June 13, 2011
Writing for Machines
by James McGirk
Writers are anxious about the Internet and all things electronic, as we worry these newfangled ways of entertaining ourselves might someday obviate our own work. The solution, perhaps, lies in understanding and adapting to this new medium. Consuming enough that we can master its complexities and render appealingly intelligent confections for our readers. But who are these readers? Are they different online than they are in print? Some of them aren’t even human. There is a new form of reader browsing the Internet. For this is no longer just the age of mechanical reproduction; we now have to contend with mechanical readers as well.
William Gibson, who coined the term “cyberspace” imagined it as a mass consensual hallucination, rendered as a cityscape, the prominence of each shape on the horizon an index of how much data was passing through a single point; a point which in 1982 a reader might have thought of as a mainframe computer, and what today, nearly thirty years later, we might identify as an html address or site. On Gibson’s Internet Google would glow the brightest, soar the highest; be an Empire State Building to the Internet’s Manhattan. Most users don’t look at the Internet by volume, however, they read it pane by pane, navigating from bookmarks or through searches, feeding keywords into an ‘engine,’ a series of algorithms, to retrieve lists of linked addresses to the information they seek. These lists are customized to the user, the results tweaked by the user’s location and previous searches. The more searches you make, the more information about yourself you reveal, the more customized the experience becomes.
From a content provider’s point of view (as opposed to a more passive content user’s point of view) an ideal Internet browser might render something close to Gibson’s landscape of crystalline data sculptures, were there a way to capture such information in real time. But commercial users would rather see traffic than the mere through-output of bits and bytes. Who consumes what information, when and why is much more important to commerce than mere bandwidth. Though online sales have grown to become big business, the Internet remains a popularity contest. The real currency of the online world is attention. Being able to read the flow of attention online would mean mastering it, and reaping the ad money that comes along with that attention. But instead of trying to follow where everyone is going all at once, content providers are instead attempting to clone their readers’ minds.
As you navigate the Internet, the Internet – which is to say the entities using the Internet – navigate you. This isn’t a benign process. They want to learn as much about you as possible to snag your attention; not only by viewing content, but by diverting your time into loops of advertisements and possibly even pushing you through a point-of-sale and taking your money directly. They do so by gleaning information about you. Where you go, what you search for, what type of computer you are using… Websites leave small tracking codes on your computer called cookies, and each of these transmits data back to home base.
Keywords (also known as index terms) are the most interesting and valuable traces left by users. Cookies record the terms users use to come across a site. An entire industry has sprung up to interpret these keywords, and another to optimize content online so it can be better read by search engines (this is called Search Engine Optimization). The data they gather is a crude simulacrum of their users; an inscription of their desires for an instant. Almost like a section of brain tissue. A clue. And en masse a hologram of their users collective desires.
All writers crave attention and respond to their readers’ desires. Charles Dickens used live audiences as focus groups for his serialized fiction. Newspapers and magazines have always had to respond to circulation numbers. Electronic texts simply speed up the process. Text online can be altered immediately. There are even advanced analytics packages that use keywords and cookies to anticipate what readers want and automatically generate ‘content’ for users in response to what they ‘perceive’ readers as wanting. Other companies use similar algorithms to assign stories to human beings. When you hear the term content farms, that’s what’s going on.
Google tweaked its search algorithms a few months ago, which trimmed back the custom-generated content that had begun to choke its search results like kudzu. But beyond the first or second page of results, it comes sneaking back and you will still find page after page of sites that copy the content of other sites, or ones loaded with all the correct terminology of whatever it is you seek, but arrayed in such a way that these phrases convey little or no meaning. As replications of our desire, these simulacra are incomplete. It would take an infinite amount of data (and a correspondingly infinite amount of time to collect this data) to accurately model a human being’s wants and desires. But machines are getting closer and closer.
There are gaps between reader and author in a traditional text too. Enormous ones. Between the platonic ideal an author holds in his or her head, the text he or she extrudes into type and the reuptake and processing that takes place in a reader’s head, there is plenty of room for strange, unexpected effects to creep in. William Gibson described the cyberspace generated by a child’s calculator as a grey infinity utterly empty but for a string a few basic arithmetic equations (slim structures of liquid crystal one imagines). This unnameable sea of grey emptiness is not neutral. More of a field or something we project into and allow things to assume shape. And distended from the platonic ideal and warped by exterior forces these things become strange. Even arithmetic has its unexpected, subjective aspects. Many a calculator screen has been reversed to spell mild profanities.
Knowledge builds on memory, and all information builds off what we already know. Reading works by drawing parallels with memories, essentially unpacking an archive into that grey arithmetic field mentioned above and letting it take new forms. The way a machine reads, in this respect, is no different. Software has an archive of its own, a database that it is adding or subtracting from. It 'reads' by comparing its archive to a text, and then updating itself. An author can access this archive with his or her text; and the more sophisticated it is, the more he or she can manipulate it; perhaps even creating an aesthetic experience.
Literary forms are beginning to emerge in response to automated reading systems, searches, and databases. Online, an era somewhat akin to the pamphlet-strewn amateurism of18th Century America is in bloom. The most exotic forms can be found on the Internet’s wild fringe, in its anonymous and pseudo-anonymous chat sites. Here there is a frantic economy of monikers, memes and spoofed identities. In online forums such as the semi-anonymous Somethingawful users compete to create the catchiest, most innovative forms – most often an evolution of an earlier idea, name or other fragment of an idea. The best innovators become famous within their tiny little spheres. Other forums are anonymous and ephemeral – the most famous of these being the notorious 4chan/b ‘Random’ board – where the only recognition earned is the sheer longevity of a creation. A post can only survive as long as it is replied to. Then it is gone forever.
The best memes were once charted on the now-defunct Encyclopedia Dramatica. But now there is no reason at all to create but sheer artistic thrill. Although ‘board lore’ has developed a concept somewhat akin to ‘duende’ – a dark, nihlistic reward in the form of amusement known as ‘lulz.’
The evolution of the online literary form could well come from manipulating these mysterious semantic mechanicals. They offer the opportunity to make writing dangerous again. With the proper keywords, information is taken up into automatic readers belonging to some very interesting entities, to the point where there can be real world consequences. As a way of experimenting with this form I created a series of posts with keywords that I imagined might appeal to some of the more peculiar gleaners out trolling for information online. I posted lists of oil rigs, information about espionage, created a consulting company specializing in complex shipping orders in the Arabian Ocean, wrote about electronic warfare, and laced my work with other ‘edible’ keywords. I received visits from hedge funds, multinational banking concerns, the department of defense, oil companies, environmental organizations, the Pakistani government, the Kuwaiti government, the Iranian government, the Russian government, an unacknowledged US military facility, and a few mysterious hits from ‘Cabin John, Maryland’ (a park across the river from CIA headquarters).
I don’t think my posts ever stirred more than a few pixels. All I did was conjure another layer of anxiety about the online world, but for a writer paranoia is far better stuff than anxiety over obsolescence.
Posted by James McGirk at 12:35 AM | Permalink