The burst, 2018 (work in progress)
machine learning, copper plates and tubes, electronics
The installation (in progress) is part of an ongoing research probing for a contingent event of singularity: a moment when techno-automated intelligence and its material realities might come into friction and burst in unison. In the installation, an event of intensive friction is enacted by a machine performing a non-dualistic learning routine of matter abstraction and prediction.
Consisting of 16 metallic probes, the machine operates as an autoencoder (an unsupervised machine learning algorithm) by reading and writing a round copper sample as a latent space (compressed data), in order to predict the upcoming burst. By touching the sample, each probe reads and validates its prediction by detecting the presence of a spark. The touch, in turn, further writes the sample and increases the chances of the burst by the traces left on the sample. The sample thus becomes a record of past predictions, just as it entangles the machine’s behavior and future as its features change over time. Hence, predicting the burst further accelerates it in an evolving feedback loop.