rapture for the geeks

I grabbed this title off the shelf at the Sea-Tac Airport Borders last month. I'm obsessed with the idea of building a program that could post to anonymous political blogs on my behalf, following some instructions (or values) that I feed to it. For this reason, I decided to give Rapture for the Geeks: When AI outsmarts IQ a read.

The book provides an informal history of automation and is well footnoted; a novice wanting to learn more about computing would benefit from reading Rapture. If you're experienced with information technology, you'll benefit as well; you'll find some of the information redundant, however. I'm not a life scientist; much of the info in chapter 4 on the human brain and AI was new for me.

Does the book get to the question from the first paragraph of the book jacket: "how long is it before [computers] are as intelligent as humans?" Not really. The book is authored by Richard Dooling, an attorney and writer, which explains a lot. Attorney...some people think "Barack Obama" or "John Roberts." I think "John Edwards." (Or, to provide an updated example, "Rod Blagojevich.")

Dooling's analysis of Moore's Law is instructive, and illustrates why there's a growing gap between the technical haves and have-nots, both in terms of available infrastructure and technical knowledge/skill. If you're so engaged that you recognize how much you're missing, with the rapid evolution of programming/scripting languages, database models, etc., you're in better shape than most.

I didn't always feel this way towards technology. When I worked in naval nuclear propulsion, the technologies and information in question were tightly controlled. In nuke power school, you couldn't take your books or notes out of the building. And you couldn't take your notes from assignment to assignment (nuke school to prototype; prototype to a sea assignment). By federal law, the naval reactors program is essentially a civilian program regulated by the NRC. There was steady progress in propulsion plants, in extending core life, in building more powerful reactors (enabling two reactor supercarriers), but the pace of change seemed manageable. Not with computing and information technology. The pace of change is radically more rapid, with open source software and ubiquitous broadband the ultimate enablers in this regard.

At numerous points in the book, the author makes snarky comments regarding Microsoft and at one point asserts that Firefox is a better and more secure browser. While I'm using Firefox at this moment, I can't agree with the overall argument, and recent reviews (one example), suggest that browser functionality is closer than he's suggesting. There's no question that much of the exponential growth in software tools is largely a product of open source software; in that sense, Dooling's appreciation for OSS is justified. I'm just tired of the strange things that happen with Firefox (lengthy, unsolicited updates, add-ons that don't work once an upgrade has been performed...).

Regarding Microsoft, here's an example assertion by Dooling: "Microsoft's business model depends on its customers knowing almost nothing about computers..." (207).

The singularity is addressed in chapter 5, led off appropriately with a Vernor Vinge quote. But in the previous chapter, Dooling quotes Jaron Lanier, who provides the best explanation as to why such a transformation won't happen: "software is brittle. If every little thing isn't perfect, it breaks" (81). Even if there's a programmer who wants to "climb the strong AI mountain" and create machines bent out conquest, the code has to work first (119). While I'm a fan of Vernor Vinge and appreciate Dooling's review of Vinge's essays on singularity, I find Lanier's argument more persuasive than Vinge's vision.

Finally, the author notes the urgency of learning a programming language, with post-singularity extermination the possible result of refusing to do so. Dooling continues his anti-Microsoft slant by insisting that only languages supported on all platforms should be considered. Given the wide popularity of Visual Basic and its elevation to first-class language status with the creation of the .NET Framework, it's a pretty narrow view to assert that readers should learn Ruby or Python or PHP or Perl....

Time to wrap this up as I plow through to the end. I'd recommend this book as an introduction to AI and to the future of computing and information technology in general. As I noted above, it's a well-organized work, and I took away many ideas for further reading. Dooling's analysis of operating systems, web browsers, and programming languages is too ideological; this takes something away from his book.