CONCEPT

“Who do you think dreamed it?” Alice asks her kitten Kitty at the end of the book “Through the Looking-Glass,” by Lewis Carroll. The indeterminacy pointed there runs not only through this book, but also through the previous one, “Alice’s Adventures in Wonderland.” The theme of dreaming, or of the boundaries between dream and reality, were fundamental issues for Lewis Carroll, a writer and mathematician interested in the limits of logic and, above all, in the impossibility of fixing meaning through paradoxes. In his books, Lewis explores these limits, playing with riddles presented to Alice, who in turn observes the world (real or imaginary, oneiric) through ambiguous and likewise paradoxical constructions.

The question “Who do you think dreamed it?” also points to the issue of the dreaming “I”: in order to dream, in order to access an unconscious, there must be a “conscious.” There must be someone who identifies as an individual and who, from that position, can immerse themselves in what structures them: their unconscious. Evidently, this Lacanian version of dreaming can be questioned. But let us adopt it as a starting point for a question that might follow from the first: can a machine dream? Would a machinic consciousness structured by a machinic unconscious be necessary for dreaming to occur?

Without attempting to exhaust any definition of the unconscious in Lacan, it is worth recalling that his construction of the concept of the unconscious is tied to a specific understanding of what Language is: a dynamic structure of signs in a continuous process of differentiation. This process of differentiation closely resembles machinic/computational processes of a discretizing and combinatorial nature.

If we were to “agree” on a supposed equivalence between the symbolic structure of the unconscious and a machinic symbolic structure (logically symbolic, in this case), how could we evidence, make tangible, this deeper layer of the machine? The work DEEP ALICE explores this fictional universe in which the machinic unconscious meets the human unconscious through Alice’s dreaming. By exploring the combinatorics between elements (paper cut-outs of the characters’ illustrations) from the books (made available as images, visual signs) and the information contained in the prompts (textual symbols), we initiate a continuous hybridization between inputs from the present as perceived by the machine (through the camera) and machinic repertoire/imaginary (a database trained to translate words into images). In the case of DEEP ALICE, the strategy of recombining elements from the books begins with simple combinations, which resonate with the machinic system’s ability to associate more probable images with the prompts; that is, starting from an image formed by random pixels (initial noise), the machine gradually selects the most likely image to emerge from the relationship between these initial pixels and the message contained in the prompt. Thus, the more “probable” (more logical, predictable, grammatically and syntactically correct) the prompt, the more coherent and probable the resulting image will be. As the prompt begins to lose coherence and logic, the generated images become increasingly improbable, disconnected from our own repertoire of images. To reach a certain level of incoherence and randomness in the creation of prompts from the two books in question, we drew inspiration from William S. Burroughs’ “CUT UP” method, producing a digital version of this technique by cutting and reassembling excerpts from the books. The result is disconnected sentences, syntactically highly or entirely improbable. In parallel with the introduction of digital “CUT UPs,” we explore figure/ground variation: the images on table 1, which serve as input for the machine, range from situations in which the figure/ground hierarchy is clear (facilitating the machine’s “understanding” of image content) to images in which this hierarchy is deconstructed (through the generation of dynamic surfaces mapped in the colors of the paper cut-outs). In doing so, we destabilize machinic reading, generating interpretations that resonate with variations in the prompts.

In DEEP ALICE, in the final phase of the machinic dream, we radicalize the combinatorics of prompt symbols by inverting all the characters of the sentence randomly generated by the previous method. In this way, we require the machine to generate images dissociated from images of the human world. The generated images remain coherent with machinic combinatorial/logical processes but disconnect from the human imaginary, from our “imaginable” images. On the surface of these new artificial images, meanings hardly manage to adhere. They continuously slide along the outside of the surface/mirror and, through this incessant sliding, eventually pass through it. And us? Do we also cross the mirror? After all, ”ti demaerd kniht uoy od ohW?“