. All of the speakers were asked to comment on what will the next fifty years bring. All the other speakers except for Bill Joy and myself used the last fifty years as a model for the next fifty years. Jim Watson himself, who co-discovered DNA, said that in fifty years we will have drugs that will enable you to eat as much as you want and remain slim. I said, “Jim, we’ve already done that. We’ll have that within one decade, not five.”
Exponential growth is very surprising, but people don’t think that way. Our intuition is linear. This whole debate on social security based on linear projections of longevity increase, ignoring what I have just said about how it is now an information technology. I believe that is hardwired. When we were walking through the Savanna a thousand years ago and we saw something coming towards us out of the corner of our eye, we made a linear projection of where we should be in twenty seconds, and where we probably shouldn’t be. That does not work as well projecting where technology will be.
Now, I will show you that an exponential projection is the same as a linear projection for a short period of time. Government models based on linear projections work okay for a year or two, but we are now at a point where the acceleration is so fast that things change very rapidly in just a matter of a few years, so you cannot ignore the exponential projection. If you are planning a game, or any information-based technology for two, three, four years from now, the world is going to be completely different. You cannot make a linear projection.
Think back six or seven years ago. Most people didn’t use search engines; it sounds like ancient history. Three or four years ago people didn’t use blogs, podcasts, social networks. Massively multiplayer games only became popular fairly recently. The world is changing very dramatically, very quickly, and the world will be very different in three or four years.
Let me give you one example. About 30 years ago, I developed the first print-to-speech reading machine for the blind, a washing machine-sized device. I’ve stayed involved in this field, and each year they have become smaller and smaller, and more and more capable. In the summer of 2006 we introduced a fairly cumbersome device, but you could carry it around. There are now over a thousand blind guys and gals taking pictures of menus and signs on the wall. We just recently introduced it in a cell phone. In addition to reading documents, this is also a phone, a camera, a GPS navigation system, email, internet, and mp3 player.
The display includes synchronized highlighting, which is useful for dyslexic kids. We have a lot of research showing that a synchronized auditory and visual presentation actually helps build their reading skills. So this is a good example of trying to time projects. That is certainly critical in the game industry. A lot of games take several years to develop, and the game technology will be completely different. Games are becoming the harbinger of everything that we do. I look forward to talking to this audience, because you can really feel how rapidly things change.
I would like to talk about what the next twenty years will be. First I would like to show just how pervasive are exponentials, how smooth and predictable these trends are. It’s not just computer and communication devices. It will ultimately affect everything we do. I mentioned our biology. Industries you would think are not information technologies will become information technologies. Take energy, for example. I was just on a National Academy of Engineering panel to come up with engineering plans to address a lot of these big problems. Larry Page of Google and I came up with the energy plan to apply information technology to energy through information-controlled nanotechnology. Nanoengineering is an information technology, controlling matter at the molecular level through massively parallel information processes.
Today, 85% of our energy comes from fossil fuels. That is clearly not an information technology, but a 19th century industrial technology. Solar energy is also an industrial technology today, using cumbersome solar panels that are hard to manufacture. They are heavy, inefficient, and expensive. There is already a new generation of nanoengineered solar panels coming out that are less expensive. Larry Page and I are convinced that within five years we will reach a tipping point, where nanoengineered solar panels will be less expensive per watt than energy from coal and oil. The cross-over is not that far away.
You might say, ‘Is there really enough sunlight to meet more than a tiny fraction of our energy needs?’ There is actually ten thousand times more than we need. If we can capture one part in ten thousand of the sunlight that falls on the earth we would meet 100% of our energy needs. We have a plan to do exactly that within twenty years, and store the energy in nanoengineered fuel cells, which is another form of information technology. That is applying information technology to energy, and there is already billions of dollars in venture capital behind this form of energy. A lot of the discussions about global warming talk about how in one hundred years a certain percent of our energy will come from fossil fuels, and will be depleted, and the atmosphere will be this… completely ignoring these emerging trends.
There is actually an exponential progression there. We are doubling the amount of solar energy every two years. That’s multiplying by a thousand in twenty years. We have seven more doublings to go, and it is on this doubling ramp. I believe that within twenty years we will have largely replaced fossil fuels with these forms of renewable energy. That is another example of very predictable results from information technology.
Let me show you just how pervasive this is. I mentioned that we are actually increasing the rate of progress. According to my models we are doubling the paradigm shift rate every decade now. We will see 32 times more progress in the next fifty years than we did in the last fifty years. The telephone, the first virtual reality technology, where you could actually be with someone else in a virtual environment even if you were hundreds of miles away, took fifty years to be adopted by a quarter of the U.S. population. By the way, these are logarithmic graphs–as you go up the graph you are not adding, but multiplying by powers of ten, and each level is ten times greater than the level below it. A straight line on a logarithmic graph is exponential growth. It took fifty years for the telephone to be adopted by a quarter of the U.S. population. The cell phone did that in seven years. Radio and television took decades to be adopted; the PC, the web were adopted in a few years time.
This has continued to accelerate. I have a theory as to why this is the case. It is fundamentally a theory of evolution. The reason that an evolutionary process accelerates is it develops a capability and then adopts that capability into its own methods of evolution, so in the next stage it evolves more quickly. The products of that next stage grow exponentially in capability.
I have put key paradigm shifts that involve biological and technological evolution on this logarithmic graph. On the x-axis is how many years ago this event took place, the y-axis is how long it took for that paradigm shift to be adopted. The very first paradigm shift (life, biology, DNA, an information backbone to guide this evolutionary process) took a billion years to evolve. The next stage in biological evolution is the Cambrian explosion and all the body plans of the animals evolved. That went a hundred times faster–it took ten million years.
Biological evolution kept accelerating. Homo sapiens, the first technology-creating species, evolved in only a few hundred thousand years–a blink of an eye in biological evolutionary terms. There are actually only three simple genetic changes that distinguish us from our biological predecessors, other primates, that comprise only a few tens of thousands of bytes of genetic information.
We have a larger skull to accommodate a bigger brain at the expense of a weaker jaw–so don’t get into a biting contest with another primate. More of the brain is devoted to the cerebral cortex, so we can do abstract reasoning. We can do “what if” experiments in our minds. “What if I took that stone and that stick, and tied them together with that twine? I could extend my leverage.” And then we have an opposable appendage that actually works well. It might look like a chimpanzee’s hand–it looks similar–but evolution had not quite finished the job. The pivot point of the thumb is down one inch. If you watch a chimp, they don’t have a power grip, they don’t have fine motor coordination. They can kind of create primitive tools, but they cannot create tools that are sophisticated enough to use the tools to create new tools to start this whole other evolutionary process. Our thumb worked well enough, our cerebral cortex was sophisticated enough, that the next stage in this progression was technology.
The first stages of that were a little bit faster: tens of thousands of years for fire, stone tools, the wheel. Then we always use the latest technology to create the next technology. Half a century ago the first computers in World War II were designed with pen on paper and wired with screwdrivers. It took years. Today you can create a new supercomputer in a matter of days.
You see it forms a straight line on this double-logarithmic graph, showing a continual acceleration in the pace of biological evolution, leading to technological evolution. Some people said, “Well, Kurzweil only puts points on this graph if they fit on a straight line–if I had a paradigm shift and it did not fit on the straight line, I did not bother to put it on there. So I took fifteen different lists: Carl Sagan’s Cosmic Calendar, the Museum of Natural History, the Encyclopedia Britannica, and a dozen other lists, and saw what they thought the key events were in biological and technological evolution.
There is some disagreement: Some people feel the Cambrian explosion took 20 million years and not 10 million years. Some people include the ARPANET with the internet, so that it is twenty-five years and not ten years. There is disagreement when human language started. But you can see a very clear trend line. I don’t think the internet took a million years to evolve, and nobody thinks the Cambrian explosion took ten years. Not much happened in a million years a billion years ago. There is a very clear acceleration in the evolutionary process.
I am looking forward to this new game, which I think is coming out in September, called Spore. It’s a massively parallel multiplayer game where you can actually design your own entities at different levels of evolution: either primitive life forms that will evolve and eventually climb out of the ocean onto earth, or more advanced organisms that can then compete. I think it goes all the way to my “sixth epoch,” where very advanced posthumans go off to space and colonize the universe. That has built into it this kind of acceleration.
This is a linear graph and it shows a simple exponential linear progression, and you can see how they are confused. They look the same, for a few years. In fact, exponential growth can actually be sub-linear. That leads people to be very pessimistic , because they are thinking linearly. If they saw the exponential potential of these technologies, they would have a very different perspective.
The ongoing exponential is actually made up of an ongoing series of S-curves, because paradigms do run out of steam, and it creates research pressure for the next paradigm. That is actually another objection to my projections–that Kurzweil takes these exponentials and projects them out, and we all know that exponential growth cannot go on forever. Rabbits in Australia multiply exponentially, but then that comes to a sudden stop when they eat up all the foliage. Indeed, every resource, every paradigm in information technology does the same thing. It finally gets to a point where it runs out of its resources for exponential progression, but that creates research pressure for the next paradigm.
If we look at the history of computation, and this is a very good example of just how predictable this is, going back to the 1890 census, we see very similar exponential growth for over a century, though five different paradigms. I put here 49 famous computers. Any computer you put on here is on this curve. In the lower lefthand corner is the data processing equipment that automated the first American census, the 1890 census. These were using punch card machines, which were subsequently shipped to the Florida Elections Commission.
We cracked the German Enigma code with a relay-based computer. There is a lot of interesting wartime literature about how it was used, and not used. They didn’t want to use it too much and tip off to the enemy that they had cracked the code. A lot of subterfuge was to convince the enemy they had got the information in another way. They knew a convoy of ships was coming; they sent over a low flier and the Nazi’s would say, ‘Oh, we’ve been spotted.’ Actually, that was just a ruse.
In 1950, vacuum tubes. CBS predicted the election of Eisenhower in 1952. Then they started shrinking vacuum tubes, every year making them smaller to keep this exponential growth going. That finally hit a wall. They got to a point where they could not shrink the vacuum tubes and keep the vacuum, and that was the end of the shrinking of vacuum tubes. It was not the end of the exponential growth of computing. They went to the fourth paradigm, transistors, which were not small tubes–it was a whole different approach. Now we have had several decades of Moore’s Law, which is basically shrinking the size of components on an integrated circuit on a chip. Gordon Moore first predicted that that would come to an end in 2002. Intel has now said 2022. By that time the key features will be four nanometers, which is about the diameter of twenty carbon atoms.
Around that time we won’t be able to shrink them anymore. That will be the end of Moore’s Law, the fifth paradigm, but it will not be the end of the exponential growth of computing. It will go to the sixth paradigm, which is three-dimensional computing. Chips are flat–they’re two-dimensional. We live in a three-dimensional world. Games are going into the third dimension, why not chips? I talked about that in my 1999 book The Age of Spiritual Machines. It was then a controversial notion. Now, a decade later, there has been so much progress in self-organizing three-dimensional chips that this is now a mainstream proposition. If you talk to Intel scientists, they will tell you they have various prototypes like this working. In fact, in today’s game machines and cell phones there is already multi-layered circuitry, the first step in three-dimensional chips. Intel is now predicting that the cross-over will be in the teen years, well before we run out of steam with flat chips. So we are going to the sixth paradigm of computing.
If you look at this, going back forty years, when I was a student at MIT, that is a billionfold increase in price performance. This whole curve is a trillionfold increase in price performance. We will see a trillionfold again in much less than a century. We will see a billionfold in a quarter of a century.
You have heard these fantastic comparisons that have been in the gaming industry for more than a few years. You’ve seen it, in terms of what is available for games. Pong, which is a simulation of tennis, which is quite crude since 1972. Look at another phenomenon of this graph. Not only does it show a fantastic growth, look at how smooth and predictable this progression is. This looks like the output of some tabletop experiment, but this is millions of people’s innovations and competition around the world. There was a lot of human history in the 20th century, two World Wars, the Cold War, the Great Depression, and a few other things happened. Despite all that unpredictable human history, millions of people and hundreds of countries, you see this very smooth, very inexorable, very predictable progression.
Supercomputers are marching along, and games computers costing hundreds of dollars are only a little bit behind them. Interestingly, this graph, which was in my book The Singularity is Near, which came out two years ago, predicted that we would reach a particular threshold, 10^16, ten to the million billion calculations per second in 2013. That is an important threshold because in Chapter 3 I talk about the different estimates of how much computation it would take to simulate the entire human brain based on the regions we have simulated already. They are all about the same. They come out to about 10^14 and 10^16. I took the more conservative one. There are now several supercomputers slated to reach that within two years, by 2010. We are already only one or two orders of magnitude below that.
I don’t want to dwell on these examples of electronics, because you are familiar with them. This is another interesting graph. When I was growing up in New York I used to hang out at the surplus electronics shops on Canal Street. Other classmates of mine were hanging out at the corner drug store in front of our high school. I would buy for $50 a telephone relay, which is one circuit. In 1968, when I was an undergraduate, I could buy a whole transistor, smaller and better and faster than that relay, for only one dollar. In 2002, you could buy ten million transistors for a dollar. Today it’s three hundred million for a dollar. You have heard these fantastic comparisons, but look at how predictable this is. This looks like it’s a man-made program, but this is the result of millions of people’s innovations and a lot of unpredictable phenomena.
Unlike Gertrude Stein’s roses, it’s not the case that a transistor is a transistor is a transistor. As we made them cheaper, they are actually better because they are smaller. The electrons have less distance to travel, so they are faster. We have had smooth exponential growth in the speed. So the growth of a transistor cycle has come down by half about every year.
Now, what does that mean? That’s 50% deflation for information technologies. There is still a piece of the economy that has to do with information that has very unusual characteristics. Instead of having inflation, you have massive deflation. Depending on what week it is, the congress will worry about inflation or deflation. By the way, it is because of information technologies deflation that we do not have runaway inflation. The Fed is back on inflation, but the proportion of the economy that is comprised of information technology is growing. It will be most of the economy by the 2020′s. There will be a virtual reality environment and games will have taken over the world. Everything will be information technology, including energy.
We will have massive deflation, just like we have during the Depression. The concern is that if you can get the same stuff–and I do mean “stuff,” because nanotechnology will enable us to create very complex three-dimensional products from information files. Today you can take an information file and create a movie, a game, a book, a music album. These used to be physical products, not just information files. Every product is going to be like that. So if you can get the same stuff for half the price a year later, you will consume more, but you are not going to double consumption year after year to keep up with this doubling of price performance. So the signs of the economy as measured in currency will shrink, and for a variety of good reasons that would be a bad thing, but that is actually not what we see.
We actually more than double our consumption every year. We had 18% growth in constant currencies for the last half century each year, despite the fact that you can get twice as much capability each year in information technology. The reason for that is that when price performance reaches a certain level, whole new applications explode on the landscape. Massively multiplayer games were not feasible a few years ago. People didn’t buy iPods for $10,000 which is what it would have cost ten years ago. So when applications reach a certain level, they explode on the landscape.
You may remember fifteen years ago when someone took out a mobile phone in a movie, this was a signal from the director that this person was a member of the power elite. You had to be wealthy to afford them and they didn’t work very well. It took them ten years to put out the first billion cell phones, three years to put out the second billion, and then we put out the third billion in the last year. They are incredibly capable compared to the first ones. They will put out another three billion in the next two years, providing one for every person on the planet.
So this exponential growth actually accelerates our consumption. We more than keep up with this doubling in price performance. The biotechnology revolution, I don’t want to dwell on that, but that is a profound revolution. The Genome Project was controversial in 1990 because mainstream scientists said that using the most advanced equipment from around the world we have collected one ten thousandth of the genome in 1989. Halfway through the project, seven and a half years later, the skeptics were still going strong, saying, “I told you this wasn’t going to work. Here you are halfway through a fifteen year project and you’ve finished one percent of the project.” But that’s actually right on schedule for an exponential progression, because if you double 1% seven more times, you get 100%. That is exactly what happened.
The cost has come down from ten dollars per base pair in 1990. It is a small fraction of a penny today. The first genome cost a billion dollars; we are now close to a thousand-dollar genome. This again looks like some government-mandated program, but this is the result of the competition of thousands of companies around the world. This slope represents the doubling every year of the amount of genetic data, and it has continued past the end of the Genome Project. The project was done in time, and it is now revolutionizing our health and medicine. The new paradigm is to basically model, simulate and reprogram our biology as if it were a set of computer processes. That was never feasible before. We can add new genes. I’m involved in an experimental company where we take lung cells, scrape them out of the throat, add a gene in a petri dish, multiply it by a million (that’s another new technology), inject it back in the body. This has cured a fatal disease, pulmonary hypertension. It’s now being tested in human trials.
So we can turn off genes, we can add new genes. We are going to have not just designer babies but designer baby boomers, which I am personally more interested in. Now health and technology are subject to what I call the “Law of Accelerating Returns,” the doubling of capabilities each year. These technologies will be a million times more capable than they are today in twenty years. Communication technologies, which is certainly fundamental for games, is doubling every year. We live in a linear world. It looked like the World Wide Web came out of nowhere in the mid-1990′s, but you could see it coming if you looked at the exponential projections.
If you are planning a project today that is going to take more than six months to do, the pace of change is so dramatic. You can see that six months is a whole new generation in technology in cell phones and games that you really need to take this exponential progression into account. We are shrinking technology at an exponential rate. These illustrations from Eric Drexler‘s book twenty years ago, which shaped our modern concept of nanotechnology, are now being built. I would say someday we will have blood cell-type devices that they can actually build inside your blood stream and provide therapeutic health functions from inside your body.
You might say that sounds very futuristic, but there are dozens of experiments doing that already today with the first generation of blood-cell sized devices. One cured Type-I diabetes in rats with a blood cell-sized device with seven nanometer pores that lets insulin out in a controlled fashion, blocks antibodies because Type I diabetes is an auto-immune disorder. This is now going into human trials. At MIT there is a device that can scout out cancer cells in the blood stream and destroy them. This is a design of a robotic red blood cell, which we have reverse engineered. These have not yet been built, but they have been analyzed in great detail. These are called respirocytes. They work just like a biological red blood cell, except that they are a thousand times more capable, which brings up an interesting observation about biology. Biology is very capable, intricate and clever, but it is also very suboptimal compared to what we ultimately can build with information technology and nanotechnology.
A conservative analysis of these indicates they are a thousand times more capable. If you replace a portion of your red blood cells with these respirocytes, you could do an Olympic sprint for fifteen minutes without taking a breath, or sit at the bottom of your pool for four hours. “Honey, I’m in the pool,” would take on a whole new significance.
In terms of the Uncanny Valley, there is a lot of concentration on the realism of graphics. We have come a long way since 1972 and the simulation of tennis represented by Pong. Games are not really in the Uncanny Valley because the intelligence behind the characters is not really close enough. That is the key issue. I think we will be in the Uncanny Valley for a very short period. That is where you get close enough to realism that it is uncanny, or “creepy.” It should be called the “Creepy Valley.” If you have characters that are really intelligent, but they’re not quite there, they can seem kind of like demented humans. We want to kind of get past that stage quickly.
Alan Turing identified, I think accurately, that the key to human intelligence was language. He based the Turing test on being able to actually engage in language dialog at a human level. It is entirely a language-based test. We are not at that stage now. We can do interesting things like language translation on a human level and search engines are becoming more facile with understanding language, but we are not at human levels. I believe we will get there by the late 2020′s, because we are also making exponential gains in understanding how the brain works. The spacial resolution of brain scanning is doubling every year. The amount of data we are gathering on a living brain in vivo is doubling every year. But then an issue comes up: Can we understand this data? Doug Hofstadter has said for years that maybe the brain is just below the level necessary to understand our brain. If we were smart enough to understand it, then our brain would have to be that much more complicated, and you can kind of never catch up with it. Maybe there is a mathematical theorem in there, that a complex system cannot be so complex that it can understand its own complexity.
But that is not what we are finding about the brian. We are finding that if you get enough data about specific regions, we can model them, simulate them and test the simulation. There are a dozen regions of the auditory cortex that have been modeled and simulated, and then sophisticated psychoacoustic tests have been applied to the simulation. They get the same results as when testing human auditory perception.
The same thing has been done for the visual cortex. The visual cortex takes a massive amount of data from the human visual stream, reduces it to seven very low resolution movies with different kinds of pattern recognition-based cues, and that’s all there is the brain gets to see. So we have this illusion that we see the world in high resolution when we are actually hallucinating what we think we see based on our memory of former visual scenes from these very low resolution cues.
I had an argument with the head of the Vision Lab at MIT. He said, “Ray, we’re not getting much useful information by reverse engineering the visual cortex in machine vision.” I had been saying for years that in speech recognition work we benefit a great deal from the models of the auditory cortex. I said, “That is because you haven’t done it yet. When you actually get these models, you will find it very useful.” A couple years ago, at AI@50, the fiftieth anniversary of the 1956 conference that gave artificial intelligence its name, he said, “You know, you were right. I just got these models. We applied them to machine vision, and we get a big boost in performance.” A lot of it was actually counterintuitive to how the human visual system actually works.
We have models of simulations of the cerebellum. That is where we do our skill formation. This has been modeled, simulated and tested, and we get similar results. We always wondered, How does a ten year old kid catch a fly ball? I mean all she has to do is solve a double simultaneous differential equation in a couple seconds, and most ten year-olds haven’t taken calculus. So, how does that work? We actually understand how that works now. The cerebellum, using a mathematical technique, actually solves those differential equations without our conscious awareness, but we do have to actually train it, which is why we have to learn skills.
I make the case that well within twenty years we will have models and simulations of all of the regions of the brain. That will give us better facilities to fix things that go wrong. What is really going on in the brain of someone who has schizophrenia? We have these crude psychiatric models today, but we will actually have a much better understanding of how that works. Most importantly, we will have the means of simulating human intelligence. Our game characters then could really be in the Uncanny Valley, and hopefully get through that very quickly. Ultimately, we won’t be able to tell the difference between real and virtual intelligence.
All this is driving economic growth that is underlying economic expansion of the economy. We have gone from $30 to $130 in constant currency of the value of an average hour of human labor over the last 45 years. The adoption of this technology is exponential. You can see that in the gaming world. E-commerce is now $2 trillion. It’s the fifth or sixth largest nation in the world, but it actually has no concept of national boundaries, and there has been smooth exponential progression.
You might say, Now wait a second. Wasn’t there a boom and a bust around the year 2000 in this dot com phenomenon? Yes, but that was not a Main Street phenomenon, that was a Wall Street phenomenon. The investment community looked at the internet and said, “Wow, this is going to take every business model and turn it on its head.” That was actually an accurate insight, but “exponentially” doesn’t mean “instantly.” It takes a decade to go from doubling small numbers to doubling big numbers, and it makes a big difference. So, Wall Street came back three years later, around the year 2000 and said, “Gee, it hasn’t changed everything. I guess we were wrong.” So all the values went down the other way. But meanwhile there was this actual exponential progression.
Now, you can argue, is Google’s 37 to 1 price earnings ratio higher or lower, but it’s not 10,000 to 1 like we had with the dot coms in the early 1990′s. They have $15 billion of real revenue and there are $2 trillion on the internet, and continues to grow. It will be the largest nation in the world, and ultimately it will supersede national boundaries altogether.
Education is progressing. Games has an unfortunate name, just like “artificial intelligence” is an unfortunate name, and “virtual reality” is an unfortunate name. We are stuck with them, but ultimately we can do most of our learning through these massively parallel interactions around the world. I’m on the board of MIT now. We give away all of our courses on OpenCourseWare. There are thousands of schools around the world where kids are taking MIT courses for free. There is going to be all kinds of educational material through interactions, largely from the gaming world. The amount of money we spend on education is growing exponentially because we are destroying jobs at the bottom of the skill ladder and creating new jobs at the top of the skill ladder. There are actually more jobs being created than are destroyed.
If I were a futurist in the year 1900 and someone were to ask me what will happen in a hundred years, I would say, “Today you have a third of your population working on farms and a third working in factories, but in a hundred years that is going to be three percent on farms and three percent in factories.” People would say, “That’s massive unemployment. What are people going to do?” And I would say, “Don’t worry, they’ll find employment in the gaming industry.”
So we are creating these new jobs that are much more interesting and form a part of the definition of who we are. They pay six times as much in constant dollars, and we have gone from 60,000 college students in 1870 to 6 million today. We spend ten times as much in constant currency per capita on K through 12 education. We have exponential growth in education.
Let me give you another example of some technology we have created. There has been continuous development of text-to-speech synthesis since the 1970′s. We took the latest version of that, which sounds quite natural, but it’s synthetic combined it with speech recognition, which we first created in the ’80s, combined it with language translation, which has gotten quite good, and created a translating telephone. I have used this to talk to people around the world. I spoke to this woman in Germany–I spoke English and she heard me in German, and vice versa. We could understand each other quite well… at least as well as we understand each other when people talk the same language. This will be a routine service of your cell phone.
Early in the next decade, devices are going to disappear into our clothing and our belt buckles. We are going to solve this problem of how people like large displays but they also like to watch movies and play games this size by putting them in their eyeglasses. This already exists. They are a little bulky and expensive today, but these will be very low cost and ubiquitous, particularly for full-immersion virtual reality games. It’s very interesting; it becomes very convincing. If you are standing in front of a cliff and there is a big valley in front of you, people will not jump, even though they know they are in a real room. It becomes very convincing. We are going to have to solve this problem of how to spot people and keep track of real reality so we don’t knock down lamps and start fires in real reality. Real reality will continue to be a little irksome for a few years.
We are going to have effective language translation. It will keep track of your conversations–in case you get stuck with something, it won’t wait to be asked. There will be pop-ups telling you it’s their birthday next Tuesday. Reminding you what people’s names are, that will be a big help right there. As you go out to 2029, one of my predictions is that we will have human-level intelligence by that time–Turing test-capable computers. We will have completed the reverse engineering of the human brain, but this will not be an alien invasion of intelligent machines that will compete with us. We are going to augment our own intelligence and nanobots will be able to go into our brains through the capillaries and interact with the biological neurons.
All this has been demonstrated on some level. We already have devices that can go inside your brain and interact with biological neurons. If you take this billionfold increase in the price performance ability, this hundred thousandfold shrinking that we will see over the next quarter century, there will be blood cell-sized devices that can really expand our intelligence. You want to go into virtual reality or a virtual game, the nanobots will shut down the signals coming from your real senses, replace them with the signals that your brain would be receiving if you were in the virtual environment. You go to move your hand and it moves your virtual hand. You are an actor in a virtual environment–you can be yourself or be other people. You don’t have to be the same boring person all the time. You could experience a relationship from the other’s perspective. A kid could become a virtual Ben Franklin, there are all kinds of capabilities.
We are doing that today in our games. This is going to become more and more realistic, more and more full-immersion. We are going to expand our intelligence by merging with technology, but it’s already happening. The fact that we can take out a device this size and with a few keystrokes access all of human knowledge, that is an expansion of our intelligence. We are the only species that does that. We didn’t stay on the ground, we didn’t stay on the planet, and we have not stayed with the limitations of our biology.
It was not in the interest of the species for us to live past child-rearing a thousand years ago, because resources were precious, there was very little food. A few people lived longer, but life expectancy was in the twenties. It was 37 in 1800. There was no sanitation, no antibiotics, no understanding of the germ theory of disease. Schubert and Mozart died in their thirties, and that was typical. It was 40 in 1900.
We have now been adding three months a year to human life expectancy, but that is with health and medicine having been a hit-or-miss affair. It has not been an information technology. Now that we have the means to turn on or off genes, turn on or off enzymes, reprogram our biology the same way we reprogram our games, it is now an information technology. We will be a million times more capable in twenty years.
So, we will be adding four months a year and five months a year. According to my models, in fifteen years from now we will be adding more than a year every year, not just to infant life expectancy but to your remaining life expectancy. As you go forward a year, your remaining life expectancy will move on away from you. The sands of time will be running in, not out. So if you can hang in there, we may get to experience the century ahead. Thank you very much.