Between 2020 and 2024, the artist duo Kata Kovács and Tom O’Doherty—often also referred to by their
abbreviated moniker KKTO—went on a series of walks. They walked along the pathways of now-abandoned railway
routes in Berlin and Brandenburg, which they filmed as they went. In these walks, they also carried portable
speakers, playing the sound of machine-learning-generated audio of invented trains: sound that has been hallucinated
by machines, or by what we have generally come to refer to as artificial intelligence.
In parallel, they recorded and produced a series of compositions on trumpet, guitar, and bass, and combined them
with these alien train sounds. These recordings were first presented as part of the artists’ exhibition at Hošek
Contemporary in Berlin, Germany, in September 2024, alongside videos, prints, film-still grids and other materials.
These recordings foreground
the slow swells, the humming amps, and the tangible messiness of analogue tactility. Careful listeners will hear not just iridescent
drones, but also the scraped strings, the shuffling footsteps, and the wavering imperfections of soundwaves moving air molecules in
a room with microphones.
At the same time, these sounds form a crucible to hold the presence of another entity, a layer of dreamlike and
eerie sound—non-existent trains, or rather a ghostly apparition of machines that have only ever been
computationally dreamt. Drawing on lineages of drone, dub, and
ambient music, Woven All of Dream and Error aims to overlay a bed of thrumming drones and swells with these
contemporary sonic machine ghosts.
Versions of these sounds were initially presented as a soundtrack for videos, during the original exhibition. The
machine-learning-generated part of the sounds are the results of
a patient process of building and training a machine-learning model from scratch, which Kata and Tom
undertook together with their collaborator Kris Slyka. They nicknamed their model Dreamsloth (it dreams, but lazily and
slowly). This model spits out sounds that are recognisable, but sounds that also have glitches and flaws—they strike our
ear as something familiar, yet uncanny.
But... why trains? What makes these sounds worth generating, worth dreaming? Woven All of Dream and Error was
partly inspired by the idea of overlaying two eras of technology—one historical, and one contemporary: railway networks,
and machine-learning networks. One has bequeathed a layer of remnants and vestiges to the world, over the course of the last two
centuries, both physical and social; the other contains potential future ruins and vestiges that we can only begin to guess at.
These sounds, and the exhibition from which they derive, draw their title from a line of a sonnet written by the Portuguese
modernist Fernando Pessoa, first published in 1918. Pessoa’s words, juxtaposing dream and error—hallucination and
disaster—were the conceptually apt encapsulation that the project adopted.
These compositions are now being given a more general release, and this sonically unique set of recordings has
been given a similarly-unique physical form: as a set of fourteen lathe-cut records, one for each track on the album.
The fourteen records in the series correspond to the fourteen lines of
the sonnet, with each track titled after a line in turn.
These records, released by Soft Noise
Acousmatics, in collaboration with Estovers Editions, were cut on the unique transducer-based SL-24 lathe,
at Nuclear Lighthouse Studios in Berlin, by Kris Slyka. Each record in the series is cut into eight-inch square transparent
polycarbonate sheets, housed in custom hand-made covers.
Woven All of Dream and Error is not only a unique set of records, and the conclusion of a sustained
multi-year project, but it is also an attempt to grapple with our emerging technological reality, to consider
the artistic implications of thinking machines, and to
imagine it as a kind of trace or remnant left for the future. Pessoa’s poem concludes:
Yet we think on, knowing we ne’er shall know.
Woven All of Dream and Error · lathe-cut record series
