Wednesday, September 28, 2005

Dissecting terrorism

Ed Lazowska, Steve Maurer, Stefan Savage, and Geoff Voelker are once again running a multi-institutional course on public policy and technology. This term's focus is on cybersecurity, but the first few weeks have basically focused on terrorism and security more generally.

Just like last year's offering, there's an online lecture archive with PowerPoints and videos, or synchronized video/slides using WebViewer.

Anyway, I just got home from this evening's lecture (which isn't entirely online yet, but will be soon), and it bears watching. Delivered tag-team style by Gary Ackerman and Jeffrey Bale of the Center for Nonproliferation Studies, the lecture was an unusually clear-eyed and (relatively) apolitical look at the definition of terrorism, the notion of "weapons of mass destruction", the motivations and deterrability of various styles of terrorist groups, and the likelihood of the various kinds and scales of terrorists attacks using non-conventional weapons. The lecture was sobering and occasionally darkly comic, as such discussions tend to be --- there's something inherently Strangelovian about weighing the probabilities of greater or lesser disasters. The overall message I came away with was: preventing devastating terrorist attacks is vastly complex, and can't be reduced to a simple formula.

Which makes me pretty worried, because America's current political discourse ain't exactly nimble when grappling with complicated issues.

Wednesday, September 21, 2005

Authors Guild demonstrates technological illiteracy

The Authors Guild is suing Google for attempting to innova--- er, for copyright infringement. If you've been reading this blog for a while, you already know my opinion on the subject. I hereby call for all Authors Guild members either to denounce the actions of their organization, or to stop using all web search engines, or confess to being complete hypocrites. Those are the only options.


UPDATE 22 Sept.: As usual, Lawrence Lessig says it better than I could.

Tuesday, September 20, 2005

A brief note on "free will"

Every conceivable behavior is either deterministic, or random, or some combination of the two. Neither determinism nor randomness can possibly qualify as free will --- intuitively, neither a clock's gears nor a pair of dice possess free will --- and therefore the very concept of free will is philosophically incoherent.

To elaborate slightly, behavior either follows a set of predictable rules, or it does not. If behavior follows a set of predictable rules, then those rules constitute a deterministic algorithm, which clearly lacks free will. If behavior does not follow a set of predictable rules, then it is a random process, which clearly also lacks free will. In practice, most nontrivial behaviors must be modeled by a combination of deterministic and random elements, but combining the two into a randomized algorithm doesn't produce free will either.

Incidentally, all known physical processes --- or, in other words, all possible computing devices in the universe, including human neurons --- can be modeled using some combination of deterministic and stochastic processes. So, even if free will could conceivably exist, actual human beings certainly would not possess it. However, I believe that the very concept of free will is incoherent, an argument that does not depend on properties of our physical universe.

If you disagree with this analysis, it can only be because you misunderstand algorithms, or you misunderstand randomness, or you do not understand how these two combine, or you are defining free will in some tendentious fashion. This seems to be the case, for example, for most of the philosophical positions described in the Wikipedia entry on free will. Notably, none of the pro-free-will arguments that I've ever seen manage to distinguish human decision-making from, say, a program that repeatedly generates a random array, and then sorts that array using an implementation of quicksort that selects a random pivot on each recursion. If you believe that quicksort with a random pivot possesses free will, I suggest that you've redefined free will away to the point of uselessness.

Now, given that free will is philosophically incoherent, I'm not surprised that definitions of free will turn out to be useless, but still, it's sort of sad. Just give up already!

Tuesday, September 13, 2005

Two days of tribulation

Programming language research groups around the globe are currently gnashing their teeth and grumbling, for today and tomorrow comprise the "author response period" for POPL'06, the world's foremost theoretical programming language conference.

For those who don't know what this means, early this morning (or, more precisely, 19:00 12 Sept, Samoa time) the reviews for all submitted papers were made available to the authors, who have 48 hours to submit a 500-word response. This is an innovation recently introduced to the top ACM SIGPLAN conferences --- actually, it started at PLDI of the year my advisor was program chair --- and most agree that it has been a success. Formerly, authors had no opportunity to reply at all, and only received reviews after the program committee had met, discussed the papers, and made a final decision to accept or reject.

Of course, given the great personal investment of authors in their work, author responses could have turned into an occasion for endless counter-sniping tirades. So, a couple of safeguards were built into the process. First, software strictly enforces the 500-word limit on responses. Second, authors are cautioned to avoid subjective ranting, and address only factual errors or questions posed by reviewers. From what I hear, these restrictions have mostly kept author responses useful.

But oh, a mere 500 words! How dearly researchers would like to discourse at length, correcting all the mistaken impressions, giving comprehensive and intuitive answers to the subtler questions, and really slamming down that one reviewer who was clearly (clearly!) smoking something illegal while writing their review.

In the unlikely event that any program committee members are reading this, I speak, of course, purely in hypothetical terms. Any resemblance to reviews that I, personally, have received for POPL or any other conference is purely coincidental.

Except for that one review...

Just kidding. In all seriousness, if you're curious, I'm actually pleasantly surprised with the reviews for our POPL'06 submission. It's still not likely to be accepted, but POPL's the toughest conference in programming languages, with an accept rate around 15%, so rejection's just the default situation. As far as I'm concerned, we've already beat the spread on this one, even if we don't pull out a victory.

Monday, September 12, 2005

A useful feature that no digital camera will ever have

CompactFlash cards have reached ungodly sizes lately. This seems great at first --- shoot all day on a single card! --- but recently I discovered the downside: it's all-too-easy to lose an entire day's worth of photos if your card goes south.

This happened to me on a recent visit to the Denver Aquarium (incidentally, for a city that's a thousand miles inland and a mile above sea level, Denver has a shockingly well-stocked aquarium). I shot a hundred-odd photos and about a dozen video clips. Then, without warning, the card got corrupted so thoroughly that my camera actually began acting wonky, taking more than a minute to power up and power down and making it impossible to navigate the menus. The card wouldn't even mount as a drive on my USB card reader, rendering software-based disk reformatting and recovery tools useless. There are recovery services, which charge an arm and a leg to crack open your card and poke the hardware with circuit testers and such --- a mere $154 to recover a 512MB card! --- but it would almost be cheaper to buy another plane ticket to Denver and re-shoot the aquarium using a fresh card. I'll just eat the loss, I suppose, and be grateful that it wasn't my Rocky Mountain National Park pictures that got destroyed (now those would be irreplaceable).

I could carry multiple smaller cards, but I don't want to be bothered switching between memory cards during the day; plus, if one of the cards got misplaced or corrupted, I'd still lose a half day's worth of pictures.

Which brings me to my point. I'd much rather have a camera with two CompactFlash slots, and have it write one copy of each file on each card.

Of course, no digital camera maker will ever, ever sell a camera with redundant storage like this. Given how small and cheap cameras have become, this would add a lot of bulk and cost to the camera. I've experienced firsthand the frustration of losing my pictures, so I wouldn't mind, but only a vanishingly small fraction of consumers would pay for this feature.

What might happen in the future is that flash memory manufacturers and camera makers might get together and decide to make individual cards more reliable, by trading storage space for reliability. CompactFlash, like all computer storage media, contains some measures to protect against minor forms of corruption, but they could go a lot further in this respect.

In the meantime, I wonder if there are any fast, portable card-to-card duplicators out there, so that I could manually take backups of my photos while shooting "in the field"...