Statistical Machine Art

In the appendix to his book on Foucault, the French philosopher Gilles Deleuze outlines in teasingly short form the transformations of life, labour and language in the age of advanced machines. Deleuze uses the term technology only fleetingly even if he hints that the new form of Man has to do with a special relation that it enters with forces of the outside now suddenly as if inside Man. This means references to molecular biology, cybernetics and information technology as well as silicon. What Foucault had analyzed as the encounters of the form of Man with the folds of the outside are now entering the regime of the Superfold, that somehow seems to have to do with the technoscientific basis on which the material and expressive basis of humans is being reorganized. In other words, now in the age of algorithmics, genetics science and cybernetics, humans are not what they used to be. In terms of language, Deleuze briefly refers to the forces of unlimited finity hence, not just the infinite, but the finite number of components yields a practically unlimited diversity of combinations. Indeed, combinatorics, as one of the characteristics of word-art in the emerging age of computing, characterized so many of the examples Deleuze gives: Mallarms book, Pguys repetitions, Artauds breaths, the agrammaticality of Cummings, Burroughs and his cutups, fold-ins, as well as Roussels proliferations, Brissets derivations, Dada collage). Especially in terms of post WWII avant-garde, it can be connected to an interest in the algorithmic logic governing language and positing a new sort of unconscious that is machinic. Deleuze tracks in a new materialist fashion the new form language takes beyond what it designates and signifies, beyond even the sounds, which resembles indeed a very algorithmic take on language; language as detached from the living human body, and as a form of statistically calculated structural wordplay instead that might not sound like anything as it is mediated only between machines. Hence, think of Baden Pailthorpes Lingua Franca exhibition, its various transformations and remediations of Orwell's text Nineteen EightyFour, as one such exemplification of the new forms of language in the age of software. As he outlines, the project is about the various translations of the original text; from the filmic translation (or what we might normally call adaptation) by the director Michael Radford, featuring John Hurt as the protagonist Winston Smith, to a further remediation of that filmic material through a removal of dialogue, for instance. In the words of Baden Pailthorpe, what is left is a series of intense glances, breathing and physical gestures. The body, stripped

of its wholeness, becomes an event. A glance. A breath. A move. Statistical rationalization was a key feature in the framing of the body as stripped to its bare functions in rational work management. But what if what remains is not the perfected operation at the conveyor belt, but just a breath? And its not only that translation; further layers are added through Stefanie Posavecs data visualization (http://www.itsbeenreal.co.uk/), an investigation of how text can be visualized into new sorts of diagrams, rhythms and patterns outside the conventional ways of categorizing texts to genres, authors and meaning. Furthermore, the text of Nineteen Eighty-Four is exposed to the Google-translator and with a back and forth movement is translated back into a mock version of the original. First from the original to a variety of languages, and then translated back hundreds of times, and here is what you get: the original starting line It was a bright cold day in April, and the clocks were striking thirteen has become April that he recognised three times. And the more famous line It was one of those pictures which are so contrived that the eyes follow you about when you move. Big Brother Is Watching You, the caption beneath it ran is after its Translator handling: But, you will the logarithm of the eye to be moved, if a man files. Big Brother is keeping yourselves, you and to flee. As one feature of the exhibition, Baden Pailthorpes interests link with the process of Googles translation machine, which besides translation, engages in the removal of words and compression of text. Baden Pailthorpe addresses this as a certain sublanguage that is being worked with, a layer that is the material of the statistical machine translation, at the core of such processes. Indeed, if we look at the idea of Statistical Machine Translation, what is revealed is this algorithmic level of language? At the moment primarily supporting 58 languages, Googles Translator is a mediatic link between the various human languages, even if often with dubious end results. Finnish rarely sounds Finnish after the translation, and same thing applies to many other languages. No wonder, as already the linguist Roman Jakobson pointed out the futility of the task, incidentally in the year 1948: "no system of mathematical logic is adequate for the description as used in the Finnish language." As its theoretical grounding, statistical machine translation relied for a long time on classical information theory, and the idea that language is redundant. In Warren Weavers words in his Recent Contributions to the Mathematical Theory of Communication (1949):

Since [the] English [language] is about 50 per cent redundant, it would be possible to save about one-half the time of ordinary telegraphy by a proper encoding process, provided one were going to transmit over a noiseless channel. When there is noise on a channel, however, there is some real advantage in not using a coding process that eliminates all of the redundancy. For the remaining redundancy helps combat the noise. This is very easy to see, for just because of the fact that the redundancy of English is high, one has, for example, little or no hesitation about correcting errors in spelling that have arisen during transmission. In fact, Weaver himself was one of the pioneers in the investigation of machine translation in the 1950s. As outlined in Peter F. Brown et. Al, The Mathematics of Statistical Machine Translation, the fact that computer power allows a more thorough scanning of possibilities of suitable resonances between, lets say, English and French languages, allowed the development of those early ideas. Indeed, and in a very condensed fashion, what we have according to Brown et al. is this in his words, Fundamental Equation of Machine Translation.

For those of us not so thoroughly equipped with mathematical skills, the short version is this: it basically outlines the relation of strings of English language with strings of French in accordance to the statistical possibilities that these specific strings (for instance sentences) would have a proper relation that we would deem relevant and meaningful. This relation stems from a tension between what is already out there the translated examples to which the algorithm can refer to and how languages mutate, live and transform. And yet, as any algorithmic translation (and Baden Pailthorpes Eighty-Four Doors shows), languages mutate on the non-signifying, algorithmic level too. Where signal was, semantics might follow. To quote Brown: In statistical translation, we take the view that every French string, f, is a possible translation of e. We assign to every pair of strings (e~ f) a number Pr(fle), which we

interpret as the probability that a translator, when presented with e, will produce f as his translation. We further take the view that when a native speaker of French produces a string of French words, he has actually conceived of a string of English words, which he translated mentally. Given a French string f, the job of our translation system is to find the string e that the native speaker had in mind when he produced f. We minimize our chance of error by choosing that English string 6 for which Pr(elf ) is greatest. So in short, what governs French and English language relations is here not so much the long history of wars and emotionally loaded cultural suspicion of the strangeness of the others habits and customs, the world of human affairs only, but the elegant seeming equation for the probabilities that our dialogue partner actually understands that I was asking her to pass the salt and the sentence not being a Freudian slip, which of course now is possible as a misunderstanding on an algorithmic level. In more serious terms, the algorithmic of language is of course connected in the case of Google to a wider economy of global digital media. The technical conditions that Google is able to cater us are a quick and dirty way to tap into approximately the 58 languages we do not understand. It is not only a nice way for us users to pretend we understand what the other person says, but also part of the wider themes of mediation in the global age. And yet, what is being mediated, is not only for human eyes and human ears. The digital media economy of Google is, as we know, based on the massive collection of data and filtering, mediating the data to suitable purposes with income streams. The actual Translate and the Translate API services can be used to interrogate wider themes of digital culture. This is what of course Baden Pailthorpe does, and what invites me to call his project Statistical Machine Art: it is interested in translation, transformation and remediation as the fundamental features in terms of the work being remixed and reused (Orwells Nineteenth EightyFour) as well as the way software itself works: a force of translation, transformation and change that is more closely to be approached through this difficult localizability than a mere fact of it as source code or execution. Such themes are brought out so well in Wendy Chuns recent work on software. But the translation research and business that Google engages in are not only restricted to this one service. Indeed, it is part of their wider augmented humanity (Erik Schmidts words) program that attaches to the industrialization of language, memory, emotions, location and

more. As a way to understand the comprehensiveness of the task, it is good to point out that Google is not just about the algorithms, but about the folds of the algorithm as part of everyday actions, habits and locations. Indeed, such a realization of the possibilities of speech and language related services characterize its stronghold. In the words of Wade Roush, who summarized the three assets behind Googles aesthetico-political digital economy: One is the gradual displacement of rules-based approaches to processing speech and language by statistical, data-driven approaches, which have proved far more effective. Another is the creation of a distributed cloud-computing infrastructure capable of holding the statistical models in active memory and crunching the numbers on a massive scale. Third, and just as important, has been the profusion of real-world data for the models to learn from. As a corporate logic and impact that invites comparisons to Nineteen Eighty-Four, Google itself engages in various forms of framing of code, from the interest in the algorithms it uses to how they are employed as consumer front-end products and business models. Further, the framing of code that Baden Pailthorpe tries in this work to engage as the sublanguage of language, is itself in the case of Google furthermore removed from the consumer. With cloud computing, we are facing the removal of that crucial process from our fingertips: When we do voice translation, when we do picture identification, all [the smartphone] does is send a request to the supercomputers that then do all the work, is how Erik Schmidt pitched this. Computers do the processing of language, image and what we used to think of our media or arts. Further removed from human hands, ears and eyes, we get served it back as consumers. Hence, the new powers of the Superfold that Deleuze addressed are not only ones inside us, but in server space and computing services out there. Obviously, even this apocalyptic tone should adjust to the more complex ecology in which we are drawn as part of that remediation loop. Indeed, as an investigation of what is extracted, what is passed on, and what remains after visual or written content finds itself in a new environment, Baden Pailthorpe works as a cartographer interested in movement of expression in the age of software. What remains is a breath, a glance. Jussi Parikka

References Brown, Peter F. et al. (1993) The Mathematics of Statistical Machine Translation: Parameter Estimation Computational Linguistics vol. 19, No. 2., 263-311. Chun, Wendy Hui Kyong (2011) Programmed Visions. Software and Memory (Cambridge, MA: The MIT Press) Deleuze, Gilles (1988) Foucault trans. Sen Hand (Minneapolis: University of Minnesota Press). Shannon, Claude E. and Warren Weaver, The Mathematical Theory of Communication (Urbana: University of Illinois Press, 1949), Roush, Wade Inside Googles Age of Augmented Humanity, 28 February 2011, Xconomy, http://www.xconomy.com/sanfrancisco/inside-googles-age-of-augmented-humanity/. Weaver, W. (1955). Translation (1949). In Machine Translation of Languages. (Cambridge, MA: The MIT Press). Bio Jussi Parikka is a media theorist who writes on software cultures, media archaeology and contemporary cultural theory. His books include Digital Contagions (2007), Insect Media (2010) and the forthcoming What is Media Archaeology (2011), in addition to several edited volumes. His website: http://www.jussiparikka.net.

The Speaking System ///////////////// the more I say the less you understand: (an argument for ebbing entries in the Dictionary or of the future expansion of language in a lexicon of interpretation) the more you say the less I have to: (the difference between using a thesaurus which has of course an index and the dictionary which is an index but Ill explain that to myelf another time) the more we say the less opportunity there is for dialogue: (language is a system according to its Science but, as Frederick Nietzsche agonised, to the contrary, the condition of understanding is a baroque discipline) the more I say about nothing the more you listen: (a fugue echoes in decorative space yet a word has no depth in which to reverberate unless Im in a Devious corner where the retort is evident to everyone apart from the orator) the more I say less, the more you listen: (other longstanding arguments: maybe youre doing things with words it could be that theyre passive or active or its an interpretation through Form ekphrasis to the Ancient Greeks) the more you listen, the less I say: (a political question is raised: the Labour of meaning who occupies its territory? it takes a lot of effort to comprehend it would destabilise Time and Capital if the obligation were upturned) the more we say nothing, the more we say: (poesis is the long way round and elision doesnt eradicate or evacuate what was meant but you try to excuse yourself: you drank too much and blabbed about things better off supressed) say nothing, say it: (forget: Body fillers puffed up, iron-on t-shirt slogans) Language politesse, page

did I leave something out? (apart from Expletives the realm of non-verbal Indistinction is ardently underrated) I said nothing: (which happens even to the most verbose of Characters myself included only the conversation is one-sided) nothing: (O 0 o Google Translate) I saw you today. I saw what you meant: (its radical recursive strategy has come and what remains behind, the Excess, is subsumed into an organising principle and its the same System of Dialogue) ///////// Lets start simply Lingua Franca establishes that there is no common language in its presentation of image, text or voice because: 1. The textual content of the book is a structure unto itself, relinquished as it is to algorithms; 2. As a System, mimesis has no content or meaning; 3. While the on-screen characters in the film of Nineteen Eighty-Four have relationships, there is no room for mutual intelligibility (a circumstance in which comprehension can occur between people of different language groups); 4. What is produced in the effort to understand is the event of a language, but its content is never present in the real, nonetheless the bridging or crossing made is a language and this is also a substance. The reprise beyond mimesis Does Eighty-Four Doors simply repeat the text, or does it inflect or reflect on George Orwells Nineteen Eighty-Four? It doesnt. If youve read the original book, or seen the film, you do. And what happens when Google Translate is employed, not as a tool for the sensible, but as an apparatus in itself? This assumption is the same for the machinic tool, which acts like an erstwhile human, yet, while the text is processed via dozens of languages, often from and to English, it is not actually translated. The difference with the lingua franca is the human

quotient, where a third language makes communication possible, systematically bridging the incomprehensible that usually comprises the divide of alien tongues. Pailthorpes Lingua Franca vacates all dialogue from director Michael Radfords 1984 cinema production of Nineteen Eighty-Four. The footage is roughly cut just at the point of the characters speech, and is projected on an A5 piece of paper, using a split screen technique. John Hurts Winston Smith is seen on one page, as Julia, played by Suzanna Hamilton, appears on the other. Unlike the film, both characters are viewed at once. Below this, the ultimate Google translation of Orwells text is typed out in white on a black band at the bottom of the page. The actors go to speak but splutter and choke as their utterances are cut out. What should be a play of thwarted desire is instead inarticulation and disjunction. Richard Burton enters as O'Brien, the torturer. And the literal violence is refigured here into a somewhere between gagging and disembodiment. In both cases the voice that has been banished is still present, just spatialised. Not hearing, being the equivalent of not seeing, makes for a new passage to thinking, so that, according to Gilles Deleuze, thought which is born in thought, the act of thinking is a thought without image. But what is such a thought, and how does it operate in the world? (Deleuze 1968: 167) For Plato, poesis was at the crux of the dilemma of representation, and in The Republic he spent some time explaining that truth could be exemplified as it was manifest in the world of form but not simply as a one-dimensional depiction, for we may be able to think of it but it might not be visible. The injunction of perception made access to this thought difficult. Especially in representational forms, which could only partly replicate truth. Poesis introduced a third projective image, an imitation of an imitation, at another remove from the truth. Algorithm: a system for the undecidable elements of the system In Algorithmic Logic this impasse is spanned by a lingua characterica or 'formula language', its unique symbols able to engage in unadorned thought. No human being can write fast enough, or long enough, or small enough (smaller and smaller without limit ...you'd be trying to write on molecules, on atoms, on electrons) to list all members of an enumerably infinite set by writing out their names, one after another, in some notation. But humans can do something equally useful, in the case of certain enumerably infinite sets: They can give explicit instructions for determining the nth member of the set,

for arbitrary finite n (Boolos and Jeffrey 1974, 1999:19) Emanating from this same mathematical school, in the mid-1930s, the conundrum of the decision problem (Entscheidungsproblem ) was posed. Algorithm was consequently used to define the role of decidability, which remains crucial as the root of formal systems and their origins in a small set of axioms and rules. In Eighty-Four Doors an undecidable quotient drives the translation procedure and is also the immaterial product of the crossing. Yet translation itself disavows both the zone of dissonant productivity at the intersection: one cannot go beyond the ideological matrix of language in or through language. Again, Plato rejoinders: ekphrasis, the rhetorical device in which one art form translates another thus encapsulating more compellingly the essence of the thing. Alain Badiou theorises the ethics of deciding on the undecidable Badiou's primary ethical axiom, directed at the human subject, states that we must make a choice, in order to 'decide upon the undecidable'. But the big problem for philosophy would be how the subject might identify the indiscernible. Badiou turns to mathematics and set theory in particular, to solve the problem of where multiplicity and infinity of nothing can be put, or alternatively, how it should exist outside of itself. Badiou does not stop there, he advocates for the return of the subject, in temporal changing openness (he calls this the event). Love, science, politics and art offer the four domains in which a subject might become an active subject, holding in a choice action of fidelity, as he argues, to the transforming life-truth of an event. If youre wondering what the link is between algorithms and Badious theories, both determine that a system better serves to release the human subject from its self-determined containment in thought. Eighty-Four Doors thus crosses the junction of two philosophical paradigms, and one does not of course preclude the existence of the other. You dont have to subscribe to any such school of thought (I dont) but simply use the very ethic of the undecidable to shift or transfer through the conceptual and perceptual space. Projected images return as algorithms of dissonant (or undecidable) thought The piece of paper in Pailthorpes installation, on which the footage and the text project, displaces the dialogical text from the on-screen character in much the same way that, conversely, the algorithm left the page, allowing for the computer calculation of infinite sums. In Cinema 2: The Time Image (2005), Deleuze argues that the post-war condition and the material experience of society is one and the same

with the shift towards a cinema of fragmentation and montage and the (particularly perceptual) effect of screen images. The allegorical operation of cinematic and other perceptually related uses of montage is also central to German philosopher Walter Benjamins conceptual methodology. The notion that montage generates a zone of dissonance was, for Benjamin, a procedure that he applied to writing and to images and to the nexus of the two. Something akin to the undecidable is bred in the space where meaning cannot be fixed, which Susan Buck-Morss, in The Dialectics of Seeing: Walter Benjamin and the Arcades Project, calls dialectical thinking (Susan Buck-Morss, 1989). This, Buck-Morss explains, is realised in Benjamins juxtaposition of writing and dialectical imagery, where past and present interact with one another, as the method and subject of a critical revision and reprisal of history, enacted through the practice of montage, of which Benjamins epic yet unfinished Passagen-Werk (or Arcades) project (19271940) was an embodiment. The late cinematic work of Jean-Luc Godard is instructive of this method too. In a dissertation from 2009, titled Border Crossings: writing, confinement and the voice, I examined the role of dissonance in Godards films. The following two paragraphs are an extract from that text. In Godards Nouvelle Vague (1990) sound is given as much if not more credence than the image, with an abundance of acoustic sources that tumble over each other. The effect is cacophonous, especially as one dialogue overrides and effaces another, building to a bedlam of six, seven and even eight independent voices speaking at once. At other points, silence intervenes and swamps the diegesis. Constant shifts between internal and external sound also exacerbate the most natural of sounds; and naturalistic sounds turn up out of context, or slightly out of synch. A squawking seabird is heard just after the camera has moved indoors, the sound of the sea slapping on a boat is heard in very close proximity to the ear, a womans high-pitched throat singing is contrapuntal to a tranquil lake shot, and loud trumpets sound in the middle of a restaurant scene. The murmuring of Nouvelle Vagues off-screen voices and their dissonant conjunction takes the ties that logically bind the image to the audio track, and stretches both sound and image across film. What emanates from the polyphonic personae of him (Alain

Delons slippery character), and the classy multi-lingual woman, her (played by Domiziana Giordano), is a critique of the attempt to corral the mind and how this can prevent a person from experiencing the liberation of unguarded otherness (and efforts to suppress this openness end up in death or disposal of the other). The couple builds a hall of mirrors: they are not divided nor are they united. We hear their voices, but also the voices of many others even ourselves so that during the process of identification with the figures (man/woman, him/her) we gradually perceive the form of the individual as utterly contingent. This otherness arguably brings about a dissolution of borders, for in Nouvelle Vague there are no boundaries between any of the characters, times or places. The films structure, moreover, is not overtly articulated at any point but is conveyed by splices of dialogue, inner thoughts, floods of music, acoustic effects, field recordings, quotations without references, voices, title pages and an array of visual and aural impressions. [The acoustic component of this concoction is so affective that it is easy to imagine it as a separate work, a film without pictures. As testament, in 1997 ECM Records released the complete soundtrack of Nouvelle Vague, including its dialogue, music, sound and silences.] The dialectic of imagetext that plays out in Lingua Franca is equally explicit in its visual arrangement of footage and type text, as it steps in to rescue us from undecidable (ill)literacy. Its dissolving non-space of image-text relations throws us into the perceptual zone of intelligible dialogue, but only because the words have been excised (and perhaps, as Badiou might conceive of it, have been placed in an empty set), allowing the observer, arguably a generative participant, to enter into and become a montage and see or voice themselves within the imaginary intersubjective domain of WinstonJulia and Winston O'Brien. Im quite comforted by this, as Roland Barthes seemed to be, because Id rather love the him that is I than the he that I love (Barthes, 1978). Disbanded as I am from the image of myself; a subject and an entity lost after having tried to listen to the empty sounds of Lingua Franca.

References Badiou, Alain (2005). Being and Event, trans. Oliver Feltham. New York: Continuum. Barthes, Roland (1978). A Lover's Discourse: Fragments, trans. Richard Howard. New York: Hill and Wang. Boolos, George; Jeffrey, Richard (1974, 1980, 1989, 1999). Computability and Logic (4th ed.). London: Cambridge University Press. Buck-Morss, Susan (1989). The Dialectics of Seeing: Walter Benjamin and the Arcades Project. Cambridge, MA: MIT Press. Davis, Martin (1965). The Undecidable: Basic Papers On Undecidable Propositions, Unsolvable Problems and Computable Functions. New York: Raven Press. Deleuze, Gilles (1968). Difference and Repetition, trans. Paul Patton (1994). New York: Colombia University Press,. Deleuze, Gilles (2005). Cinema 2, The Time-Image, trans. Hugh Tomlinson and Robert Galeta. London: Continuum. Godard, Jean-Luc (1990). Nouvelle Vague, film, directed by. Paris: Cine Video Film; DVD, 2007, Cahiers du Cinema. Hibberd, Lily (2009). Border crossings: writing, confinement and the voice, unpublished PhD exegesis, Monash University, Melbourne. Turing, Alan M. (19367). "On Computable Numbers, With An Application to the Entscheidungsproblem". Proceedings of the London Mathematical Society, Series 2 42: 230265. doi:10.1112/plms/s2-42.1.230. Reprinted in The Undecidable . Bio Lily Hibberd works across live performance, writing, painting, photography, video and installation. Her practice is concerned with questions of time and memory, and historical research or encounters with specific communities and sites. Lily is founding editor of the independent journal un Magazine and holds a PhD in Fine Art.