[Slide 2] This is a letter Blake wrote to Reverend Trussler in August of 1799. Jason Whittaker and I used this quote in William Blake and the Digital Humanities to argue that Blake saw his creations as not belonging to him, but as things that were passed from one creator to the next. Similarly, Ian Balfour contextualizes Blake’s work as a form of prophetic citation or an “echo” that resounds with others. I feel this understanding of creativity is extremely important, particularly as we consider what it means to preserve the past.
[Slide 3] Mike Goode’s characterization of Blake’s work as having a “delinquent tendency” to “de-compose” itself is also pertinent. Goode is arguing against the notion that Blake’s poetry is absolutely inseparable from his visual art articulated by critics like W.J.T. Mitchell and Robert Essick. I would suggest that the completeness argument is, in fact, enabled by print-assumptions associated with the notion of collecting as well as Foucauldian assumptions about authorship as a site for individuality and persecution. If we look at phenomena like Blakean citations on Successories posters (as Mark Lussier has shown us) or the various twisted ways the Jerusalem hymn has come to signify both British nationalism and punk-rebellion (as Whittaker and Keri Davies have noted), we can see that many of these “de-compositions” have political lives of their own apart from their appearance in Blake’s work. I think there is also a way that this understanding of “de-composition” is linked to what Andrew M. Cooper has called Blake’s “revisionary logic,” and his understanding that Blake’s system can be seen as a huge set of differing recursive loops. The Blakean system could be seen not as a narrative, historical, or mythological structure, but as what Kari Kraus has called a “possibility space” for imagining other realities: in other words, a template for creativity or making.
[Slide 4] I’d like to appeal to the emergent school of ‘critical making’ to make a case for understanding computation as a creative practice. Matt Ratto has coined this term to challenge the maker movement to include more critical and politically progressive discourses into its work. But it also extends to fields like design, zine culture, even programming. I argue that this term is especially applicable to Blake, who embedded his arguments in his creative work. When thinking about the recursive and de-compositional structure of Blake’s work, I believe this term also helps to uncouple the idea that digital scholarship has to be merely archival. Marcel O’Gorman has already talked quite a bit about the “archive fever” of the digital humanities, but what might the alternative be?
[Slide 5] One example of this creativity is this project by one of Kraus’s students, in which a material book is wired to a computer. At first glance, this looks like a somewhat grotesque Frankensteinian monster, with wires growing out of it. But it’s actually using the book’s materiality to construct an interface with the computer. This is a Makey Makey, which is a small circuitboard using a USB cable that allows you to remap the keys of your computer. The Makey Makey was designed by engineers at MIT in order to get children interested in programming, but it has a wide variety of uses.
[Slide 6] As you can see here, from another one of the photographs, the student wired the Makey Makey to create a multimedia edition of Coleridge’s Kubla Kahn. So, if you have enough graphite, that graphite conducts electricity, which creates a circuit with multimedia content stored on your computer. “Touching the top of the first page of ‘Kubla Kahn,’ Cameron Mozafari describes “causes calming Chinese hammer and dulcimer music to play and thus sets a complementary reading mood” (94). As a purposeful altering of the reading situation. Mozfari elicits a parallel alteration in what it means for a reader to be immersed in a text.
[Slide 7] Immersion is, of course, a primary fascination with the recursive strands in Blake’s work. Often, as Donald Ault has pointed out, Blake’s later prophetic works actively work against readerly immersion into a consistent fundamental reality — content instead on generating “a narrative field in which the past is not finished and closed but incomplete and open-alterable and revisable” (4). I decided to create a Markov-chain enabled Twitterbot in order to explore the limits of Blake’s recursive nature, and to see if there was a way to fit together a counterfactual Blake: one that sounded like the poet we know, but who’s work is generated by a computational algorithm. You can see the twitterbot (and its work) live by going to https://twitter.com/autoblake.
[Slide 8] Markov-chain Twitterbots became famous mostly due to the @Horse_ebooks Twitter account (even though this was later revealed to be a real person imitating a computer). This account was originally created as a spam Twitter account to entice people to click on possibly dubious links. It started using non-sequitor statements generated by a Markov-chain algorithm in order to avoid detection from anti-spam software and has consequently captured the imagination of many Twitter afficianados. Programmers like Mark Sample (who’s created several Markov-chain Twitter accounts like @SaveHumanities), Moacir Pranas de Sa Pereria (who created @KarlMarxovChain), and Zach Whalen (whose bot tweets strings of ROM ASCII characters) started using the Markov-chain much to generate forms of poetic expression.
[Slide 9] A Markov-chain works much like Google auto-complete by creating chains of text based upon probability. Dan Catt explains it well here. Within a given textual corpus, there’s a certain probability that one word will follow another. This probability is often determined in a Markov-chain algorithm by what’s known as Markov order. If we, for example, set the Markov order to 0, then the algorithm spits out nonsense, since it assumes that all words have the same probability of appearing next to each other. If Markov order is set to 2 or 3 (the order closer to the English language) it will assume that certain words are more often paired with each other, then search for words based upon that probability. The resulting chains are displayed as tweets. By the way, this is a introductory explanation of how Markov-chains basically work but it certainly doesn’t cover every minute detail of the theory – which is complex and is a still-developing part of the Natural Language Processing field. Autocomplete is still pretty new, and gets things wrong often. I’m no expert in Markov-chain theory. My tweets aren’t perfect yet, and I still have some learning to do – especially when constructing complete sentences as tweets. Fortunately, given Blake’s particular interest in recursive structures, this isn’t as much of a problem as it would be with other authors.
[Slide 10] As you can see, it takes quite a bit of “tinkering” to mold Blake’s voice out of the corpus you present to the algorithm. You have to be careful to not be too random, lest you lose the patterns that makes Blake’s voice unique. On the other hand, you can’t be too structured, because you don’t want to just repeat lines from his poetry.
[Slide 11] The trick, and you can see that I was still working this out, is to create unique tweets that “sound” like Blake could have written them. Imagine an alternate universe Blake, who says things he never wrote in our universe, yet sounds similar.
[Slide 12] Overall, I believe the bot not only functions as a form of what Chuck Rybak calls “found poetry,” but provokes some extremely important questions. First: can we use algorithms like this one to conceptualize authors as computational functions? Foucault talks about them in terms of their social function, but if we can string together a stylistically consistent set of words comprising a text, how long before we can start creating forgeries of Blake’s text? Second, if we take the de-compositionist and recursive nature of Blake’s work seriously, how might we leverage Blake’s work (and this bot) to understand the mystical or historical valences of recursive structures that are so central to computer programming?
Thanks!