“Speculations on Perceptrons and Other Automata”, 1959-06-02 (; backlinks):
Harold Jeffreys once said that the brain may be an imperfect thinking machine, but is the only one available. For over 2,000 years men have devised tools that aid or to some extent replace thought. Most of these men were philosophers. There were Aristotle’s logic; Boole’s “Laws of Thought”; the theory of probability founded by Cardan, Leibnitz, Pascal, Fermat and others; simple calculating machines designed by Leibnitz and Pascal; and a complicated calculating machine designed by Charles Babbage but never completed, owing to the short-sightedness of the British Treasury.
But it was not until the advent of electronic computers that many scientists began to suspect that robots may be feasible.
The justification for this view is not so much that existing electronic computers have yet been programmed to do much that is very close to thinking, as that electronics provides a very rapid and reliable method of handling information.
…When considering how much should be spent on developing artificial brains, it is necessary to try to guess the probability of various degrees of success. My own guess for the time it will take to develop a really useful artificial brain is 20 years multiplied or divided by 1.5×, if it is done during the next 100 years (which I think is odds on).
An argument against is the large number of neurons in the cortex of the human brain (and the chimpanzee has almost as many). On the other hand, electronic units may work more reliably and perhaps a million times as fast in 10 years’ time, and very small units will exist and will be inexpensive. Moreover, natural evolution is extremely wasteful and blind, whereas artificial selection can be controlled intelligently. It is true that evolution has gone on for a long time and over a wide area, and it could be argued that to make life by chemical means may be easier than to construct an artificial brain.
On the other hand simulation of intelligence on general-purpose computers has already met with some limited success. We can analyse conscious thought by introspection, a method that does not apply to biochemistry, and also does not apply very well to unconscious thought. I understand that some schizophrenics are good at analyzing unconscious thoughts; perhaps, as occupational therapy, they should be given work on the construction of artificial brains.
Once a machine is designed that is good enough, say at a cost of $809,969,618.34$100,000,0001959, it can be put to work designing an even better machine. At this point an “explosion” will clearly occur; all the problems of science and technology will be handed over to machines and it will no longer be necessary for people to work. Whether this will lead to a Utopia or to the extermination of the human race will depend on how the problem is handled by the machines. The important thing will be to give them the aim of serving human beings.
It seems probable that no mechanical brain will be really useful until it is somewhere near to the critical size. If so, there will be only a very short transition period between having no very good machine and having a great many exceedingly good ones. Therefore the work on simulation of artificial intelligence on general-purpose computers is especially important, because it will lengthen the transition period, and give human beings a chance to adapt to the future situation.