|
|
| Mon 26 Sep 2005 16:11:29
| | How Near is the Singularity? | Few writers have inspired as much controversy, hope, or incredulity as Ray Kurzweil. Respected as a successful entrepreneur and clever designer, Kurzweil is better known for his prognostications than for his inventions. Kurzweil, spent much of his youth programming digital computers and creating optical character recognition machines, scanners, synthesizers, and speech recognition programs. Kurzweil has had considerable success turning these innovative ideas into spinoff corporations. During the past two decades, however, he has focused his efforts on predicting future technological trends and developments. Along that vein, Kurzweil created the Kurzweilai.net website, and now regularly gives keynote speeches at myriad technology symposiums. He has written several other books on Artificial Intelligence, including the Age of Intelligent Machines and the more recent The Age of Spiritual Machines. Kurzweil’s current tome, The Singularity is Near: When humans transcend biology, expands on his earlier writings, and puts the finishing touches on what can only be described as mind-boggling predictions.
| Kurzweil’s main thesis, originally described in earlier works but elaborated upon in this book, is that key technologies – computing, biogenetics, and communications – are improving at a double-exponential rate, and that this pace of technological development will continue throughout the 21st century. Indeed, Kurzweil argues that the pace will quicken further, and will result in a technological “singularity” within the next forty years. The author defines the singularity as “technological change so rapid and profound it represents a rupture in the fabric of human history” which will result in nonbiological intelligence (computers) becoming the dominant life form within this century. Humans at that point will need to engage in an intimate and extensive merger with computer technology in order to remain relevant. Kurzweil predicts that this singularity should occur by about 2045 because by that point “the nonbiological intelligence created in that year will be one billion times more powerful than all human intelligence today”.
Kurzweil’s sanguine predictions regarding the future of computer hardware form the core of his beliefs that the GNR (Genetics, Nanotechnology, Robotics) troika will utterly transform every aspect of civilization within the 21st century. Some time in the next few decades, humanity will reach a crossover point, when computer hardware/software surpasses the innate capabilities of biological neurons. At that point, computer hardware will quickly soar past human level intelligence, and these smarter than human computers will then be directed to recursively improve themselves. Individually, humans will have a choice: either accept computers (or computer components) in their brains as surgical or modular implants in order to augment their intellectual capacities, or become increasingly archaic and irrelevant. According to Kurzweil and his proponents, bacteria sized nanobots will become permanent brain fixtures, allowing humans not only to enhance their intellects and remain viable, but also to intimately and dramatically affect their perceptions of themselves and the outside world.
As an intriguing example of the ultimate computing potential of matter, Kurzweil notes that a plain one kilogram rock contains 10^25 atoms, and if all the atoms are used for computing, is theoretically capable of 10^42 calculations per second. The author estimates that the individual computational capacity of a human brain can be functionally emulated by a machine performing 10 quadrillion(10^16) calculations per second. He therefore notes that a one kilogram computer could theoretically “perform all human thought over the last 10,000 years (assumed at 10 billion brains for 10,000 years) in 10 microseconds”. These numbers are, according to Kurzweil, neither academic exercises nor amusing facts. Rather, they are extrapolations based on long-established trends and careful analysis of future capabilities. To Kurzweil, the pertinent question isn’t if we will ever be able to create a one kilogram computer capable of 10x42 calculations, it is what decade in the 21st century such capabilities will arrive.
In order to believe any of these forecasts, the reader will first need to accept that a number of speculative and unproven technologies – such as molecular electronics and advanced molecular manufacturing – will soon be developed. To his credit, Kurzweil presents a formidable quantity of concrete evidence (the book presents 100 pages of footnotes alone) of progress along multiple technical fronts. Furthermore, it is a well established fact in the computer field that semiconductor technology is progressing exponentially. Researchers in the biochemistry, nanotechnology, and robotics fields may not be as acutely aware of exponential growth (there are no equivalents to Moores law in those fields) but scientists are still cognizant of the rapid and increasing pace of development in those nascent industries. Moreover, Kurzweil correctly states that progress in one field, such as semiconductors, has a ripple effect on other industries. For instance, the Blue Gene supercomputers, a marvel of modern semiconductor technology, will be able to simulate precisely the interatomic forces that determines how a protein folds. This capability will have a profound and direct impact upon the microbiology, biochemistry, and genetic engineering industries. Finally, Kurzweil highlights the burgeoning resources that can be directed towards given technology projects. The surge in labor resources (technically trained scientists and researchers are coming out of worldwide Universities at an ever greater rate) combined with the increased financial funding (industries have growing R&D funds, in inflation-adjusted dollars,) and the inexorable rise in computing power, all lead to a substantial aggregate increase in available R&D resources, which in turn allows seemingly intractable problems to be surmounted.
Kurzweil is keenly aware of his plethora of critics, and in Singularity has devoted an entire chapter to dealing with naysayers. These panoply of technical criticisms include the criticism from Malthus (constrained resources) and software (software isn’t progressing as fast as hardware) to the criticisms of lock-in (complex infrastructures will hinder technological development) and failure rates (complex systems won’t work reliably). Yet Kurzweil has responses to all of these issues: Computation becomes more efficient as it progresses, thereby ameliorating if not circumventing resource constraints. Although software development is not progressing as rapidly as corresponding hardware, it is still steadily improving, and should continue to improve for the foreseeable future. Complex computing and communications infrastructures already exist, but they have not prevented or significantly altered the relentless forward march of computing or communications capabilities. Although complex systems initially have a variety of problems, and vexing problems often persist, extremely complex yet highly reliable computer systems are becoming increasingly common.
Along with these technical counterarguments come multiple critiques along philosophical lines. These criticisms seem more trenchant than the technical counterpoints, for most cannot be dismissed as easily as the technical criticisms. Kurzweil has amassed a mountain of technical evidence to show at least exponential growth among the relevant technologies. Even if one doubts that these technologies are progressing at a double-exponential pace, a simple continuation of exponential trends will result in many of Kurzweil’s predictions occurring sometime within the 21st century. Yet discussing philosophical rebuttals is somewhat akin to attempting to nail jelly to the wall. Kurzweil deserves credit for effectively dismantling one of the most famous AI critiques, John Searle’s famous “Chinese Room” argument. In Searle’s “Chinese room” individuals are given a complex set of instructions for converting English to Chinese. Given a sufficiently complex set of instructions and enough time to perform the conversions, the “room” would appear to know Chinese, even though none of the workers in the room spoke anything other than English. Kurzweil makes the obvious point that such a room, in order to work effectively, would need to contain billions of workers, and would have to be approximately as complex as a human brain in order to function effectively. Searle’s Chinese room, Kurzweil notes, essentially describes the working of a neuron.
Other issues regarding consciousness, cognizance, sentience, and spiritual matters, cannot be dealt with as decisively, and it is here that Singularity is weakest. The concept of downloading human minds into computers is enough to make even the most open-minded futurist incredulous. Moreover, Kurzweil’s sanguine assumption that people will be able to seamlessly incorporate myriad artificial components in their skulls presents practical as well as philosophical issues. Although Kurzweil is keenly aware of these issues, and addresses them at length, his explanations may fail to convince or even mollify skeptical readers.
Even if only a small fraction of Kurzweil’s predictions turn out to be accurate, then Kurzweil’s basic thesis is correct. Kurzweil’s main point is that continuing rapid change in multiple technological disciplines will cause profound and disruptive changes in human civilization, and will cause nonbiological intelligence to dominate by the end of the 21st century. The changes that Kurzweil forecasts are so great that, even if he is off by many orders of magnitude in his analysis, the effects on society will still be overwhelming. If, for instance, technical hurdles prevent us from utilizing every atom in a one kilogram device for computing, and we are only able to achieve 10^22 calculations per second from such a device instead of 10^42, we will still have computers that are one trillion times faster than current laptops. Similarly, if sentient, cognizant computers become hundreds of times, rather than quadrillions of times, smarter than humans, they will still inevitably become the dominant life form.
Anyone who has ever speculated on what will become of humans in the 21st century will find this work informative if not enlightening. Although some readers may find Kurzweil’s conclusions preposterous, The Singularity is Near should nevertheless be considered a must read for anyone interested in computers and the future.
|
|
|
Blog | Nano
Small Tech Company Research
site search
