Traditional Freudian psychoanalysis, of course, posits the unconscious as the inherently inaccessible mental bedrock of conscious mental activity. It is, as the basic argument goes, the chaotic, non-logical, all-desiring source of mental energy—demanding at once the absolute proliferation of vitality and the absolute rest of death—the mad little homonculous which is controlled by the psychological machinery of conscioussness: repression, sublimation, projection, etc. With the insights of modern linguistics and especially the vocabulary of structuralism at his disposal, the French psychoanalyst Jacques Lacan was able to come right out and say that the unconscious is structured like a language.
That language, the way we talk and what we understand as sensible for speaking, represents a certain rationality, and thus a certain kind of morality. If you can’t say certain things because they don’t make sense given our current state of words and rules for words, then you can’t or are discouraged from doing things that correspond to such impossible sayings.
Algorithms of articifical intelligence have an inherently political dimension and at their frontier this dimension represents a precarious pivot between death and life, i.e. conservation and revolution. On the one hand, they interpret new information based on what they know about the past, and there is thus a rational tendency—founded on the very true, real inertia of real history—to assume that the new is like the old and to thus to impose old formulas on novel articulations. Thus, one component of the political unconscious revealed by Google’s speech-to-text algorithm will be the following: it will reveal at points the conservative obverse of linguistic novelties, how exactly the inert speech of the status quo is mathematically most likely to impose the weight of dead customary formulae onto even the most radical, innovative articulations. Thus, on one hand, Google’s transcription service is a kind of molecular linguistic Thermidor machine, mathematically reducing freedom of linguistic invention to that which is most common and dead
On the other hand, because of practical imperfections and simple technological inadequacies—errors induced by background noise, irregularities in human speech (speaking too fast or eliding syllables, slurs, stutters, etc.), differences in dialect, the algorithm’s inability to deal with any languages other than English despite penetration of English by many foreign tongues, etc.) misfired transcriptions of common, dead, bourgeois formulae will become accidental slips of the most profoundly revolutionary social imagination, as in dreams, the delirious ravings of schizophrenics, and so-called “Freudian slips.” Why do we claim that such erroneous transcriptions will have a profoundly revolutionary content? It is because, as errors, they will be outside the limits proscribed by status-quo reason, but by virtue of the best effort of the algorithm, they will be the best errors, the nonsensical, prohibited-by-reason errors which are nonetheless the very closest to reason itself: the errors contained within reasonable language.
One can invent, sui generis, radically innovative theories or images but these tend to be utopian, disconnected from the imminent logic of history and in fact such utopianism is usually a function of socially and politically manipulated fantasies if not designed then at least permitted and encouraged by the dominant powers, as when corporate advertising pitches us so many varieties of freedom only to make us work harder and spend more, to submit more, in order to obtain them. The most radically dangerous and truly revolutionary historical ruptures are those that are already imminently contained in the weighty, inescapable, all-pervasive, strictures of reasonable reality itself. Thus, when common expressions are transcribed erroneously the result represents a floating to the conscious top of the most saliently precarious links in the chain of signifiers that represents the symbolic order of reason, that status-quo itself. Not only can we rightfully deduce that such points of slippage are the most salient and precarious (precarious because prone to imminent reversal or subversion, salient because precarious) but the positive content the transcription assigns to such slipping links is by definition the socially and historically weightiest possible alternative, in other words the most practically realizable. Thus, the practical conclusion of this consideration is to be particularly keen with respect to politically inflected errors in the transcription of common expressions. The mathematics of artifical intelligence and natural-language learning suggest that such errors not only announce themselves as so many heels of Achilles in the rational defense of the established order but also reveal, scientifically, the most imminently practical line of flight.
The following poem was produced by reading a passage of Deleuze and Guattari’s Anti-Oedipus into Google’s transcription service. The errors and slippages are surprisingly coherent and suggestive, distinct and decisvely different than the original text, but connected to it by a cipher which is certainly not random.
When you can, Johnson,
Friends and family,
The meeting of the simple, you know,
My daddy: the residual territorial using my money.
And Catherine, divided, split, you go.
Father, mother, trial of a triangle
With a river as a brief answer
And images of capital, which, short of this, arrive.
Apart from that, it’s to be determined.
It is our intimate Claudio,
Forms of awful solvency.
We’re all Callaway’s and
It is at this that I call my cellphone
And at this that the production, reduction,
Extension of ten finds a family,
Meeting on the simple unit of competition.
It is on the mother that we can, soon,
And if it is sent everything is changed to get the family
On the factors of sofa reproductions,
Of conventions fine,
And all of these factors.
He’s observer. However, not all of his desire
Is to determine the best.
I will be there, should be back on the familiar determinations.
With a reach personal, what the polls hopefully reveal:
A person, my clothes off,
A domain of scribbled images,
No less than configuration services for the point-signs,
The bricks, flows, sure figures, a couple of them
Sound like Apple computers.
I just want to drive, wake up on the workers,
Talk to Dr. Slope, labor
In this way and it’s going to get in touch with you soon,
Just terrible things on the other side of the landscapers,
An image of a test to see you soon in five terrible crashes.
This initial example provides striking preliminary evidence for the hypothesis that the cipher responsible for this encoding is indeed something comparable to universal consciousness. How else would it be possible to obtain such a combination of fidelity and distortion at once puzzling and compellingly decipherable?
A more detailed analysis can be left to literary critics, but let us simply seize on the last several lines of this initial experiment. Obviously Deleuze and Guattari knew nothing of Apple computers, but here we learn that “Bricks, flows, sure figures… sound like Apple computers.” Apple computers are indeed crucially important, solid bricks keeping the creative energies of society harnessed and territorialized. They are “sure figures” whose continued technological development we can count on, our anticipation of which performs the critical function of keeping privileged and creative young adults in love with contemporary technocapitalism. The creative classes of Apple users do not labor but rather personify it, they furnish an acceptable image of labor. These Apple users are neurotically self-conscious of their distance from manual laborers: “terrible things” exist on that repressed, other side where landscapers labor with their bodies. The poem itself is only an image of a test, a portrait of what it might mean to confront these realities rather than repress them, a mere image of a test that everday we continually defer but which always finds us through the perpetual terror it provokes.
Murphy, Justin. 2011. "Using Google's transcription service as a universal consciousness, with an example," http://jmrphy.net/blog/2011/05/14/using-googles-transcription-service-as-a-universal-consciousness/ (October 17, 2017).