New Orleans (day 2 of 7)

So today was almost entirely taken up by SIGGRAPH. Which is reasonable, as it's why I'm here, but it does mean I'm putting in longer hours than I would if I were back in Houston, actually coding.

The day started at 8:30, which is reasonable enough, with a complete waste of time, which wasn't. One of the topics of this year's SIGGRAPH is "Informational Aesthetics" — a perfectly reasonable concept, especially in the context of HCI, but unfortunately this opening talk was... notsomuch, and I'll cover it in detail simply because I'm in an exceedingly lousy mood.

The talk opened with a description of the construction of various visualizations of a graph of the intercitations of articles published in various scientific journals. (An overview is at well-formed.eigenfactor.org, if you're curious.) The speaker referred to this as a "graph of science," which grated on my nerves — use of "science" to refer to the social phenomenon of the interaction of scientists, rather than the abstract process of learning about the world possible in a social vacuum, has never sat well with me. (On a similar note, he referred to the emergence of the discipline of 'neuroscience' — shown on one graph as a recategorization of journals originally from several distinct disciplines (1997) into a distinct cluster not previously extant (2005) — as a "story." This sets off all sorts of alarm bells.)

This was the most relevant talk of the three.

The next speaker presented this. While it was a reasonable application of computer graphics to augment dance, I'm not convinced it was very useful to non-choreographers. There didn't seem to be any analysis done; it was all representation and re-presentation of the data, as far as I saw. Once upon a time, this would possibly have been suitable for a video presentation, but while this may represent the cutting edge of choreography, it doesn't seem to be much of a SIGGRAPH topic.

The final speaker of the talk ... may have had potential, but did not use it. He had simply taken the output of the electricity-usage monitors in the Dartmouth undergraduate dorms and hooked them up to a gauge instantiated as the hand-drawn animated image of a polar bear — depicted as happy when usage is low, and in danger when usage is high. (Apparently this has measurably reduced electricity consumption in the dorms.) This is all well and good, but hardly seems to be SIGGRAPH material. Apparently there was some novelty, or at least nontriviality, to his data processing (determining what constituted 'low' and 'high' respectively). It might have been better if he'd gone into more detail about that...

At any rate, while fragments might possibly become useful in unforeseen future days, apparently I should have been attending Advances in Real-Time Rendering, where there was hard math to be had. Alas, when I entered around 10:00, the presenter said she was done with the integrals and filter equations and would now pass the stand to another. Indeed, the rest of the talk was relatively useless — there was a description of a deferred-lighting engine, and then a read-from-the-slides talk about light volumes (?) in Crysis.

I ate lunch between those two, around 11:00. This becomes relevant later.

After that, I wandered around upstairs, looking at the Emerging Technologies and Art sections, groups which tended to shade into one another at times. I note specifically, and with some fondness, the Funbrella — an electric motor / kinetic sensor mounted on an umbrella, allowing the wielder to feel the recorded sensations of spaghetti, marbles, and plush toys raining down upon them. More serious was the installation Hylozoic Soil, which is (as one attendee put it) "terrifying and beautiful" — a jungle made out of plastic and actuators that moves on its own in ways a (Terran?) jungle shouldn't. The creaking and buzzing vibrations that pervaded the setup (from the plastic and actuators respectively) were as much what brought it to life as the motions were, though.

The technical portion of the conference got back underway at 13:45. For me, that meant part II of the Real-Time Rendering talks, which started off with essentially an hour-long advertisement for Disney Interactive's Pure (racing game) and Split/Second (racing game with explosions explosions with a racing game attached). To be fair, Pure's segment described in reasonable detail their foliage-drawing methods... although their sort-avoidance algorithm seems kind of silly, given that you can "sort" the points in a regular lattice in a square by distance from any point not in the square in constant time. (Go old-school; think Bresenham's algorithm.)

Split/Second's segment had no such redeeming qualities, though. It was an ad.

After that, things picked up relatively quickly: the remainder of the talk had actual substance, covering depth-of-field, antialiasing, real-time raytracing (on two-generation-old hardware, no less!), and concluding with a detailed analysis and explication ("postmortem") of the graphics engine for Little Big Planet. (Note to non-PS3 owners: yeah, don't expect this to come out for any other platform anytime soon. The engine viciously exploits not merely the SPEs — essentially the entire vertex portion of the graphics pipeline is performed there — but specific quirks of the nVidia RSX's MSAA implementation.)

As this wound to a close, I went upstairs to the Sandbox, where a number of games were on display: Flower, most notably, but there were several, most being independent titles such as Akrasia and ... damn, I don't remember any of the others' names. There was an Xbox360 boxing game, a vaguely Silent-Hill-esque third-person game involving a goth girl running around in a foggy forest that I only got to see over someone's shoulder, and a game involving being a sphere that consumes other spheres by touch -- the only real mechanics for the last are 'momentum is conserved', 'large spheres consume small spheres' and 'you can eject very tiny spheres at high speed in any direction'. Fun. Also frustrating at the level I found myself playing. (Edit 2009-08-06: These are, respectively, Fight Night: Round 4, The Path, and... okay, I still don't know the last one.)

This was just to kill time in between the last talks and the previews of the Technical Papers, which lasted from 18:00 to 20:00. Two full hours of 50-second previews for some four-fifths of the papers to be presented. At least this lets me have some sense of which presentations to avoid on the grounds that their presenters will just be reading off of their slides, and whose papers I could as easily absorb without their presence.

I'm going to give those three at the beginning some credit -- none of them did that.

So I came to the convention center at 8:30; I left eleven and a half hours later; and I promptly staggered northward to the first restaurant I found, one Mulate's. (Which, for the record, had better music than food. Which is not to say the food was bad; it was just nothing to write home about. So I won't.)

To close the day, which had already been long enough that I should probably have had the sense to throw in the towel for the night, I headed to the ACM SIGGRAPH party, at the Generations Hall -- a fairly standard club, and a fairly standard party, and therefore as much of a waste of time for me as the day's opening had been. (It was worth a shot, I suppose.)

Tomorrow will hopefully be less full than today.