Ray Kurzweil is an American inventor, businessman, and graduate of MIT who has become something of a cultural phenomenon in technocratic circles throughout the United States. He is alternatively described as a: “restless genius” by the Wall Street Journal, ‘the ultimate thinking machine’ by Forbes, and Inc. magazine ranked him #8 among entrepreneurs in the United States, calling him the ‘rightful heir to Thomas Edison”. Kurzweil is the founder of the aptly named Kurzweil Computer Products, Kurzweil Applied Intelligence, Kurzweil Music Systems and invented the Kurzweil 250 Digital Synthesizer, the Kurzweil Cybernetic Poet and the Kurzweil Reading Machine. More than this, he is one of only five members of the Army Science Advisory Group (ASAG), a body responsible for advising the U.S. Army on the priorities of its scientific research, and has testified before the American Congress on his views on how scientific funding should be spent. Aside from these achievements in the realms of business and politics, he has become the icon of a branch of futurism that has a wide audience amongst the influential circles in which he travels. Bill Gates, Dean Kamen, Marvin Minsky, and a startling array of scientists and inventors have gathered around his banner of technological utopia. With such steady excitement in industrialist and political circles it is necessary to explore just what it is that Kurzweil predicts when he discusses the Singularity, why he has predicted it, and what it says about the principles underlying our present understanding of intelligence and technology. These are important things to consider, for few of us are immune to this vision. Implicitly, even if we have never even heard of it. Every day when we assume the price of electronics will go down over time, or when we trust in technological solutions to hunger, pollution, war and disease, we are in some ways tacitly consenting to Kurzweil’s understanding of the nature of technology, and the destiny of humanity.
Kurzweil’s views come from somewhere, from the heady mix of such thinkers as Alan Turing, Hans Moravec, Isaac Asimov, Vernor Vinge and Gordon Moore. Since many readers will be unfamiliar with the terms Kurzweil uses to discuss his ideas, and since Kurzweil himself often develops convoluted chains of associations between these terms, it will then be necessary to elaborate on a number of them. Kurzweil’s understanding of intelligence, artificial intelligence, the law of accelerating returns, the Singularity and what he calls the “saturation” of the universe with intelligence will thus be explained in some detail. Finally, given this background, the consequences of his understanding of intelligence will be explored. Personally, I cannot shake the feeling that it presupposes that there is an underlying teleological continuity between order, computation, evolution and intelligence which has as its ultimate conclusion the Singularity, but which may very well rest upon a tautology.
Alan Turing (1912–1954) was the single most instrumental figure in the development of artificial intelligence, being the first to lecture on computer intelligence and having written the now famous essay “Intelligent Machinery” in 1948. This paper is sometimes called the first manifesto of A.I. and explores the issue of how much a machine could replicate the higher functions of the human brain. It was not published during his life, but Turing would continue to develop the idea that the human brain is in effect a digital computer, and that in this case machines could be made as intelligent as human beings. Kurzweil has particularly high praise for the Turing Machine, Turing Test, and the Church-Turing thesis. Time and again he returns to these three concepts when attempting to draw a connection between human and machine intelligence.
The Universal Turing Machine (not to be mistaken for a Turing Machine, which is more limited) was Turing’s theoretical conception of a machine capable of performing any conceivable computation. It consists of an infinite tape and a device which can read the symbols printed on the tape as instructions. It can in turn print further symbols, based on a small number of possible instructions and provide an output. For Kurzweil, the Turing machine is “a further demonstration of the universality and simplicity of computation”. The simplicity is evident in the small number of logical functions required to perform a potentially infinite number of calculations. The universality will be made evident with a look at the Church-Turing thesis.
The thesis was developed independently by Turing and the American Alonzo Church. It posits that “[t]he universal Turing machine can perform any calculations that any human computer can carry out”. This is based in part on the already perceived equivalence of human and mechanical computation, since in this case computation is seen as nothing more than the application of systematic rules to a numerical problem. For Kurzweil, this has the further consequence that: “problems that are not solvable on a Turing machine cannot be solved by human thought, either”. Yet it is here valuable to note that the “computations” of Turing’s machines were based off the methodologies used by computers (the term then in common use to specify human beings who performed computations according to some fixed method). There is already a deep seeded sense in this work of computation being fundamentally identical with human computation.
The Turing Test is a proposed examination in which a human judge interviews an artificial intelligence and a human foil, without being able to see them. If the judge cannot tell which is the human and which the machine then the machine is considered to possess human-like intelligence. Kurzweil supports the practical, observational basis for determining artificial intelligence which he sees in the Turning Test, and praises the insight of using human dialogue as its deciding factor. This is because of the way Kurzweil perceives the complexity of human dialogue, as well as his trust that it is nevertheless subject to a reductionist interpretation. Most importantly for our discussion, the test is dependent on an outside observer to “judge” if the machine’s performance can be considered intelligent. While the criteria are purposefully left vague, there is a sense that this is the only way to avoid the difficult metaphysical problems surrounding the internal, reflective qualifications for intelligence.
The conclusion that there is no difference between human intelligence and computation is the key principle underlying Kurzweil’s support for Turing’s three proposals, which argue for the universality of computation, the observational and practical basis of intelligence and the equivalency of intelligence with computation. To this end he places the 1937 derivation of the Church-Turing thesis on his timeline of significant events because it lays down for him that: “all problems that a human being can solve can be reduced to a set of algorithms, supporting the idea that machine intelligence and human intelligence are essentially equivalent”.
Kurzweil’s own definition of intelligence as laid out in Spiritual Machines owes a great deal of its form to his reading of Turing, along with the somewhat tautological understanding of the meaning of computation. Despite this, however, it is important to point out that Turing never did propose a universal definition of intelligence. At best he was offering a criterion for human thinking, not of thinking as such, and asked the question:
“May not machines carry out something which ought to be described as thinking but which is very different from what a man does?”
In 1965, the inventor Gordon Moore postulated that the surface areas of transistors were being reduced by about fifty percent every twelve months, resulting in a consistently exponential increase in price-performance approximately every two years. In his own words:
The complexity for minimum component costs has increased at a rate of roughly a factor of two per year […]. Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years.
Kurzweil goes much further than this in stating that a similar law has been in effect throughout all of history. He explains that his law describes “the acceleration of the pace of and the exponential growth of the products of an evolutionary process”. These products include computing technologies, but extend much further than Moore’s Law into biological and evolutionary systems.
Hans Moravec, writing in 1988, likewise used an extrapolation of Moore’s law to argue that machines with greater than human intelligence are the ultimate culmination of biological intelligence, and will ultimately be humanity’s “mind children”. As with most of the thinkers discussed here, he argues for the unity of technological and biological evolution, provides an array of exponential graphs with somewhat subjective points to demonstrate his points and proposes a state in which the universe will at some point “wake up”. He is also very keen on circumventing the second law of thermodynamics, this time through a “subjective infinity” which results from the increasing ease of computation at lower and lower temperatures, allowing our mind children to exist in an information state that would be, essentially, eternal.
This leads to the crux of Kurzweil’s argument, the Singularity. It is a concept which has been gaining in support since its inception in 1963 with I.J. Good’s lecture “Speculations Concerning the First Ultraintelligent Machine”, which attempted to trace out the consequences of advanced computers for our understanding of meaning and intelligence. It is important, then, to remember that Good was a British cryptographer and statistician who worked under Alan Turing on the Colossus project as a member of the secret World War 2 group Hut 8. In concluding this talk, he commented that:
It is more probable than not that, within the twentieth century, an ultraintelligent machine will be built and that it will be the last invention that man need make, since it will lead to an “intelligence explosion.” This will transform society in an unimaginable way. […] The design of the machine will be partly suggested by analogy with several aspects of the human brain and intellect. In particular, the machine will have high linguistic ability and will be able to operate with the meanings of propositions, because to do so will lead to a necessary economy, just as it does in man.
The linguistic component of this approach, as well as the relationship between human and machine intelligence, as we have already seen in Turing, is a reoccurring theme. It goes beyond Turing however, in positing an intellectual impetus which will snowball into an “intelligence explosion”, a notion that does not necessarily occur in Turing’s meditations on machine intelligence, but which forms the heart of modern theories of Singularity.
It seems like there is a considerable gap between the sixties and the eighties in which theorists did not turn their attention to the subject of Singularity. This changed with mathematician and science fiction writer Vernor Vinge’s 1983 article “First Word”, published in the January edition of Omni Magazine, after which a number of other writers and scientists renewed their interest in the subject. In Vinge’s 1993 paper “Technological Singularity”, he again put forth the position, arguing that when: “greater-than-human intelligence drives progress, that progress will be much more rapid”. With the consequence that progress itself would result in the creation of even more intelligence entities on an even shorter time scale.
Vinge had written about the phenomenon of the Singularity before 1983, but instead of doing so in an essay he chose to do it with a story. His 1966 work “Bookworm Run!” was his first articulation of the concept. This brings up the important connection between Kurzweil, the Singularity and science fiction. Even in his 1963 lecture, Good is clearly cognizant of this debt to science fiction when discussing the value of ultraintelligent machines. He warned his listeners to be wary of assuming the benevolence of smarter than human intelligences. Good commented that he found it strange that this point was only ever raised in science fiction and concludes that it: “is sometimes worthwhile to take science fiction seriously”. Vinge also notes that science fiction writers were in fact the first to feel the effects of the Singularity:
More and more, these writers felt an opaque wall across the future. Once, they could put such fantasies millions of years in the future. Now they saw that their most diligent extrapolations resulted in the unknowable … soon. Once, galactic empires might have seemed a Post-Human domain. Now, sadly, even interplanetary ones are.
This relationship of science fiction to science is applicable to Kurzweil himself, as he chooses to begin The Singularity is Near with a discussion of the Tom Swift Jr. series of books and their influences on his early development. The series ran from 1954 to 1971 and focused around the adventures of Tom Swift Jr, the son of the inventor Tom Swift, himself the wealthy CEO of the imaginary Swift Enterprises. The overall message of the series was one of scientific optimism, in which Tom would regularly retreat to his basement laboratory, only to emerge with the solution to the world’s biggest problems (conveniently also finding ways to make more money for Swift Enterprises in the process).
While foregrounding this scientistic optimism, however, it is surprising that Kurzweil makes so little reference to the writings of Isaac Asimov (1920-1992), only mentioning him briefly in The Age Of Spiritual Machines to point out that any superintelligent or self-replicating machine would have to have something like Asimov’s three laws of robotics in order to prevent it from eliminating the human race either through intention or accident. In his 1956 short story “The Last Question” Asimov wrote about a series of ever more complex computers attempting to solve the problem of how to reverse entropy in the universe. The story concludes with an ultraintelligent machine making the Godlike command “let there be light”, reversing the inexorable flow of entropy and restarting existence itself. More than this, a consistent theme through Asimov’s writing was the relationship of robots (Artificial Intelligences) to concepts of empire, which culminates first in a planetary intelligence called Gaia and then in the prediction of a universal intelligence called Galaxia, which bears a striking resemblance to Kurzweil’s own depiction of a universe “saturated” with intelligence. Yet what exactly does he mean by this intelligence?
According to Kurzweil’s definition, intelligence is: “The ability to use optimally limited resources –including time– to achieve a set of goals (which may include survival, communication, solving problems, recognizing patterns, performing skills)”. The relationship of artificial intelligence and intelligence for Kurzweil is a straightforward one. Artificial intelligence is the same as human intelligence except for the fact that it does not rely upon a biological substrate. This point is made clear in a twenty thousand dollar bet Kurzweil made with Mitchell Kapor on a computer’s ability to pass for human by the year 2029. In the terms of the bet both competitors agreed on the definition that a computer “is any form of nonbiological intelligence […] and may include any form of technology, but may not include a biological Human (enhanced or otherwise) nor biological neurons”. Thus the distinction between human and machine intelligence is merely one of substrate, and not of kind.
Kurzweil’s trust in biological intelligence’s ability to produce non-biological intelligence then leads him to the conclusion that not only is artificial intelligence an inevitable outcome of biological intelligence, but it will also be both quantitatively as well as qualitatively better than its predecessor. This is because while there is fundamentally no difference between human and machine intelligence, machine intelligence has the benefit of being free from the localized corporealness of an animal body. It can rapidly share its knowledge, make copies of itself and modify itself in ways that human intelligence can not. While Kurzweil does praise the workings of the human mind insofar as it is a stepping stone to bigger and better things, he nevertheless assures his readers at every step that “the architecture of the human brain is […] profoundly limited”. This is in contrast to how he envisions machine intelligences, which will have freedom of “design and architecture ([…] they won’t be constrained by biological limitations, such as the slow switching speed of our interneuronal connections or a fixed skull size) as well as consistent performance at all times”. As a result of this “[c]omputers will prove more capable of solving unsolvable problems than humans will”, which in the end, he deems to be the pinnacle of biological intellectual achievement.
Kurzweil justifies this view on the trajectory of intelligence through his “Law of Accelerating Returns”, which lays down that: “As order exponentially increases, time exponentially speeds up (i.e., the time interval between salient events grows shorter as time passes)”. In other words, order begets more order exponentially, creating ever shortening gaps between events that represent substantial increases of order. For example: when one compares the development of computers over time, or our understanding of biological processes. A consequence of this is that, while entropy in the universe is increasing, there exists “pockets” in which it actually decreases, leading to ever higher forms of complexity, and, ultimately, intelligence. What Kurzweil means by salient events are events in the history of computation and evolution which demonstrate accelerating growth curves in their degree of complexity. In The Singularity is Near he provides a small army of graphs plotting such things as computing power versus cost in Moore’s Law, brain scan resolution versus year of development (time), and, most dramatically, time to next event versus time before the present for milestones as varied as the birth of the Milky Way, the emergence of life, stone tools, democracy and modern physics. By placing these events on one accelerating continuum he hopes to demonstrate how his law is a universal one which can account for the observed increasing order in the universe. To do so he begins with the order and formation of the universe itself, but rapidly focuses on the earth, and then the human, and then even goes so far as to consider the United States as the pinnacle of this development. This telescopic universality will be vital to his argument, for Kurzweil does not hide the fact that his law of accelerating returns is directly inspired by Moore’s Law on Integrated Circuits.
The Singularity, or technological singularity, for Kurzweil is “a future period during which the pace of technological change will be so rapid, its impact so deep, that human life will be irreversibly transformed”. This theoretical period during which technological progress will be measured in seconds, rather than months or years, is the ultimate result of smarter than human A.I. continuing the law of accelerating returns past the point of human comprehension. The Singularity derives its name from the singularity observed in black holes, whose masses are entirely compressed into a point containing essentially no volume. This results in an infinite gravitational pull that does not permit information from escaping its event horizon, and that represents the point at which our ability to make meaningful predictions breaks down
Despite the historical black hole represented by the Singularity, Kurzweil uses it to justify his view that when a planet: “yields a technology-creating species and that species creates computation […] it is only a matter of a few centuries before its intelligence saturates the matter and energy in its vicinity”. What he means by the saturation of the universe is that we will in effect be “utilizing the matter and energy patterns for computation to an optimal degree, based on our understanding of the physics of computation”. In order to understand this position it is vital to know that Kurzweil possesses an almost Pythagorean understanding of information as being at the heart of all things. Everything from stones to people represents information in its varying degrees of complexity. Thus, what is “saturating” the universe in this scheme can be understood as an ultra-refined pattern of order, which ultimately expresses itself in some universe-wide intelligence.
It is apparent from the previous discussion that for Kurzweil intelligence is by its very nature a kind of computation and is not substantially different in humans or machines. Kurzweil, as for most of the previously mentioned thinkers, sees the work that Alan Turing did through the 1940s and 50s as being instrumental in making this connection. However, it is important to remember that Turing himself did not believe in any one standard for intelligence and instead postulated that while a machine will be intelligence, its intelligence may be of a substantially different kind than that of its human counterparts. One consequence of this disconnect between Kurzweil’s understanding of Turing and Turing’s thought is that Kurzweil turns intelligence into one universal phenomenon similar to light or heat, whereas Turing himself made no such claims and could only describe “criterion” for thought. Since intelligence as computation is like a unified force of nature it is possible to place it under some equally universal understanding of evolution. Yet this understanding of universal evolution is one which is driven by direct goals rather than through random mutations and genetic selection, as in Darwinian evolution. For example: in terms of Darwinian evolution there is, technically speaking, no direct difference in the “fitness” of a slug or a human, whereas under Kurzweil’s computational understanding of evolution there is, since its direction leads to higher forms of complexity and intelligence.
This view is supported by Kurzweil’s understanding of evolution and its equivalency with evolutionary algorithms now in use to engineer intelligent strategies for commercial problems. As has already been seen, for many of the thinkers associated with Kurzweil, there is no set distinction between biological and technological evolution; it is seen as the processes in which some kind of higher order and sophistication in achieved in a chaotic system because of evolutionary pressures, moving first in the biological world and then the mechanical one. Evolutionary algorithms are computer programs made to design “intelligent” solutions to engineering problems. They do this based on emulating a process like natural selection, with offspring, “genetic” mutations, inheritance, and selective pressures, but in a situation in which the programmers can dictate the terms of “survival” (i.e. a more aerodynamic wing structure, or efficient engine). He makes the association between algorithms and evolution very clear when he states that over generations of evolution:
the order (suitability of information for a purpose) of the design of organisms increases, with the purpose being survival. In an ‘evolutionary algorithm’ […] the purpose may be defined to be the discovery of a solution to a complex problem.
Since he sees evolutionary algorithms as being essentially equivalent to evolution, there is nothing conceptually preventing him from transferring the same solution oriented structure of an algorithm onto all evolutionary processes.
The question which then arises is what is the solution to the evolutionary algorithm working in nature? Kurzweil sees evolution as the means of increasing order in an otherwise chaotic system, and order itself is defined as “information that fits a purpose”. In the end, this purpose is the expansion of intelligence, which would seem circular if it did not have its culmination in the Singularity. This is because the circularity of the expansion of intelligence through evolution ultimately cultivates in the “critical mass” of universal intelligence. Kurzweil then feels justified in making the claim that everything from rocks to people to galaxies are fundamentally composed of information being slowly ordered to higher degrees of computation through evolution, with the ultimate goal being the expansion of intelligence in the universe. Evolution thus understood is why the law of accelerating returns must redirect and ultimately marginalize the role of entropy in the universe. It seems fair to say that for Kurzweil an entropic conclusion to existence could not allow for the progress (understood by increases in order and computing power) experienced by human civilization throughout its history. In short, the history of evolution, which we have to some extent already experienced, is only made possible by its conclusion in a Singularity and a saturated universe, rather than an entropic one.
Kurzweil can make his radical claims about the Singularity because of his view that intelligence, computation, order and evolution are essentially equivalent. They are collectively seen to be a force of nature striving for accelerated expansion throughout the universe. These four distinct, but closely related, phenomena may or may not be as identical as the Kurzweil implies, for it is an association which Kurzweil never directly makes and seems to assume as a given about the natural world. Yet the axiomatic nature of this standpoint is vital to understanding his position, since it is the underlying principle of a goal which allows the equivalency between these four concepts. Computation moves towards an answer and order toward a greater suitability of the information within it to some purpose. In this way evolution is a mere expression of order insofar as the “suitability” in question is survival, or, in Kurzweil’s algorithmic understanding: any goal at all. Intelligence then, understood as the ability to optimally use resources to achieve goals, is in a fundamental way not substantially different from order, evolution and computation, but only differs in the plurality and complexity of its various goals.
Ultimately, it seems we are left with two alternatives, both which derive their argumentative force from purely axiomatic understandings of the world as teleological or otherwise. One possibility is that we must accept that intelligence and the history of intelligence is teleological in nature, in which case it can be equated to goal-fulfillment in its various manifestations, and seems to be subject to Kurzweil’s laws. The other option is that it is not teleological, in which case we will have to come to terms with the fact that everything about what we perceive as progress and purpose are largely incidental and accidental, and ultimately represent no great losses, gains or certainties. In either case the argument has little bearing on the possibility or desirability of artificial intelligence or the Singularity and has a great deal more to do with how comprehensible and manageable these phenomena will ultimately prove to be should they ever manifest themselves. Even if intelligence is not teleological, a random, purposeless system of intelligence could still produce effects similar to the Singularity. Thus if Kurzweil is right or if he is wrong, the implications of his work reaches far beyond the realms of engineering, computer science and science fiction from which they sprang. Rather, they reveal the extreme ambiguity at the heart of our understanding of ourselves and our relation to the world around us.
For More Information:
Dyson, George B. Darwin Among the Machines: The Evolution of Global Intelligence. New York: Helix Books, 1997.
Gardner, James N. Biocosm: The New Scientific Theory of Evolution – Intelligent Life is the Architect of the Universe. Hawaii: Inner Ocean, 2003.
Good, I.J. “Speculations Concerning the First Ultraintelligent Machine”. (1999. Aeiveos Research Library. 21 March. 2008.) <http://web.archive.org/web/20010527181244/http://www.aeiveos.com/~bradbur y/Authors/Computing/Good-IJ/SCtFUM.html>
Kurzweil, Ray. The Age of Spiritual Machines: When Computers Exceed Human Intelligence. New York: Penguin Books, 2000.
—. The Singularity is Near: When Humans Transcend Biology. New York: Penguin Books, 2005.
Kurzweilai.net. “About Kurzweil – Speaking Engagements, Press, and More”. 2008. http://www.kurzweilai.net/about/frame.html
Kurzweil, Ray and Mitchell Kapor. “By 2029 no computer – or “machine intelligence” – will have passed the Turing Test”. 2001. The Long Now Foundation. 23 March 2008. <http://www.longbets.org/1>
Moore, Gordon. “Cramming More Components onto Integrated Circuits”. In Electronics (Vol 38. April 19, 1965).
Moravec, Hans. Mind Children: the Future of Robot and Human Intelligence. Massachusetts: Harvard University Press, 1988.
Strauss, Linda M. “Untitled Review” in Science, Technology, & Human Values, Vol. 17, No. 3. (Summer, 1992), pp. 396-401.
Turing, Alan Mathison. The Essential Turing: Seminal Writings in Computing, Logic, Philosophy, Artificial Intelligence, and Artificial Life, Plus the Secrets of Enigma. New York: Oxford University Press, 2004.
Vinge, Vernor. “Technological Singularity”. 1993. ROHAN Academic Computing 21 March. 2008. <http://www-rohan.sdsu.edu/faculty/vinge/misc/singularity.html>