Books by Kurzweil, Ray

Kurzweil, Ray. The Age of Spiritual Machines. New York: Penguin Books, 1999. ISBN 978-0-14-028202-3.
Ray Kurzweil is one of the most vocal advocates of the view that the exponential growth in computing power (and allied technologies such as storage capacity and communication bandwidth) at constant cost which we have experienced for the last half century, notwithstanding a multitude of well-grounded arguments that fundamental physical limits on the underlying substrates will bring it to an end (all of which have proven to be wrong), will continue for the foreseeable future: in all likelihood for the entire twenty-first century. Continued exponential growth in a technology for so long a period is unprecedented in the human experience, and the consequences as the exponential begins to truly “kick in” (although an exponential curve is self-similar, its consequences as perceived by observers whose own criteria for evaluation are more or less constant will be seen to reach a “knee” after which they essentially go vertical and defy prediction). In The Singularity Is Near (October 2005), Kurzweil argues that once the point is reached where computers exceed the capability of the human brain and begin to design their own successors, an almost instantaneous (in terms of human perception) blow-off will occur, with computers rapidly converging on the ultimate physical limits on computation, with capabilities so far beyond those of humans (or even human society as a whole) that attempting to envision their capabilities or intentions is as hopeless as a microorganism's trying to master quantum field theory. You might want to review my notes on 2005's The Singularity Is Near before reading the balance of these comments: they provide context as to the extreme events Kurzweil envisions as occurring in the coming decades, and there are no “spoilers” for the present book.

When assessing the reliability of predictions, it can be enlightening to examine earlier forecasts from the same source, especially if they cover a period of time which has come and gone in the interim. This book, published in 1999 near the very peak of the dot-com bubble provides such an opportunity, and it provides a useful calibration for the plausibility of Kurzweil's more recent speculations on the future of computing and humanity. The author's view of the likely course of the 21st century evolved substantially between this book and Singularity—in particular this book envisions no singularity beyond which the course of events becomes incomprehensible to present-day human intellects. In the present volume, which employs the curious literary device of “trans-temporal chat” between the author, a MOSH (Mostly Original Substrate Human), and a reader, Molly, who reports from various points in the century her personal experiences living through it, we encounter a future which, however foreign, can at least be understood in terms of our own experience.

This view of the human prospect is very odd indeed, and to this reader more disturbing (verging on creepy) than the approach of a technological singularity. What we encounter here are beings, whether augmented humans or software intelligences with no human ancestry whatsoever, that despite having at hand, by the end of the century, mental capacity per individual on the order of 1024 times that of the human brain (and maybe hundreds of orders of magnitude more if quantum computing pans out), still have identities, motivations, and goals which remain comprehensible to humans today. This seems dubious in the extreme to me, and my impression from Singularity is that the author has rethought this as well.

Starting from the publication date of 1999, the book serves up surveys of the scene in that year, 2009, 2019, 2029, and 2099. The chapter describing the state of computing in 2009 makes many specific predictions. The following are those which the author lists in the “Time Line” on pp. 277–278. Many of the predictions in the main text seem to me to be more ambitious than these, but I shall go with those the author chose as most important for the summary. I have reformatted these as a numbered list to make them easier to cite.

  1. A $1,000 personal computer can perform about a trillion calculations per second.
  2. Personal computers with high-resolution visual displays come in a range of sizes, from those small enough to be embedded in clothing and jewelry up to the size of a thin book.
  3. Cables are disappearing. Communication between components uses short-distance wireless technology. High-speed wireless communication provides access to the Web.
  4. The majority of text is created using continuous speech recognition. Also ubiquitous are language user interfaces (LUIs).
  5. Most routine business transactions (purchases, travel, reservations) take place between a human and a virtual personality. Often, the virtual personality includes an animated visual presence that looks like a human face.
  6. Although traditional classroom organization is still common, intelligent courseware has emerged as a common means of learning.
  7. Pocket-sized reading machines for the blind and visually impaired, “listening machines” (speech-to-text conversion) for the deaf, and computer-controlled orthotic devices for paraplegic individuals result in a growing perception that primary disabilities do not necessarily impart handicaps.
  8. Translating telephones (speech-to-speech language translation) are commonly used for many language pairs.
  9. Accelerating returns from the advance of computer technology have resulted in continued economic expansion. Price deflation, which has been a reality in the computer field during the twentieth century, is now occurring outside the computer field. The reason for this is that virtually all economic sectors are deeply affected by the accelerating improvements in the price performance of computing.
  10. Human musicians routinely jam with cybernetic musicians.
  11. Bioengineered treatments for cancer and heart disease have greatly reduced the mortality from these diseases.
  12. The neo-Luddite movement is growing.

I'm not going to score these in detail, as that would be both tedious and an invitation to endless quibbling over particulars, but I think most readers will agree that this picture of computing in 2009 substantially overestimates the actual state of affairs in the decade since 1999. Only item (3) seems to me to be arguably on the way to achievement, and yet I do not have a single wireless peripheral connected to any of my computers and Wi-Fi coverage remains spotty even in 2011. Things get substantially more weird the further out you go, and of course any shortfall in exponential growth lowers the baseline for further extrapolation, shifting subsequent milestones further out.

I find the author's accepting continued exponential growth as dogma rather off-putting. Granted, few people expected the trend we've lived through to continue for so long, but eventually you begin to run into physical constraints which seem to have little wiggle room for cleverness: the finite size of atoms, the electron's charge, and the speed of light. There's nothing wrong with taking unbounded exponential growth as a premise and then exploring what its implications would be, but it seems to me any forecast which is presented as a plausible future needs to spend more time describing how we'll actually get there: arm waving about three-dimensional circuitry, carbon nanotubes, and quantum computing doesn't close the sale for me. The author entirely lost me with note 3 to chapter 12 (p. 342), which concludes:

If engineering at the nanometer scale (nanotechnology) is practical in the year 2032, then engineering at the picometer scale should be practical in about forty years later (because 5.64 = approximately 1,000), or in the year 2072. Engineering at the femtometer (one thousandth of a trillionth of a meter, also referred to as a quadrillionth of a meter) scale should be feasible, therefore, by around the year 2112. Thus I am being a bit conservative to say that femtoengineering is controversial in 2099.

Nanoengineering involves manipulating individual atoms. Picoengineering will involve engineering at the level of subatomic particles (e.g., electrons). Femtoengineering will involve engineering inside a quark. This should not seem particularly startling, as contemporary theories already postulate intricate mechanisms within quarks.

This is just so breathtakingly wrong I am at a loss for where to begin, and it was just as completely wrong when the book was published two decades ago as it is today; nothing relevant to these statements has changed. My guess is that Kurzweil was thinking of “intricate mechanisms” within hadrons and mesons, particles made up of quarks and gluons, and not within quarks themselves, which then and now are believed to be point particles with no internal structure whatsoever and are, in any case, impossible to isolate from the particles they compose. When Richard Feynman envisioned molecular nanotechnology in 1959, he based his argument on the well-understood behaviour of atoms known from chemistry and physics, not a leap of faith based on drawing a straight line on a sheet of semi-log graph paper. I doubt one could find a single current practitioner of subatomic physics equally versed in the subject as was Feynman in atomic physics who would argue that engineering at the level of subatomic particles would be remotely feasible. (For atoms, biology provides an existence proof that complex self-replicating systems of atoms are possible. Despite the multitude of environments in the universe since the big bang, there is precisely zero evidence subatomic particles have ever formed structures more complicated than those we observe today.)

I will not further belabour the arguments in this vintage book. It is an entertaining read and will certainly expand your horizons as to what is possible and introduce you to visions of the future you almost certainly have never contemplated. But for a view of the future which is simultaneously more ambitious and plausible, I recommend The Singularity Is Near.

June 2011 Permalink

Kurzweil, Ray. How to Create a Mind. New York: Penguin Books, 2012. ISBN 978-0-14-312404-7.
We have heard so much about the exponential growth of computing power available at constant cost that we sometimes overlook the fact that this is just one of a number of exponentially compounding technologies which are changing our world at an ever-accelerating pace. Many of these technologies are interrelated: for example, the availability of very fast computers and large storage has contributed to increasingly making biology and medicine information sciences in the era of genomics and proteomics—the cost of sequencing a human genome, since the completion of the Human Genome Project, has fallen faster than the increase of computer power.

Among these seemingly inexorably rising curves have been the spatial and temporal resolution of the tools we use to image and understand the structure of the brain. So rapid has been the progress that most of the detailed understanding of the brain dates from the last decade, and new discoveries are arriving at such a rate that the author had to make substantial revisions to the manuscript of this book upon several occasions after it was already submitted for publication.

The focus here is primarily upon the neocortex, a part of the brain which exists only in mammals and is identified with “higher level thinking”: learning from experience, logic, planning, and, in humans, language and abstract reasoning. The older brain, which mammals share with other species, is discussed in chapter 5, but in mammals it is difficult to separate entirely from the neocortex, because the latter has “infiltrated” the old brain, wiring itself into its sensory and action components, allowing the neocortex to process information and override responses which are automatic in creatures such as reptiles.

Not long ago, it was thought that the brain was a soup of neurons connected in an intricately tangled manner, whose function could not be understood without comprehending the quadrillion connections in the neocortex alone, each with its own weight to promote or inhibit the firing of a neuron. Now, however, it appears, based upon improved technology for observing the structure and operation of the brain, that the fundamental unit in the brain is not the neuron, but a module of around 100 neurons which acts as a pattern recogniser. The internal structure of these modules seems to be wired up from directions from the genome, but the weights of the interconnections within the module are adjusted as the module is trained based upon the inputs presented to it. The individual pattern recognition modules are wired both to pass information on matches to higher level modules, and predictions back down to lower level recognisers. For example, if you've seen the letters “appl” and the next and final letter of the word is a smudge, you'll have no trouble figuring out what the word is. (I'm not suggesting the brain works literally like this, just using this as an example to illustrate hierarchical pattern recognition.)

Another important discovery is that the architecture of these pattern recogniser modules is pretty much the same regardless of where they appear in the neocortex, or what function they perform. In a normal brain, there are distinct portions of the neocortex associated with functions such as speech, vision, complex motion sequencing, etc., and yet the physical structure of these regions is nearly identical: only the weights of the connections within the modules and the dyamically-adapted wiring among them differs. This explains how patients recovering from brain damage can re-purpose one part of the neocortex to take over (within limits) for the portion lost.

Further, the neocortex is not the rat's nest of random connections we recently thought it to be, but is instead hierarchically structured with a topologically three dimensional “bus” of pre-wired interconnections which can be used to make long-distance links between regions.

Now, where this begins to get very interesting is when we contemplate building machines with the capabilities of the human brain. While emulating something at the level of neurons might seem impossibly daunting, if you instead assume the building block of the neocortex is on the order of 300 million more or less identical pattern recognisers wired together at a high level in a regular hierarchical manner, this is something we might be able to think about doing, especially since the brain works almost entirely in parallel, and one thing we've gotten really good at in the last half century is making lots and lots of tiny identical things. The implication of this is that as we continue to delve deeper into the structure of the brain and computing power continues to grow exponentially, there will come a point in the foreseeable future where emulating an entire human neocortex becomes feasible. This will permit building a machine with human-level intelligence without translating the mechanisms of the brain into those comparable to conventional computer programming. The author predicts “this will first take place in 2029 and become routine in the 2030s.”

Assuming the present exponential growth curves continue (and I see no technological reason to believe they will not), the 2020s are going to be a very interesting decade. Just as few people imagined five years ago that self-driving cars were possible, while today most major auto manufacturers have projects underway to bring them to market in the near future, in the 2020s we will see the emergence of computational power which is sufficient to “brute force” many problems which were previously considered intractable. Just as search engines and free encyclopedias have augmented our biological minds, allowing us to answer questions which, a decade ago, would have taken days in the library if we even bothered at all, the 300 million pattern recognisers in our biological brains are on the threshold of having access to billions more in the cloud, trained by interactions with billions of humans and, perhaps eventually, many more artificial intelligences. I am not talking here about implanting direct data links into the brain or uploading human brains to other computational substrates although both of these may happen in time. Instead, imagine just being able to ask a question in natural language and get an answer to it based upon a deep understanding of all of human knowledge. If you think this is crazy, reflect upon how exponential growth works or imagine travelling back in time and giving a demo of Google or Wolfram Alpha to yourself in 1990.

Ray Kurzweil, after pioneering inventions in music synthesis, optical character recognition, text to speech conversion, and speech recognition, is now a director of engineering at Google.

In the Kindle edition, the index cites page numbers in the print edition to which the reader can turn since the electronic edition includes real page numbers. Index items are not, however, directly linked to the text cited.

February 2014 Permalink

Kurzweil, Ray. The Singularity Is Near. New York: Viking, 2005. ISBN 0-670-03384-7.
What happens if Moore's Law—the annual doubling of computing power at constant cost—just keeps on going? In this book, inventor, entrepreneur, and futurist Ray Kurzweil extrapolates the long-term faster than exponential growth (the exponent is itself growing exponentially) in computing power to the point where the computational capacity of the human brain is available for about US$1000 (around 2020, he estimates), reverse engineering and emulation of human brain structure permits machine intelligence indistinguishable from that of humans as defined by the Turing test (around 2030), and the subsequent (and he believes inevitable) runaway growth in artificial intelligence leading to a technological singularity around 2045 when US$1000 will purchase computing power comparable to that of all presently-existing human brains and the new intelligence created in that single year will be a billion times greater than that of the entire intellectual heritage of human civilisation prior to that date. He argues that the inhabitants of this brave new world, having transcended biological computation in favour of nanotechnological substrates “trillions of trillions of times more capable” will remain human, having preserved their essential identity and evolutionary heritage across this leap to Godlike intellectual powers. Then what? One might as well have asked an ant to speculate on what newly-evolved hominids would end up accomplishing, as the gap between ourselves and these super cyborgs (some of the precursors of which the author argues are alive today) is probably greater than between arthropod and anthropoid.

Throughout this tour de force of boundless technological optimism, one is impressed by the author's adamantine intellectual integrity. This is not an advocacy document—in fact, Kurzweil's view is that the events he envisions are essentially inevitable given the technological, economic, and moral (curing disease and alleviating suffering) dynamics driving them. Potential roadblocks are discussed candidly, along with the existential risks posed by the genetics, nanotechnology, and robotics (GNR) revolutions which will set the stage for the singularity. A chapter is devoted to responding to critics of various aspects of the argument, in which opposing views are treated with respect.

I'm not going to expound further in great detail. I suspect a majority of people who read these comments will, in all likelihood, read the book themselves (if they haven't already) and make up their own minds about it. If you are at all interested in the evolution of technology in this century and its consequences for the humans who are creating it, this is certainly a book you should read. The balance of these remarks discuss various matters which came to mind as I read the book; they may not make much sense unless you've read it (You are going to read it, aren't you?), but may highlight things to reflect upon as you do.

  • Switching off the simulation. Page 404 raises a somewhat arcane risk I've pondered at some length. Suppose our entire universe is a simulation run on some super-intelligent being's computer. (What's the purpose of the universe? It's a science fair project!) What should we do to avoid having the simulation turned off, which would be bad? Presumably, the most likely reason to stop the simulation is that it's become boring. Going through a technological singularity, either from the inside or from the outside looking in, certainly doesn't sound boring, so Kurzweil argues that working toward the singularity protects us, if we be simulated, from having our plug pulled. Well, maybe, but suppose the explosion in computing power accessible to the simulated beings (us) at the singularity exceeds that available to run the simulation? (This is plausible, since post-singularity computing rapidly approaches its ultimate physical limits.) Then one imagines some super-kid running top to figure out what's slowing down the First Superbeing Shooter game he's running and killing the CPU hog process. There are also things we can do which might increase the risk of the simulation's being switched off. Consider, as I've proposed, precision fundamental physics experiments aimed at detecting round-off errors in the simulation (manifested, for example, as small violations of conservation laws). Once the beings in the simulation twig to the fact that they're in a simulation and that their reality is no more accurate than double precision floating point, what's the point to letting it run?
  • Fifty bits per atom? In the description of the computational capacity of a rock (p. 131), the calculation assumes that 100 bits of memory can be encoded in each atom of a disordered medium. I don't get it; even reliably storing a single bit per atom is difficult to envision. Using the “precise position, spin, and quantum state” of a large ensemble of atoms as mentioned on p. 134 seems highly dubious.
  • Luddites. The risk from anti-technology backlash is discussed in some detail. (“Ned Ludd” himself joins in some of the trans-temporal dialogues.) One can imagine the next generation of anti-globalist demonstrators taking to the streets to protest the “evil corporations conspiring to make us all rich and immortal”.
  • Fundamentalism. Another risk is posed by fundamentalism, not so much of the religious variety, but rather fundamentalist humanists who perceive the migration of humans to non-biological substrates (at first by augmentation, later by uploading) as repellent to their biological conception of humanity. One is inclined, along with the author, simply to wait until these folks get old enough to need a hip replacement, pacemaker, or cerebral implant to reverse a degenerative disease to motivate them to recalibrate their definition of “purely biological”. Still, I'm far from the first to observe that Singularitarianism (chapter 7) itself has some things in common with religious fundamentalism. In particular, it requires faith in rationality (which, as Karl Popper observed, cannot be rationally justified), and that the intentions of super-intelligent beings, as Godlike in their powers compared to humans as we are to Saccharomyces cerevisiae, will be benign and that they will receive us into eternal life and bliss. Haven't I heard this somewhere before? The main difference is that the Singularitarian doesn't just aspire to Heaven, but to Godhood Itself. One downside of this may be that God gets quite irate.
  • Vanity. I usually try to avoid the “Washington read” (picking up a book and flipping immediately to the index to see if I'm in it), but I happened to notice in passing I made this one, for a minor citation in footnote 47 to chapter 2.
  • Spindle cells. The material about “spindle cells” on pp. 191–194 is absolutely fascinating. These are very large, deeply and widely interconnected neurons which are found only in humans and a few great apes. Humans have about 80,000 spindle cells, while gorillas have 16,000, bonobos 2,100 and chimpanzees 1,800. If you're intrigued by what makes humans human, this looks like a promising place to start.
  • Speculative physics. The author shares my interest in physics verging on the fringe, and, turning the pages of this book, we come across such topics as possible ways to exceed the speed of light, black hole ultimate computers, stable wormholes and closed timelike curves (a.k.a. time machines), baby universes, cold fusion, and more. Now, none of these things is in any way relevant to nor necessary for the advent of the singularity, which requires only well-understood mainstream physics. The speculative topics enter primarily in discussions of the ultimate limits on a post-singularity civilisation and the implications for the destiny of intelligence in the universe. In a way they may distract from the argument, since a reader might be inclined to dismiss the singularity as yet another woolly speculation, which it isn't.
  • Source citations. The end notes contain many citations of articles in Wired, which I consider an entertainment medium rather than a reliable source of technological information. There are also references to articles in Wikipedia, where any idiot can modify anything any time they feel like it. I would not consider any information from these sources reliable unless independently verified from more scholarly publications.
  • “You apes wanna live forever?” Kurzweil doesn't just anticipate the singularity, he hopes to personally experience it, to which end (p. 211) he ingests “250 supplements (pills) a day and … a half-dozen intravenous therapies each week”. Setting aside the shots, just envision two hundred and fifty pills each and every day! That's 1,750 pills a week or, if you're awake sixteen hours a day, an average of more than 15 pills per waking hour, or one pill about every four minutes (one presumes they are swallowed in batches, not spaced out, which would make for a somewhat odd social life). Between the year 2000 and the estimated arrival of human-level artificial intelligence in 2030, he will swallow in excess of two and a half million pills, which makes one wonder what the probability of choking to death on any individual pill might be. He remarks, “Although my program may seem extreme, it is actually conservative—and optimal (based on my current knowledge).” Well, okay, but I'd worry about a “strategy for preventing heart disease [which] is to adopt ten different heart-disease-prevention therapies that attack each of the known risk factors” running into unanticipated interactions, given how everything in biology tends to connect to everything else. There is little discussion of the alternative approach to immortality with which many nanotechnologists of the mambo chicken persuasion are enamoured, which involves severing the heads of recently deceased individuals and freezing them in liquid nitrogen in sure and certain hope of the resurrection unto eternal life.

October 2005 Permalink