For the last few months I’ve been getting quite interested in datacentres, and the algorithms that live in them. I’ve been focussing on the architecture of the former as a way of understanding the way we see (or don’t see) the physical, phenomenal products of the numinous network. This is what I talked about at FutureEverything ten days ago—I haven’t written it up yet, but I’m indebted to Claire Welsby for doing so: Let’s make friends with the robots.
I’ve really been enjoying talking about this, but I haven’t been entirely comfortable with the algorithm chat. It makes sense, but Kevin Slavin does it better. It’s not the point.
A couple of weeks ago, I wrote about the New Aesthetic over at the RIG blog (and I am continuing the research over at Tumblr). I wasn’t really sure what it was about, but Matt Jones figured it out in Sensor-Vernacular. One core aspect of the New Aesthetic is that it shows us the world as the machines see it: “the grain of seeing/computation… Of the gaze of another nature on ours… the robot-readable world.”
Digitisation is the process of making a text machine-readable.
(You should read The Internationale’s Software tunnels through the rags ‘n refuse, and I’m reminded too of the shortlived Bookrabbit project, Automatic Bookcase, to photograph bookshop shelves and render them machine-recognisable, and thus searchable.)
In a fascinating talk at Unbound Book in the Netherlands last week, Bernhard Rieder, an assistant professor at the Département Hypermédia at the Université de Paris VIII and author of the Politics of Systems blog, drew a bright line from algorithms to books, neatly pinging the poles of my obsession.
Rieder pointed out that what is occurring is the transformation of the book into a data object, something which can be acted on computationally, and in turn producing what he called “computational value”. “Systems that digitise books, like Amazon and Google, transform books into information, and then unbind and rebind it again as an interactive, social and semantic interface.” (Thanks to Masters of Media for the write-up.)
Two key quotes, via Rieder:
“Is the digital library a machine, or an institution?”—Agre, 2003
“Incite, derail, make easy or difficult, enlarge or limit, render more or less probable… These are the categories of power.”—Deleuze, 1986
(In my FutureEverything talk I noted several times that I was being deliberately, overtly paranoid: I’ve long been fascinated by conspiracy theories and feel that it’s worth deploying them strategically on occasion in order to examine things from a particular angle. After all, one man’s Bilderberger golfing holiday is another man’s secret rulers of the world, and both these things are basically true. So if I go overboard on the Google thing, remember it’s just speculative theory.)
Google Book Search is an exercise in—a process of—feeding the machine. Opposition to it has largely focussed on the intellectual ownership of orphan works. Perhaps instead it should have focussed on the computational value of such texts: an immense corpus of human and societal language, interests and connections.
What Google gains from GBS is not merely the ability to “organize the world‘s information and make it universally accessible and useful”, but to act on that information, to process the largest collection texts ever assembled in history, and out of it to create new data objects and data products. The texts themselves are a substrate; information is materiel.
The value of books to Google is not like the value of books to a library, it is like the value of books to people, who might understand them and act on that understanding.
We should be asking about the algorithms which are being developed not for search but for understanding, the value of n-grams and the capabilities of financial news-readers and recommendation engines.
What does a book look like to a machine? The Roombas are learning to talk to one another; what happens when they’re well-read?