920728 Counting Your Failed Simulations Before They've Crashed A.E. Wilder Smith, in 'Th

Master Index Current Directory Index Go to SkepticTank Go to Human Rights activist Keith Henson Go to Scientology cult

Skeptic Tank!

920728 Counting Your Failed Simulations Before They've Crashed A.E. Wilder Smith, in "The Creation of Life: A Cybernetic Approach to Evolution", says: Thus, Paley's work was destroyed by a theory for which the experimental evidence was -- and still is -- lacking. Darwinism survived simply because it was difficult to disprove and because it neatly and conveniently destroyed the divine hypothesis to which intellectuals were unwilling to submit. THE "SUPER-COMPUTERS" It is only in recent years, with the advent of the "super-computer" which could automatically, swiftly and surely deal with the astronomical numbers in which Darwin enshrouded his theory, that the denouement of this grand scheme became possible. The astronomical numbers of random changes, the long time spans and the alleged evolutionary "trends" in the midst of randomness have been programmed and fed into super-computers. The result has been dramatic, for the machines jam in their efforts to unravel such tangled masses of informational "noise." No wonder that the mathematical experts have crowded around the site of these experiments just as physicians crowd around the bed of a patient sick with a rare disease, to ascertain the cause of the excitement. The biologists have mocked from a distance and denied the result proclaimed by the mathematicians -- that the theory will not work but merely jams the best machines. [pp. 232-233] [End quote] Wilder Smith does not give any references for where these "experiments" took place, who conducted them, or even any reasonably technical description of the results. (No self-respecting computer scientist would leave the results as, "It jams the machine." You would be told, "There was a segmentation fault recorded at address 634592," or something like that.) Wilder Smith later claims that this experiment or experiments reinstates Paley's "argument from design". However, David Hume knocked off this argument long before computers were available. I saw no refutation of Hume's arguments here, so Wilder Smith was a bit premature in his rehabilitation of Paley. But even if we ignore Hume, and more recently Dawkins, Wilder Smith was premature for another reason: natural selection has been adapted to computers and is being used on platforms ranging from mainframes down to personal computers. The whole field of genetic algorithms remained to unfold in the future at the time that Wilder Smith wrote the above quoted lines, which tells us that it is always proper to reserve the conclusions of science to the domain of science, which is natural mechanisms of operation. Appended here is a list of references collated by Mike Rudnick, which is simply papers that address artificial neural networks and genetics algorithms together. Additionally, there are several GA demos and simulation packages available from CNS which do not "always" or even "consistently" do the Wilder Smith "jam" on the computer. [Quote on] This is a bibliography of work relating artificial neural networks (ANNs) and genetic search. The Unix bib format is used throughout. It is organized/oriented for someone familiar with the ANN literature but less familiar with the genetic search literature. It includes some artificial life references. I am a phd candidate in computer science at Oregon Graduate Institute interested in tackling ANN architectural design using genetic search. My particular orientation is to view minimizing interconnections as a central issue, partly motivated by VLSI implementation issues. I have started a mailing list for those interested in using genetic search and ANNs together in various ways. Mail to Neuro-evolution-request@cse.ogi.edu for administrivia, answers to questions, or to have your name added to the list. Much thanks to the people who helped provide this bibliography by providing references. Corrections and additional references are welcome. Mike Rudnick CSnet: rudnick@cse.ogi.edu Computer Science & Eng. Dept. ARPAnet: rudnick%cse.ogi.edu@relay.cs.net Oregon Graduate Institute BITNET: rudnick%cse.ogi.edu@relay.cs.net 19600 N.W. von Neumann Dr. UUCP: {tektronix,verdix}!ogicse!rudnick Beaverton, OR. 97006-1999 (503) 690-1121 X7390 ****************************************************************************** %A David Ackley %T A Connectionist Machine for Genetic Hillclimbing %D 1987 %I Kluwer Academic Publishers %X Ackley's phd thesis. Compartive study between Ackley's fairly baroque "connectionist" learning architecture and several alternatives which include genetic search and random. %A David H. Ackley %A Michael L. Littman %T Learning from natural selection in an artificial environment %A J. M. Baldwin %T A new factor in evolution %J American Naturalist %D 1896 %V 30 ***** continued in next message *****


E-Mail Fredric L. Rice / The Skeptic Tank