“The invention of the ship was also the invention of the shipwreck,” the philosopher Paul Virilio wrote in 1999. All technology not only becomes obsolete with time, but it also contains hints of its own future state of disuse and abandonment.

In our current work, we are interested in the points of overlap of two areas of technology — machine learning, and railways. One is relatively new, and the other much older.

The social and industrial innovation of railways, which was once the radical cutting-edge of human technology, has had the time to become mature, and to also develop layers of ruins, disused remnants, and buried vestiges.

At present, artificial intelligence and machine learning are in the process of steadily becoming ubiquitous, and for many, they have come to represent the forefront of current technological possibility. At the same time, these technologies contain their own inherent biases and flaws, and we are curious about what possible future integral “shipwrecks” these technologies may contain.

With this in mind, we have been researching the possibilities of using machine learning algorithms to generate audio of trains which have never existed, but are generated in software — computer-dreamt imaginings of non-existent trains.

In parallel with this, we are also retracing, on foot, the routes of a series of abandoned, disused, or disappeared railway lines in Berlin and Brandenburg. On these walks, we have been playing the machine-learning-generated audio from small battery-powered speakers — thus combining the physical route of nonexistent, now-absent trains with an accompaniment of nonexistent, computer-imagined sound. This research process is currently ongoing.

The dual video that can be seen here is a two-channel work-in-progress excerpt, without sound, from this research process.