Latent Imprints. That’s the name of this piercingly profound album by Sveið, a free jazz improv trio of international musical sputniks. The name of the album is pertinent. There is something latent in this music, and in the sounds it comprises. Any sound is of course latent: sound can be framed through all kinds of phonographic inscription techniques and when we listen there are features in the sound that arouse us, depending on context, the reproduction equipment, the space we find ourselves in, the intensity of our attention. No sound is the same, even if recorded, because we are never the same when listening to it.
But now there is a new phenomenon in music called latent space. This is the neural soup that gets cooked when large quantities of sounds are fed into a neural network that grows synaptic superhighways of sample statistics; of what sample might follow another. The resulting models, trained on sonic data, thus become pregnant with infinite possible offsprings of sound depending on the trajectorial movements within the multidimensional space, through the navigation of gestural parameters or live sound. The models are mystical, dormant, silent yet booming, and full of virtual potential. They are in need of exploration, investigation, discovery, and performativity. This is what Sveið have done: they have trained a series of particular sonic models which they investigate through the laboratory practice of musical improvisation. And this we can hear very clearly on the album: there is a high intensity attention to latent sounds emerging from the models, which focuses, colours, and shapes the intense music we hear. Playing with brimming latent potential is clearly a fun process.
Sveið. Three musicians—a drummer, a live coder and a saxophonist—describe themselves as “powered by AI,” and it is clear that there is something eery going on that infuses their music with uncanny predictive unpredictability. The album is described as blurring “the boundaries between free jazz, improvisation, electronic music, and artificial intelligence,” where one is a musical non-style, the other a performance method, the third an organological descriptor, and the fourth a description of a machine with a mind, of sorts. Respectively. Respect. Because the description is fitting and we can predict how the music would sound to a certain degree from that designator, but not really. The unpredictability of the neural audio synthesis, influencing an intense and energetic musical performance, yields a mindset of concentration and care that centres on this very exploration of latent sounds.
Then there is the imprint. An engraving that becomes a product after a process of inscription. On vinyl, CD, and stream. The phonographic act of recording this human-AI synthesis in the form of musical performance is a stamp in time, an etching, an imprint on its own. AI is not only reading for us, writing for us, drawing for us, composing for us, playing for us, but it now shapes our thinking and being in the world. In an age where neo-liberal tech bros line up with fascists inaugurating our future, we cannot yield full agency to the artificial output, the manmade manmade, the simulacrum, but challenge it, form it, negotiate it, and document through this inscription where the live music is lapidified. Thus the models become latent again: dormant, powerless, and silent, except haunting us in the mythic aigenic track names that title this music of much silence and much noise. Sveið.
—THOR MAGNUSSON