About Douchebag Brainwaves

Overview

Anger is upsetting to smarm—real anger, not umbrage. But so is humor and confidence. Smarm, with its fixation on respect and respectability, has trouble handling it when the snarkers start clowning around. Are you serious? the commenters write. Is this serious? On Twitter, the right-thinking commenters pass the links around: Seriously?

— Tom Scocca, On Smarm

Douchebag Brainwaves is a blog consisting of automatically generated text. It's generated by my own Markov chain–based text generator after training it on the essays of Paul Graham. I take Paul Graham to be the titular douchebag of the blog, and think that this is a fair description of his own character as it's shown in those essays (though you may, of course, feel differently). The code that actually creates the individual posts by calling the text generator is available here, if you'd like to look at it.

So: this blog is consists of essays created by algorithmically recombining phrases based on a mathematical model created by analyzing the essays Graham has published on his website. Though the occasional phrase written by Graham himself may pop up, this is not in itself appropriation; it's just that the text is generated by performing the virtual equivalent of flipping coins, and occasionally the virtual coin-tosses result in the happenstance production of the original text on which the generator was trained. (Of course, the virtual coin flips are also constrained by the fact that they're traversing chains constructed from Graham's own writing, so it's not precisely the same as the proverbial monkeys banging on typewriters, either.)

So why are you doing this?

Every text is a network of roads taken and not taken. Some of the roads have never been taken, so far as we know, and of the roads known to have been taken, some are well traveled and some hardly traveled at all. Who traveled which roads, and when, and where, are matters of consequence to anyone studying the texts. Roads identical in one respect or another may be seen as very different roads if viewed from a different vantake—and of those different points of view, many will be possible. […] We don’t want to discover what the texts mean but what they might be imagined to mean or to have meant. Those meanings are a function of what texts might or might not do, given their rules of engagement; and those rules are determined from what they have and have not done, as well as what they might have done or might be made to do given their historical descent.

— Jerome McGann, conclusion to Radiant Textuality: Literature After the World Wide Web (p. 152 in ISBN 1-4039-6436-X)

Well, several reasons:

It's fun.
It gives me a chance to practice my coding chops on a quick project (I threw this together on a Sunday).
Feeding a text into a text generator tells you things about it: it's what Jerome McGann calls a deformative reading, which throws features of the source text into relief that have previously been invisible.
This is a primary motivation for many of my other textual recombination projects: it shows me new things about the texts that I already know well (or fairly well). This is, for instance, the case with my other blogs Ulysses Redux and The Worst of Bad Lovecraft, both of which have shown me a great deal about the texts that they consume and remix.
For some texts—and, again, I'm primarily thinking here about about James Joyce's Ulysses and the way that it is consumed and reconstructed in Ulysses Redux—this process constitutes a continuous source of amazement about how texts that were written when the abstract notion of computabilty was barely an area of theoretical pondering with absolutely no practical application were still imbued with the kind of semantic richness that results in the construction of dense Markov paths in ways that produce a constantly evolving set of new texts that surprise and delight without ever being predictable.
On the other hand, when other texts are fed into the textual generator, its reconstructions produce the kind of dreary, entirely predictable rehash of elements that are, to a surprising extent, devoid of meaning. I take this to be the case not only for Douchebag Brainwaves, but also for my Sarah Palin quotes quiz. Feeding these texts into the textual generator and getting back (over and over) texts that are surprisingly difficult to distinguish from the texts that were fed in says a lot about the (lack of) quality and sophistication in the source texts being fed into the generator.

Essay construction

This rarity of statements, the incomplete, fragmented form of the enunciative field, the fact that few things, in all, can be said, explain that statements are not, like the air we breathe, an infinite transparency; but things that are transmitted and preserved, that have value, and which one tries to appropriate; that are repeated, reproduced, and transformed; to which pre-established networks are adapted, and to which a status is given in the institutions; things that are duplicated not only by copy or translation, but by exegesis, commentary, and the internal proliferation of meaning.

— Michel Foucault, The Archaeology of Knowledge

Each brainwave posted on the blog consists of:

A title, generated according to one of several patterns, some of which are themselves trained on Graham's own titles.
An essay of two or more paragraphs, created by the text generator that has been trained on the bodies of Graham's essays.
(On some essays:) a set of notes to accompany the text of the essay.
(On some essays:) an expression of gratitude to various people, organizations, and other nouns, generated partly based on nouns thanked by Paul Graham himself in his own essays, and partly by pulling lists of popular topics from Wikipedia.

Some quick links

Douchebag Brainwaves, an automatically generated Tumblog.
Code generating the blog.
Paul Graham's essays.
Slightly more information.
I occasionally blog about Douchebag Brainwaves here and here.