Review: Mind Children

treatment for people over 70, even though their taxes have paid for the care.

5. And as life is extended, there will be an ever greater gap between the birth rate and the death rate. So there will be ever more people on this planet. I think that this planet is already overcrowded. To see it more so, with more people crawling over it than there are worms on an apple, is a horrifying prospect. I would like to reduce the population to what it was before 1900, when people could at least have a bit of living space, though there is no way to do that without killing lots of people, and that of course is out of the question. But we have to avoid increased numbers in the future; and having a birth rate that doubles the death rate is NOT a way to do that!

Anyway, good luck with your new venture.

Best Wishes,

John Hospers

(Editors’ note: Dr. Hospers is a professor emeritus of philosophy at USC, author of several books and countless articles on subjects such as aesthetics and political philosophy, and in 1972 became the first presidential candidate to run on the Libertarian Party ticket. His points are well taken, but we’ll respond to them in upcoming articles. We deal with points 3,4, and 5 in this issue’s articles on supercomputing and nanotechnology.)

Hans Moravec -

Mind Children: The Future of Robot and Human Intelligence.

Harvard University Press 1988.

Reviewed by Max O’Connor

This is a book that every extropian not only should read but would enjoy immensely. It is packed with fantastic, imaginative, wild ideas which are almost all solidly based in real science. The ideas in the book revolve around the coming massive boost in intelligence resulting from advanced artificial intelligence (AI) and an upbeat, optimistic view of our future. It is impossible to do justice to all the invigorating ideas and style of presentation here but I will try to convey some impression of the treasures to be found in the book.

EXTROPY

#2 - Winter Issue, 1989

5

The first chapter sets the stage with a survey of the history and current state of robotics and AI. In the second chapter, ‘powering up’, Moravec forcefully demonstrates the growth in computational power and speed this century and convincingly argues that we already know that it can continue its onward charge without slowing down for decades to come. Careful calculations of the computational of the human brain (based on the retina’s processing capabilities) are made to give us a benchmark by which to compare computing power. According to Moravec the retina’s processing can be matched by one billion computer calculations per second. The brain has 1,000 times as many neurons as the retina but they are packed into a volume 100,000 times as large. A guesstimated calculation suggests that (using a compromise value of 10,000) the human brain could be matched in processing power by a computer performing 10 trillion ($10^{13}$) operations per second (teraops). This is 1 million times faster than the machines used to run robots now and 1,000 times faster than the most powerful supercomputers around. If the standard processing-to-memory ratio continues to hold our humanlike computer would require about 10 trillion words of memory, or about $10^{15}$ bits (computer words are between 16 and 64 bits long). It is only an estimate, although one based on

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Since 1900 there has been a trillion-fold increase in the amount of computational power that a dollar will buy.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Moravec has a beautiful and exciting chart on p.64 of computing power from 1900 to the present and projected onward at the same, constant rate until 2030. Since 1900 there has been a trillionfold increase in the amount of computation a dollar will buy. At the same rate of increase the 10 teraops needed for a human equivalent computer will be available before 2010 for $10 million, and in a $1,000 PC by 2030. A cautionary note - Moravec’s calculations are only for advances in hardware; we don’t know whether the ability to program (or teach to self-program) a computer to work like a human brain will come about as

what is known of the brain, the $10^{15}$ bit memory of a humanlike computer should be comparable to the information encoded in the $10^{14}$ synapses of a human biological brain. From these calculations we can say that the most powerful supercomputers (costing $10 million dollars) of the late 1980s are a match only for the one gram brain of a mouse. This is quite a testament to evolution. How long before the humanlike computer will be possible?

EXTROPY

#2 - Winter Issue, 1989

6

quickly. Moravec does give plenty of examples of current research and research paths that show how vast advances are still ahead of us. To take us out on the far reaches of ∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

possibility Moravec mentions nanotechnology and the plans for nanocomputers by Eric Drexler and Conrad Schneiker. These computers would use small molecules for their parts and so would have switching times a million times faster than the fastest computer components now. Nanocomputing would allow the human-equivalence criterion to be

We can leave our weak, vulnerable, stupid bodies and brains for something enormously superior.

exceeded a millionfold! A true extropian, Moravec speculates further: Recent physical theories involving superstrings suggest that, in principle, ultradense matter would make possible switching times much faster than the speed of light. Moravec speculates that someday, our non-human progeny may exploit the ultradense matter of collapsed white dwarfs and neutron stars to build machines with a million million million million (10³⁰) times the power of a human mind. I intend to be around long enough to become one of those creatures.

Chapter 3, ‘Symbiosis’, discusses the trend towards ease of use of computers, with better menus, icons and so on. Much research is underway in trying to make computers more interactive with humans. Moravec tells us about early versions of ‘magic glasses’ (leading to an entire ‘magic wardrobe’) which will be able to create any artificial environment of sight, sound, and touch. The consequences of such a technology will be the subject of an article in a forthcoming issue of Extropy.

The fourth chapter - ‘Grandfather Clause’ - takes us into areas with which many readers of this magazine will be familiar. The discussion is set up with a description of a robot ‘bush’ with a trillion fingers, each able to move a million times per second. Just one possible use of such a device would be surgery at the cellular level. Moravec shows his extropic, pro-survival orientation in not being happy at the prospect that we will be left behind as our vastly superior progeny explore the inner and outer universe. Fortunately, for those of us with the desire and guts to do it, that need not be our fate. Moravec presents us a particularly compelling scenario in which a robot surgeon maps your still-functioning brain layer by layer as it writes a program that models the behavior of the brain tissue. At the end of the process, you brain has gone but you (the information that is your identity) resides in a computer brain which is

EXTROPY

#2 - Winter Issue, 1989

7

housed in a body of your choice. This is only one possible method of technological transmigration but the point is that we can leave our weak, vulnerable, stupid bodies and brains for something enormously superior. Once in the new computer-brain you could flip a switch to speed up your thinking by a factor of a thousand or more. Moravec talks of many other changes that would then be open to us, from retarding the onset of boredom so that we could fully use the immense new speed and computational power, to making copies of yourself. Copies may be made to ensure against catastrophic accident resulting in total destruction of a unit. Merging with other people or with copies of yourself that have spent some time accumulating independent experiences would also be options. Taking up temporary residence in bodies of animals such as whales might be an entertaining activity, and matter transportation would be possible although, rather than transmitting matter they would actually transmit the information necessary to reconstitute you at the other end.

A wide-ranging discussion in the chapter on ‘Wildlife’ brings in present and computer viruses and provides an intriguing warning for SETI advocates about possible parasitic computer or information viruses from space. After having got the reader alarmed by the worrying possibilities that come with the virtually unlimited new power, Moravec ends the chapter on a highly optimistic note. He discusses Robert Axelrod’s ideas from his (highly recommended) book The Evolution of Cooperation in which it is demonstrated that rational, self-interested agents will cooperate and be ‘nice’ to each other because this is their best strategy. Axelrod found that, out of a large number of strategies submitted by many people, the most effective one was the ‘tit-for-tat’ strategy. This involves always cooperating with others for mutual advantage when they do not cheat, never cheating yourself, punishing cheaters, and not holding grudges after the punishment has been given. Cooperation and absence of hostility is boosted to the extent that agents understand the long-term consequences of their actions, remember past interactions, and understand the workings of the world around them. This leads to the extremely extropic and optimistic likelihood that a universe populated by incredibly intelligent, unforgetting agents pursuing their own interests will be one of maximal peace, harmony, and productivity.

The final chapter, ‘Breakout’, is the most far out (yet well-connected to reality) and so hardest to describe. I urge the reader to go out and discover Moravec’s ideas in their full glory for him or herself. Speculations are presented as to how we can experience an infinity of subjective experience and thinking in a universe which will have a final end, and, more radically, how we may eventually overthrow the tyranny of the ‘laws’ of nature and set our own terms. More speculation appears in the third appendix on ‘The Outer Limits of Computation’. Here we are told

EXTROPY

#2 - Winter Issue, 1989

8