Weekly Summary of Noospheric Echolocutions
This past week I finished my reading of Mandelbrot’s most recent book The (Mis)Behavior of Markets. I actually didn’t like it that much. I found the book to be especially light on details; for a mathematical empiricist Mandelbrot didn’t actually explain, in unambiguous terms, the patterns that he sees in market data. He did a good job of showing that price charts appear roughly the same independent of company and time-interval. (That is, if you forget to label your axes then a chart covering a few years doesn’t look appreciably different from a chart covering only a few seconds.) But the details of how he knows about the multi-fractal patterns, and how to measure such patterns, or even the basic mathematical characteristics of mult-fractals went entirely ignored. Some of these details were given in a reference appendix for each chapter, but not nearly enough to satisfy my avid mind. The editor also did a poor job of citing Mandelbrot’s references to peer-reviewed papers throughout the text; you had to simply follow along in the appendix simultaneously while reading the main text (an annoying usability bug). Instead of discussing such details within the main text, he opted to give generalized overviews with inadequate pictures of fractals that have nothing to do with finance. Now, I know that since I read graduate-level textbooks on Computer Science in my spare time, I’m not the target audience for this book, but I would have expected to see many, many more price charts and elucidating discussion of features within such charts, esp. considering that he’s so big on empirical data. As it was, even when the book had pictures to convey the geometric ideas within the text they came out piss-poor and confusing. Tufte would be appalled.
Mandelbrot also seems to have a healthy dose of self-assurance (deservedly so, for he is quite accomplished). He didn’t fail to mention that one of his students (Fama) formulated the Efficient Market Hypothesis, and that his early paper on cotton prices help to form a pillar of econophysics. He certainly deserves recognition for developing a very important part of mathematics, but I consider his abilities to write for the layperson to be severely lacking.
I’ve also been reading up on Component-Oriented Software Construction. It’s not a panacea that will solve all the industries woes, but I consider it a significant step towards solving many of our current problems. The traditional Object-Oriented inheritance paradigm is fundamentally incompatible with parallelism, and OO in general suffers from systemic rot as shown by the Circle-Ellipse Problem and the Fragile Base Class Problem. The move toward a Component-Oriented architecture is, in my opinion, a beneficial one. But it comes at a high cost; instead of having a (complex and rigid) framework for message passing and method calling, the programmer must handle request forwarding and chaining themselves; functionality is derived through composition instead of inheritance. In my experience, programmers are the most unreliable piece in the software construction business, and relying on humans to successfully adapt their minds (already trained on the OO model) to a new methodology is a mistake. Furthermore, the Component-Oriented message passing system is so flexible (best formulated as a general directed graph) that a programmer could easily get themselves into the opposite of a deadlock (messages getting passed around continuously in a Hamiltonian path). Debugging could be a real nightmare in such a system, and we’d really best develop techniques for tracing and visualizing information flow throughout the program.
I watched a recording of Guy Steele‘s infamous lecture Growing a Language (because of the poor video quality, you really want to read the article as you watch), which he gave at OOPSLA ’98. This was a delightfully instructive video, I recommend that everyone interested in programming language watch it. I also watched a presentation he gave about his newest language, Fortress, because it was mentioned at LtU. I especially like the mathematical nature of the language, and his willingness to go back and try syntax experiments (APL, PL/I) that failed due to requirements of specialized hardware.
I also read Vannevar Bush‘s famous article As we may think, wherein he introduces the memex. I found the proposed interface to the memex to be both antiquated and mechanically cumbersome, when compared to the modern computer. But, given the technology in his day this is to be expected. Some of the issues related to search, recall, and annotation, were unfortunately glossed over; in fact, these same issues still plague us today, everybody consistently underestimates the complexity of the task; reliable associative search is a hard problem. He did have the remarkable foresight to envision how people would wish to use such a system, and, were he alive today, would probably be greatly enamored with the collaborative efforts of knowledge sharing that places such as wikipedia represent.