In the Age of Spiritual Machines (Review)
Felix Stalder
© by Computers & Society Vol.30 Nr. 4 [Dec. 2000]
Homepage ||| Contact
Kurzweil, Ray: The Age of Spiritual Machines: When Computers Exceed Human
Intelligence. Viking: New York. 1999. pp.388
ISBN: 0-670-88217-8This book is as annoying as it is interesting. In this combination it might be a better indicator of the strange world view prevalent among certain computer scientists, particularly the proponents of so-called strong artificial intelligence (AI), than of the future it purports to forecast.
Let's start with the interesting part. Kurzweil forecasts three technical developments to convergence, leading to the creation of inexpensive computers that are intelligent in the way most humans are. First, around 2020, $1000 dollars will buy a computer that will have roughly the same processing power as the human brain. The underlying equation, though, surprisingly simplistic. Roughly, the human brain consists of 100 billion neurons linked into 100 trillion connections, each capable of a simultaneous "calculation", i.e. transmitting or blocking an impulse. Each connection is capable of 200 such calculations per second, hence the total processing power of the human brain is about 20 million billion calculations per second. A recent desktop computer makes about 1 billion calculations per second (1 gigaflop). Kurzweil argues that processing power will continue to increase at the same rate than it has in the last few decades (Moore's Law). This equation rests on the underlying assumptions that the two types of "processing power", one based on neurons and the other based on silicon, are by and large equivalent and, second, that the processing power is the single most important variable of intelligence. Both of these are heavily contested, even within AI.[1]
Second, advances in non-invasive brain scan technologies (magnetic resonance imaging (MRI) and optical imaging) will increase our knowledge of how the brain works, that is, which of the 100 trillion connections are activated to create a particular mental process, for example, the visual recognition of an object, or, the storage of a memory. It is already possible to use scanners to distinguish between the sets of neurons responsible for the perception of depth, shape and colour. Ultimately, it will be possible to create a complete model of the particular configuration of an individual human brain based on a precise mapping of all connections among all neurons, and the calculations they produce. There are two scenarios for the use of such detailed scans within computer science. They can be used to design simulated neural nets that operate similarly to the human brain, or, and this is a bit more far fetched, "but also ultimately feasible" (p.124), to re-create or "download" individual brains, including the content of its memory, on a sufficiently fast computer.
The third development stems from advances in artificial intelligence, particularly in the field of neural networks, evolutionary algorithms, and other approaches based on theories of self-organization. The underlying idea is the following: We are able to do a lot of things that we do not understand fully. The classic example is how we catch a flying ball. We do not know the objective variables of the flying ball (speed, angle, weight, wind, etc.) but we possess enough subjective experience to continually adjust our movements to those of the ball, so that we end up standing in the right spot to catch it. We are not born with this ability, we have to acquire it. However, we do not do this in the physics class, but by training on the playing field. In a similar way, a computer can be trained, through repeated feedback that reinforces successful decision-making strategies and weeds out others. In this accelerated evolutionary process, complex patterns can be made to emerge. Their particular configurations are not programmed in a top-down fashion (as in a traditional expert system) but are rather bottom-up induced. Many Wall Street firms manage their portfolios based on recommendations from such artificial intelligence systems. According to Kurzweil, the combined effect of these three developments will be ubiquitous computers with human-like, or even superior, intelligence. It will be increasingly difficult to define clear boundaries between humans and machines. On the one hand, computer implants will become more common, augmenting everything from sight, to hearing to memory, replacing failing organs and extending the life span of human beings. On the other hand, intelligent machines, baffling even their creators, will become more and more similar to human beings in their mental complexity, a similarity that will be enhanced by the advances of virtual reality technology. Kurzweil's sketches of the technical development are in the line of, though less outlandish those of, say, Hans Moravec and his space-conquering robots.[2]
Now to the annoying part.
Kurzweil goes to great lengths to explain why all of this happens, or rather, why all of this needs to happen. What he comes up with is what one might call the eternal rule of evolution. Starting with the first fractions of a second after the big bang, Kurzweil draws a straight line into the future, picturing an evolution jumping from one level of complexity to the next: first physics, then chemistry, and later biology. The higher the level of complexity, the faster is the pace of evolution. However, biological evolution is still slow and linear, that is, it cannot break with its own history. To overcome this limitation, evolution, in its drive to accelerate itself, has added another layer of complexity: technology. Why? Because "once life takes hold on a planet, we can consider the emergence of technology as inevitable" (p.77). One wonders how many planets did Kurzweil sample to detect this "law"? Back on planet earth, "evolution has found a way around the computational limitation of neural circuitry [i.e. the human brain]. Cleverly, it has created organisms that in turn invented a computational technology a million times faster than carbon-based neurons" (p.101). In other words, computer scientists are the leading edge of "carbon-based" evolution, fulfilling the destiny of this planet, which has been manifest since the beginning of time.
This is a strange role that Kurzweil casts for himself. He and his colleagues are at the same time merely raw material for an almighty evolution, no different from amoebae cleverly designed to overcome the limits of chemistry, and, at the same time, god-like fulfillers of destiny and creators of the future. One does not need to be David Noble [3] to see strong religious themes sweeping through the book. "We will be software, not hardware" (p.129) could have been uttered by an ecstatic prophet announcing: The final judgment cometh and thou shalt be spirit, not flesh. Interestingly enough, though, in this contemporary eschatological vision, there is no day of reckoning. In best engineering tradition, each problem will be solved, separately, one by one, though in a very quick pace. For Kurzweil, paradise is only thirty years away when "the basic necessities of food, shelter and security are available for the vast majority of the human population" (p.222), and, "a variety of neural implant technology has essentially eliminated the handicaps associated with most disabilities" (p.221). In light of this bright future, any critique can only be irrational, misguided Luddism whose impact, however, "is limited by the level of prosperity made possible by the new technology" (p.196). The book is characterized by the stark contrast between a highly sophisticated technological vision and a rather obtuse, reductionist and ultimately naive social vision of the use of these technologies. This would be unproblematic were its aims purely technical. But they are much broader. Kurzweil strives "to reflect on the gradual, yet inevitable, emergence of true competition to the full range of human though in order to comprehend the world that lies ahead" (p.6). Particularly unsettling is the insistence on the evolutionary necessity and inevitability of computer technology which seems motivated, consciously or unconsciously, by the desire to claim credit and, at the same time, disclaim responsibility.
[1] Lanier, Jaron (2000). One Half Of A Manifesto. The Edge (September 25) [28.09.2000]
[2] Moravec, Hans (1999). Robot: Mere Machine to Transcendent Mind. New York, Oxford: Oxford University Press
[3] Noble, David F. (1997). The Religion of Technology - The Divinity of Man and the Spirit of Invention. New York: Alfred A. Knopf
End of file