Faith in the Machine
Ian H. Hutchinson*
20 Horace Rd.
Belmont, MA 02178
From:Perspectives on Science and Christian Faith 52 (December 2000): 260--262.
Ray Kurzweil, The Age of Spiritual Machines: When Computers Exceed Human Intelligence (New York: Viking Penguin, 1999). 352 pages. ISBN: 0670882178.
Most secular futurologists set out with a few plausible, but unverified, principles of development and try to imagine their consequences as far into the future as permitted by the reader’s patience. Much futurology, during the past fifty years, has been developed as science fiction. With fiction, it is understood that readers will suspend disbelief, allowing authors to develop their ideas relatively unhindered by immediate factual criticism. More recently, the popular imagination has been captured by overt futurology, like Ray Kurzweil’s The Age of Spiritual Machines. Kurzweil presents his predictions of the developments in Artificial Intelligence for the next century as factual. Judging by recent coverage in the magazine, Technology Review, they are taken that way. However, despite its many stimulating qualities, this book requires as much suspension of disbelief as many a work of fiction; or rather—and this is perhaps the main point—it draws upon, and promotes, faith in the ultimate triumph of Artificial Intelligence.
The principle of development Kurzweil adopts is that the power of computers will continue to grow exponentially at the current rate (usually summarized by "Moore’s Law") for the next century. Now, extrapolations based on assumptions of continued exponential growth have largely been discredited in many other fields. So why should computer technology be different? Kurzweil’s main argument is presented as a proposed "law" of nature.
The Law of Time and Chaos: In a process, the time interval between salient events … expands or contracts along with the amount of chaos.
One problem here is that Kurzweil’s vague and shifting use of the word "chaos" leaves the reader unable reliably to concretize this law. But he helps us with two "corollaries:" (1) a "Law of Increasing Chaos," which purports to explain processes that slow down—his primary example being the universe; and (2) "The Law of Accelerating Returns," that is, "as order exponentially increases, time exponentially speeds up." Kurzweil regards this second law as explaining what he sees as an acceleration in evolution, of which computer technology is the latest form. He says: "The Law of Accelerating Returns applies equally to the evolutionary process of computation, which inherently will grow exponentially and without limit." How the undoubted fundamental limits of miniaturization are to be circumvented is left as an open question, but beyond a brief allusion to use of three-dimensional chips and longer discussions of nanotechnology and quantum computing, Kurzweil feels no need to demonstrate that there is a way to sustain the exponential growth. For him, it is a proven consequence of his Law of Accelerating Returns. Eventually the required innovations "will come from the machines themselves."
Moore’s law, that the speed (and capacity) of digital computers doubles approximately every 1.5 years, has had an impressive run since it was first proposed in 1965. These thirty-five years have been the age of integrated circuits. Kurzweil argues that the exponential growth has been occurring much longer, since about 1900. His supporting data includes mechanical, electromechanical, and vacuum-tube digital computers. Within those individual technologies, the data actually shows little growth as a function of time in computing speed per dollar, but the improvements from technology to technology can be fitted by a curve of slowly increasing exponential slope.
The data becomes far less convincing, though, if one includes analog computing devices. Examples might include special-purpose computers such as the steam "governor" or a host of astronomical instruments and models, for example, the astrolabe as well as general purpose electronic analog computers or the lowly slide-rule. Analog computers were quite widely used in the mid-twentieth century to solve complicated, ordinary differential equations. They solved problems at speeds that would require a digital computer capable of roughly 100 kflops (per $1000), approximately 105 times higher than Kurzweil’s fit to digital capabilities at that time. Kurzweil is inconsistent when he ignores such points and restricts his data to purely digital computation, because his hopes for future computational gains are based in part on non-digital algorithms. For example, he discusses specifically neural nets that can be approximated by digital computers, but which are inherently analog computers, though highly nonlinear. So only by arbitrary selection of the data does he obtain his result.
Perhaps Kurzweil recognizes that many of his readers are likely to be unpersuaded by his pseudo- scientific "laws." So he offers us his expert credentials as a further justification of his authority. He recounts a brief history of his outstanding entrepreneurial career—starting several, successful small businesses based on emerging computer technology. The book’s fly leaf also lists some of his prizes: Outstanding Computer Science Book of 1990; Dickinson Prize, 1994; and MIT Inventor of the Year, 1988. Very impressive. The partial list he provides of fulfilled predictions from his 1990 award-winning book is somewhat less impressive, however, both because he does not provide the full scorecard (how many were wrong as well as right) and because some of the successes require substantial interpretation or actually have yet to be fulfilled.
Perhaps one ought rather to assess the reliability of Kurzweil’s future predictions on the basis of the accuracy of his historical and scientific reporting and the logic and consistency of his arguments. Such an assessment would probably be less favorable. Examples of historical and scientific misrepresentations include:
"The British government organized … under the intellectual leadership of Alan Turing, with the mission of cracking the German military code … the cracking of Enigma was enough to enable the Royal Air Force to win the Battle of Britain" (pp. 67–8). [The early Enigma cipher used by the Luftwaffe was cracked by Polish cryptanalysts in 1933, and an Enigma machine constructed by them was provided to Britain at the start of the war. So the role of the Bletchley Park team in the Battle of Britain was mostly deciphering and translating messages sent using this already cracked code. In itself, this was a critical contribution to the war effort. The U-boats used a more secure version of Enigma. It was cracked by the Bletchley team but not until 1943 after an Enigma cipher machine was recovered intact from a U-boat—by the Royal Navy, not the U.S., as portrayed recently by Hollywood!]
"The physical world only bothers to manifest itself in an unambiguous state when one of us conscious entities decides to interact with it" (p. 114). [Recent research on Quantum Decoherence largely disproves this extreme interpretation of quantum physics.]
Examples of logical non-sequiturs or self-contradictions include:
"The human brain presumably follows the laws of physics, so it must be a machine" (p. 5). [By this argument anything in the universe is a machine.]
"The establishment of basic body plans of modern animals in the ‘Cambrian Explosion’ … allowed evolution to concentrate on higher-level features such as expanded brain functions" (p. 35). [Attributing intentionality to a sort of personalized evolution—as if it said, "This is finished, now we can move on to the next task"—is logically contrary to the entire Darwinian program.]
"Regardless of the nature and derivation of a mental experience, spiritual or otherwise, once we have access to the computational processes that give rise to it, we have the opportunity to understand its neurological correlates" (p. 151). [If we assume that computational processes give rise to a mental experience, we have already assumed its nature and derivation.]
One should also note the highly debatable assumptions inherent in all of the book, for example, that minds are nothing but elaborate computer programs embodied in biological brains, and that minds can therefore be abstracted, read out by a scanner, and re-embodied in other "hardware." Of course, this assumption is a staple of science fiction, but it is hardly an accepted fact. Or again, Kurzweil takes as inescapable the Technological Imperative—what can be done, will. This is possibly a plausible view, but certainly a chilling one.
The oft-told illustration of exponentiation recounts the servant asking his reward from the emperor in the form of a chess board with one rice grain on the first square, two on the second, four on the third, eight on the fourth and so on. This seemingly modest request turns out to be impossible to fulfil because of the gigantic numbers that the series leads to. The natural lesson is that all exponential growth curves must saturate at some point because the resources to sustain the growth are exhausted. Kurzweil tells this story and notes that the early doublings are quite manageable and it was only "as they headed into the second half of the chess board that at least one of them [servant or Emperor] got into trouble." By analogy, digital computers have experienced about thirty-two doublings of speed and capacity and "we are heading into the second half of the chessboard." Rather than draw the natural conclusion that the exponential growth must stop sometime in the not too distant future, Kurzweil prefers to imply that the wonders of artificial intelligence will be the result. He says: "This is where things start to get interesting." The perverseness of this conclusion, like much of the book, reveals more about the world view of artificial intelligence enthusiasts like Kurzweil than it does about what will actually happen with computing in the next century.
There will undoubtedly be further large cumulative increases in computing power before the exponential curve saturates. Also, many fascinating and useful applications will arise that draw on this increased capacity. It may be that computers will pass the Turing test and come to be perceived as possessing intelligence. But, regardless of whether or not this happens in the next quarter century, as Kurzweil predicts, I predict that the artificial intelligence enthusiasts will be as unshaken in their convictions as they were by the disappointment of their early artificial intelligence expectations. The reason is that their convictions rest not so much on sober scientific or historical analysis as on faith in their particular, highly reductionist world view. This faith is, I suspect, as resistant to critical argument as the most fundamentalist religion. Given the ascendancy of computer technology at present, many people believe this faith is justified. But faith it is, not scientific fact.
In the end, then, The Age of Spiritual Machines is best seen not as factual or fictional but as a work of popular apologetics supporting the artificial intelligence faith. For all its faults, it is an engaging exposition of that faith, complete with heroes and martyrs, mysticism and humanitarianism, pride and prophecy. Time will tell whether there is any substance to its more apocalyptic visions.
In the meantime, the Christian church badly needs to do its homework. Worked-out theological understanding of the meaning of intelligence, the spiritual significance of apparently intelligent machines (if and when they are built), and the ethical dilemmas that may arise in that situation are almost non-existent. Perhaps it all seems too hypothetical to attract the interest of orthodox Christian theologians. Perhaps the Christian professionals in the computer science field are already stretched too thin to sustain this discussion. Perhaps the Lord will return before this all becomes important. But perhaps not. In any case, we need a response today to give to the artificial intelligence apologists that is more than merely incredulity. Much of what they predict is surely fantasy, and will not come to pass. But many of their imaginings are already in the process of becoming practical. How shall we answer the artificial intelligence apologists? And when it comes to the significance and ethical challenges of machine intelligence, how are we Christians to welcome the good and oppose the evil?