Writing about algorithms in long form is a project that's been on my list of things to do for some time now. I'm fascinated by how they shape our world and have always been fascinated by artificial intelligence in general as someone who, as I think I mentioned in an earlier blog post, grew up on sci-fi.
These two factors surely contribute to the reason that I became obsessed with a certain film--if one can even call it that. I become equally obsessed with the exhibition where the film was shown. The film is titled Autoencoded Blade Runner which I discovered due to it's being a part of the Dreamlands: Immersive Cinema and Art 1905-2016 exhibition which featured more work by artists that I'm in love with e.g. Hito Steyerl, Lynn Hershman Leeson, Liam Gillick, Pierre Huyghe, Bruce Connor, etc.
Autoencoded Blade Runner is a film created by a pHD at Goldsmiths in London named Terrence Broad. The film is exactly what it sounds like. Broad created an artificial neural network, specifically he created an auto-encoder. Auto-encoders do their business by picking out patterns within images and narrowing those patterns down into broader patterns and then, after however many layers, the auto-encoder builds the image again this time using the information it has gathered in it's breakdown process. Or, Ars Electronica, which also screened the film in 2017, states that an auto-encoder "learns to model all frames by trying to copy them through a very narrow information bottleneck, being optimized to create images that are as similar as possible to the original images". Broad then applied this neural network to each and every frame of the film--or, perhaps, the neural network did this on instinct, so to speak. The resulting film is juxtaposed on the right with the original frames on the left in the following image.
In the earlier summer months, when my interested in this work peaked, I stumbled upon Brian Massumi's essay "Realer than Real", in which he argues that Jean Baudrillard's simulacrum is unnecessarily pessimistic and misses the transformative nature of simulacra. Massumi opts for Deleuze and Guattari's conception instead. For Massumi,
“the challenge is to assume this new world of simulation and take it one step farther, to the point of no return, to raise it to a positive simulation of the highest degree by marshaling all our powers of the false toward shattering the grid of representation once and for all”.
Massumi comes to this conclusion through analyzing the plot of Blade Runner. In the end, Massumi argues, what makes the simulacra, the replicants, fight for their representation is not that they want to be exactly like humans. The threat is not that they will live among us, the threat is revolution. The threat is that they are "human-plus", to use Massumi's term. But, to mobilize replicant as a political mode of being seems to be Massumi's admission.
In light of all of this, the algorithm discussion by Kevin Slavin seems too pessimistic. While I agree and find astonishing the ability for algorithms to quite literally shape our world, shape the earth, Massumi's contribution and the ability for the neural net algorithm to make Blade Runner look almost cosmic, for lack of a better term, makes me wonder about the creative potential for algorithms. Even at the end of his essay, Brian Massumi claims that we should think of replicants wanting "not to be human, but to be human plus. This kind of simulation is called "art." Art also recreates a territory, but a territory that is not really territorial. It is less like the earth with its gravitational grid than an interplanetary space, a deterritorialized territory providing a possibility of movement in all directions. Artists are replicants who have found the secret of their obsolescence".
In the case of Autoencoded Blade Runner, this idea takes on a life of it's own. Can the kind of deterritorialization that Massumi discusses be mobilized politically for algorithms as Massumi suggests? The question of artists as replicant, artist as algorithm, also brings up the agency that we give to artworks made by algorithms and even wall street trading as in Slavin's talk. Slavin notes that the most unpredictable algorithms are even given names in the finance world. Algorithms learn problem solving but do they have creative agency? Who is the artist of Autoencoded Blade Runner ? Can we prosecute the Boston Shuffler?
Works Cited ---------------------------------------
Massumi, Brian. "Realer than real." Copyright no 1 (1987): 90-97. Harvard
Mari, thanks for bringing this work into conversation with our readings from this week. I think algorithms generative and creative, as Autoencoded Blade Runner demonstrates, but I also think it's important to hold that creativity in tension with its costs. Algorithms are extractive processes. Algorithm extract and essentialize data for the sake of efficiency (as discussed in Parisi, also discussed in an excellent paper on exhausted algorithms and data "overfitting" by fellow UChicago PhD student Gary Kafer). They are environmentally extractive; requiring the teraforming of land for algorithmic efficiency, as discussed by Slavin as well as in this article on the environmental costs of bitcoin mining (see https://arstechnica.com/tech-policy/2018/05/new-study-quantifies-bitcoins-ludicrous-energy-consumption/). It seems to me algorithmic creativity comes at a cost, be it…