dessalles

The video above, of Daft Punk’s summer performance at Brooklyn’s Keyspan Park, is compiled entirely from footage shot by approximately 250 audience members. When I first saw it, I assumed the constituent videos had naturally emerged from the event and appeared online. I was mistaken, the assembled footage is the product of a coordinated effort, organized by Olivier Gondry (Michel Gondry’s brother) and commissioned by Daft Punk themselves. That’s not to imply something similar couldn’t occur organically. It certainly did when the Boredoms performed 77 BOA DRUM this summer in Brooklyn Bridge Park. [As it turns out, Gondry's video is in fact inspired by the Beastie Boy's movie, Awesome: I Fuckin' Shot That] Regardless, the idea of scraping the internet for documentation of an event and compiling it into some sort of collective memory is entirely brilliant.

Software developer Blaise Aguera y Arcas has been developing an application to similarly compile photos into a single record using a process called Photogrammetry. PhotoSynth (watch a video demo from TED ‘07) can quite amazingly patch together a pile of photos taken of a location and without any human input construct a 3D model of the represented environment (e.g. give it all the photos from your visit to the most photographed barn in America and out pops a 3D model of Moulton Barn). What’s dissapointingly missing from Photosynth is the element of time. Photosynth can’t analyze photos taken at different times and construct an evolving 3D model of the environment - differences between photos are ignored. Regardless, it’s worth it to extend the concept further and fantasize about using video, instead of photos, to generate an animated model. Imagine hooking the London CCTV system into a centralized instance of PhotoSynth and maintaining a 3D recording of the entire city. What couldn’t you do with that?

[Photo: London CCTV cameras by Antonio Martinez]

All this points to an emerging trend and design challenge the 21st century seems keen on solving: How can we best compile and maintain a single recording of all of reality?

My first pass at the question leaves me thinking there are two basic approaches. The first is an exocentric one, effectively described above, where things observe other things. Then there is an egocentric, object-oriented approach where each thing “observes” itself.

Before getting into the (dis)advantages of each I want to point out how inline the second approach is with Bruce Sterling’s concept of Spimes. In brief, a Spime is a theoretical object which continuously records, measures and shares information about itself throughout it’s entire, theoretical lifespan. So in Sterling’s “Spime World” we have a situation where instead of relying on specialized objects to observe other things, each object instead just continually broadcasts a description of itself. With these descriptions we can then generate a record of reality. Three things to note about this situation: We are no longer maintaining a “recording” of reality in the photographic sense, but a description of it. Secondly, we now need a language with which things can describe themselves. And finally, since not everything is a Spime, our recording will be far from complete. However, what makes the Spime solution really interesting is how easily our recording, generated from data, can be augmented, manipulated, transformed, navigated, mashed-up, remixed and (most importantly) inserted into video games.

All of that’s not too say some combination of the two will likely prevail. And then, let’s not forget Crowdsourcing. This would be the angle from which Google intends to tackle the challenge.

That’s it for now.