The machine dreams in hyperreal hallucinatory visions emergent from the convolutions of its deep neural network. It produces a cycogeographic mapping of site, an extraction of the essence of place through a process of forensic analysis and Bayesian probabilities. Machinic Dreamings are the output of a machine learning generative adversarial network (GAN) trained on 1000 photographs of a body occupying the temporary landscape of the Limehouse Foreshore, a triangular expanse of mud, silt and rocks on the northside of the Thames, just as the river sweeps south at Canary Wharf and only visible at low tide.
All dreamings are collective acts. Machinic Dreamings link anonymous humans and non-humans across time and space. They are dependent on a technical infrastructure of GPUs housed in data centres, located across national borders and interconnected through fibreoptic cables. Machine Learning algorithms have their own ancestry and lineage; the StyleGAN2 algorithm redefined StyleGAN, which built upon wider style transfer research. It is impossible to map the network of actants whose labour has been essential in producing a single machinic image. My contribution was the gathering of a data set and the training of the GAN.
The initial results were a shock. They open up deep-seated cultural anxiety about human relations to the emergence of Artificial Intelligence (AI). The images suggest murderous intent. Severed chunks of flesh discarded on the beach, the possible abandoned residue of Capital’s human meat stranded on the Foreshore between the rising waters of the Thames and the hostile steel and glass of Canary Wharf. Surplus to requirements in an era of AI. There is a generalised fear of the replacement of labour, but the real concern should be the already existing automation of the capitalist. A simple inhuman algorithm, the appropriation of ever more surplus value, has always driven the capitalist. An algorithm very easy to replicate in code. Inside the skyscrapers of Canary Wharf, high-frequency trading, places machine learning at the centre of capitalist accumulation.
The interpretation of dreams is notoriously difficult, and the non-human perspective of Machinic Dreamings opens up multiple alternative interpretations.