X-Message-Number: 4310 Date: Thu, 27 Apr 1995 22:31:38 -0700 From: John K Clark <> Subject: SCI.CRYONICS Definitions,Computers and Beyond -----BEGIN PGP SIGNED MESSAGE----- Message #4303 (Thomas Donaldson) Wrote: >I note that he does not actually give a definition >[of a computer] or refer to one. Actually I did, I said a computer was a deterministic Turing machine. I also said that it was a broad definition so if you want to use a narrower one it's fine by me. If you accept my definition that would mean that neural nets are computers but quantum "computers" are probably are not. Richard Feynman and David Deutsch did some early work on this. Seth Lloyd published in the most respected scientific journal in the country a design for a quantum computer, although some still say it won't work, see the September 17 1993 issue of Science page 1569 " A Potentially Realizable Quantum Computer" by Seth Lloyd. Our own Charles Platt has written a popular account of Lloyd's work in the March issue of Wired magazine. Very recently Peter Shore of Bell Labs proved that IF a quantum "computer " is possible and could be built, it could factor numbers in polynomial time. A deterministic Turing Machine might be able to do the same thing if given the proper algorithm, but such an algorithm has not been found and most think it never will be. Roger Penrose thinks that electrons in the micro tubules of brain cells are doing quantum calculations, but few share his opinion. There is no question that as a mathematician and physicist Penrose is absolutely world class, but most biologists think he's a crackpot who shouldn't be poaching in their domain. Personally I doubt if he's correct, but he's certainly no crackpot. Interestingly he finds his most sympathetic audience among computer designers. >I think that some further definitions might sharpen my >question. My main problem with thinking of neural nets as >computers comes from the fact that they aren't programmed >in any traditional sense. There is no special loading >of any sequence of directions to which the machine will >respond. Instead they are TRAINED: **WARNING THE FOLLOWING IS PURE SPECULATION** If we're more like neural net style computers than Von Neuman style computers, as seems likely, perhaps that could explain why we have such difficulty coming up with formal definitions, and in fact seldom need them in everyday life, not even for very familiar concepts. As you have said a neural net does not have a stored program in the usual sense of the word, it's trained, that may be why I can point to red objects but can not give a definition (a program?) of red. The concept of "red" is still meaningful to me even though it lacks a definition, the same is true for "intelligence" and "consciousness". John K Clark -----BEGIN PGP SIGNATURE----- Version: 2.6.i iQCzAgUBL6B7MH03wfSpid95AQF70QTvdVt1aQyQdzq3DsZJSfq03hsUXkX+4Dn+ yJylkRtwMSk0PIehObN+LAvrtByDwa9WIoHZs087/pBeGkykAsY8VtWuJLEU6Gdr wfXhJ1O4YnYorfYmrqgChwCLKM20ObtBtqZVG1Jw6Clm8NMS7sLTMBuEhw2XNEMP 6UsOFiHcB7+xkz+ttGyAsjTAr9I44N/RSejaOnDlEZFr37BjCTw= =0oNj -----END PGP SIGNATURE----- Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=4310