BKAGSPMC.RVW 991010 "The Age of Spiritual Machines", Ray Kurzweil, 1999, 0-670-88217-8, C$34.99 %A Ray Kurzweil raymond@kurzweiltech.com %C 10 Alcorn Ave, Suite 300, Toronto, Ontario, M4V 3B2 %D 1999 %G 0-670-88217-8 %I Viking/Penguin/Signet/Roc %O C$34.99 416-925-2249 Fax: 416-925-0068 service@penguin.ca %P 388 p. %T "The Age of Spiritual Machines: When Computers Exceed Human Intelligence" In the beginning was optimism, and the optimism was not quite justified. Then came a great deal of experimentation in the real world, and the dawning realization that there were a few things about intelligence that we hadn't learned yet. In fact, I remember one comment to the effect that, after all of this work, we were farther from building an intelligent computer now, than when we started. The field of artificial intelligence is obviously enjoying renewed confidence. Buoyed by successes in areas such as chess (the very best computers can now beat Grandmasters) and speech recognition (top end personal computers can now transcribe speech faster than some people can type), and with a few decades of hard won experience behind them, leaders in the field are now willing to predict that usefully intelligent machines are achievable within the foreseeable future. Human levels of intelligence are predicted on the basis of ever increasing processor speeds, and, given the apparent means of getting around predictable limits on that increase, intelligence beyond current human levels is seen as inevitable. Kurzweil is no starry-eyed futurist, but his prologue does rely on some simplifying assumptions. Models of the human brain are woefully incomplete. A very large neural net may duplicate a number of functions, but even now we know that there are purely chemical bases for aspects of memory, at the very least. In addition, while computers may be able to transcribe written material, the process of "understanding" text will undoubtedly involve comparison against an enormous base of prior data, and may take a lot longer to prepare and communicate than is the case with current file transfers. Chapter one (and, basically, chapter two) posits some superficially interesting "laws" about time and evolution, but there is a nagging feeling that these laws rely on the universe caring whether we exist or not. The analysis of mind and consciousness, in chapter three, is intended to prove that machines can be intelligent, but really only indicates that we do not know what intelligence is. In addition, an early discussion, reversing the "when does a machine become human" question, passes off a very vital point as being trivial. At one point the memory of an individual is replaced with a "perfect" machine memory, and the person is said to be unchanged, even though Kurzweil himself points out that there would be a very new awareness of things we would rather have forgotten. There is a grab bag of AI history, research, and algorithms in chapter four. Chapter five is a very terse review of "expert," or knowledge based, systems. Chapter six looks at a few proposals for new generations of computers. (Including quantum computers, the discussions of which always lead to unsettling visions of Douglas Adams' Infinite Improbability Drive and A Really Hot Cup Of Tea.) The look at "bodies" meanders through nanotechnology and virtual reality (with the obligatory detour through cybersex) in chapter seven. Most of chapter eight touches on computer generated art, although most of the examples were rather disappointing in view of other, more ambitious projects that I have known, some going back twenty years. (We are also treated to a resume of Great Things Ray Has Done.) Part three presents "views" of the future at ten, twenty, thirty, and one hundred years distance. Basically, these are just predictions, without backing or justification. There is a great deal of similarity between this work and Hans Moravec's "Robot: Mere Machine to Transcendant Mind" (cf. BKRBTMMT.RVW) (which Kurzweil quotes from). Moravec, however, prepares a solid background from his research into what computers have been able to do up to now, before he extrapolates what they might do in the future. He also sticks to computer activity and behaviour. When Moravec does take off into the far future, he is very clear about what he is doing--and even then, buttresses his fancy with solid research and reasonable theory. Kurzweil, on the other hand, blithely assumes that we will all agree with his prediction of computers duplicating, and then exceeding, human intelligence, a rather loose target at best. As for the spirituality, I think I must have missed it. copyright Robert M. Slade, 1999 BKAGSPMC.RVW 991010