The Other Side


Séance; prayer beads, wood, candles

The Other Side is a ritual performance of the mathematics of an artificial neural network with pre-digital technology. The performance references the rites and nomenclature of the modern spiritualist movement whose devotees held gatherings and séances across the western world around the turn of the 20th century hoping to communicate with spirits of the dead. By manually performing the neural network, the work frames deep learning as communion with an emergent intelligence, presenting black-box neural networks as divination and questioning contemporary narratives of machine learning as artificial intelligence.

The ritual is enacted on a large analog computer made of several concentric discs, a string of wooden beads and a book which holds the weights and biases of the network.

The apparatus uses concentric discs to perform the mathematics of a feed-forward neural network. The outmost discs enable multiplication, while the inner two allow summation and activation.

Reading aloud from a book, one member of the séance instructs the others to rotate and align the discs in specific increments that represent the multiplication, summation and activation of values. Holding the beads first to their forhead and then counting back and forth, the reader keeps track of the floating point. Borrowing the interaction modes of a Ouija board, the apparatus is designed to be controlled by several users at once.

A string of wooden beads is used to follow the order of magnitude by counting back and forth.


Comparisons between machine learning and magic are common even amongst experts and practitioners. Although the mathematics of neural networks are well defined, the models they produce are invariably complex and indeciferable. They work, but it is difficult to explain why or how. When used in this way, deep learning is a form of divination, an arcane set of steps that delivers answers without explanations. Without explanation, to trust the network is an act of faith.


The work was performed at the CHI conference in April 2018.