Administrator
|
I just got a copy of NKS .. which is also available on-line:
http://www.wolframscience.com/nksonline/toc.html We've spoken a bit about this in the past, Roger in particular (attached) had helpful remarks. I find the book irritating so far due to the excessive "I then invented XXX" and that sort of thing. But it certainly has interesting points. Has anyone else looked into the book since then? I'm curious if anyone has found any value there. -- Owen Owen Densmore 908 Camino Santander Santa Fe, NM 87505 Cell: 505-570-0168 Home: 505-988-3787 http://backspaces.net On Aug 28, 2002, at 12:27 PM, Roger Frye wrote: > The most maddening thing about NKS is the lack of attribution of > results. If you read the last line of the paragraph at the top of > column 2 of page 945, you might think that Wolfram had found the > smallest counter-example to Euler's conjecture extending Fermat's last > theorem, but I did. > > I have done extensive work on the Connection Machine studying the > usefulness of Rule 30 as a pseudo-random number generator. I classify > PRNGs as good, bad or ugly. Good ones pass tests and have provable > qualities. Bad ones don't pass tests. Ugly ones like Rule 30 pass the > tests so far applied, but you can't predict anything like the period, > and you can't prevent correlation of parallel streams. Furthermore, > despite Wolfram's assertions, Rule 30 is very inefficient: the number > of operations (even with the universal bit op-code structure of the CM) > required to produce one random value using Rule 30 is much larger than > the number needed with a good PRNG. I modified the Rule 30 PRNG that > Carl Feynman had written under Wolfram's guidance in order to avoid > patterns that I had found which generated short periods. I also added > a > fast, parallelizable PRNG to the CM library. > > I have learned some interesting things from NKS about studying CAs by > considering the additive and renormalization properties of their > operations. I can't tell whether these ideas came from Wolfram, but it > was good to discover them in his book. I have also been provoked to do > experiments in order to try to disprove his results, and at one point I > thought I had, so I wrote to him but later realized that I had misread > the claims. His assistant wrote back with the correction and told me > that Wolfram remembered my name and was glad that I was reading his > book. > > I am less than half way through the book, so I haven't read the more > speculative chapters, but I have read many of the reviews > (http://www.math.usf.edu/~eclark/ANKOS_reviews.html). My impression is > that when a specialist in one of the sciences reviews NKS, that > scientist usually reports that Wolfram's speculations in that field > either are known to be wrong or have been known to be true for some > time. Brian Hayes' choice of the term "delusional history" (American > Scientist, July-August 2002) may sadly be a correct diagnosis. > > -Roger Frye |
Free forum by Nabble | Edit this page |