A Review of Robot: Mere Machine to Transcendent Mind

lar escalation of the estimates of the complexity of human intelligence. With regard to the first caveat, Moravec’s estimates of animal equivalence are based solely on hardware complexity. It is often the case that hardware alone cannot deliver performance, but it also requires software sufficient to the task. Thus, projecting “human equivalence” in computer hardware does not mean that “human equivalent software” will be ready to run on the hardware. With regard to the second caveat, neuroscientists regularly adjust upward the estimate of human brain complexity. This upward estimation has the effect of extending the estimated time at which computHans Moravec’s new book, Robot: Mere Machine to Transcendent Mind, will delight readers familiar with this work. Moravec’s strengths—his insightful data analysis, extrapolation of technology to extreme conclusions, and provocative predictions—are all here and will probably gain him some new fans. His insightful data analysis, for example, includes plots of computing trends of the last century, comparing mechanical, electromechanical, and electronic computing capabilities in terms of computational power per unit cost (MIPS per $1000 in 1998 U.S. dollars) across time. This summarizes the evolution of computer power per unit of cost. Readers of his previous book, Mind Children (Harvard University Press, 1988) will recognize the current graph as a refinement of one that appeared in the earlier book. Three things become clear from this refined analysis: (1) computing progress has been tightly coupled around a trend line; (2) this makes predictions of future technologies reasonable by extrapolation; and (3) there’s been an interesting acceleration over the last 30 years, even on this logarithmic scale. The last point is a new addition to this. These data allow predictions about when computers will cross the boundaries of complexity equivalent to a lizard (approximately the years 2002–2008), a mouse (approximately the years 2008–2017), a monkey (approximately the years 2013–2028), and a human (approximately the years 2020– 2030+). There are, however a few caveats to consider: (1) these are hardware estimates and (2) there is a reguin this graph is just a foretaste of the degree to which Moravec will go in predicting the future. This book traces robotic history, projects the field’s immediate future, and travels well beyond. Without giving away the ending, suffice it to say that in this book, the reader will be treated to such ideas as, “What is reality anyway?” (p. 195). Another idea is that