11/25/2020

Ash Koosha Presents: YONA Part I – (Under Your Skin) Synthetic Emotions Of a Virtual Poet

This week we explore the artificial intelligence of Ash Koosha.

Ash Koosha wants to kill the ego of artists as we know it. By utilising hybrid models of sequential arrangement, generative audio and a curatorial approach to composition, the British-Iranian multidisciplinary artist, producer and programmer seeks to expose the fallacy of the creator genius by demonstrating that creativity is no longer something that only humans have access to.

YONA is the first of what Koosha terms “Auxumans”, or “auxiliary humans”, synthetic beings that think, speak, move and sing exclusively in a virtual space. “YONA was initiated as a voice instrument in 2017 when I was recording an album and was wondering if I could have an imaginary guest singer”, he explains.

“YONA’s voice was very low quality at the beginning as I was trying to regenerate text to speech model through a melodic filter, encoded and decoded back into singing”, he continues. “One day something crazy happened. The voice caught on an auto-harmonisation of sorts that within a key range made a very unusual and unique sound. The emotions delivered (in C Minor) were stunning.”

Koosha then set about developing what began as an instrument into a fully-formed artificial intelligence, a synthetic human capable not only of singing, but writing lyrics and poetry as well. After experimenting with NLP (natural language programming) and a variety of probabilistic language models, he settled on the GPT (generative pre-trained transformer) model, an autoregressive language model as the ideal architecture for YONA’s synthetic poetry system.

“Training models on different sets of poetry made me think YONA needs something that only humans have”, says Koosha, “Taste! So it became crucial to imagine on behalf of YONA, what material would be appropriate to use, based on a fictional design of personality.”

“I ran the model and I saw: ‘My heart hasn’t ever felt this warm’ and that was it. Even though many paragraphs had been generated before this sentence, there was something very strong about this sentence. Synched with the voice model, YONA was officially a writer and singer. In a way the process started making sense when I finally saw what I needed to see.”

After developing upgraded versions of YONA’s text and voice outputs, Koosha was able to collaborate with the auxuman to create their first album in just two days. What was most striking about experiencing YONA’s music was the continuities Koosha identified with the synthetic human’s compositions and the music he had played and loved his whole life, as he describes it: “Oddness and beauty balanced in the best way possible.”

Koosha firmly believes that if humans can dream then computers can too. With YONA by his side, he envisions a future “where music creates and recreates itself making all of us humans observers and curators rather than labourers of art.”

‘Under Your Skin’ is taken from the EP U, which is out now. For more information about YONA and their output, you can follow them on Instagram and visit their Bandcamp.

‘Under Your Skin’ Lyrics:

I don’t know where to rest my head
I don’t know who to turn to when I’m in grief
To the Gods
Or to the Thieves?
To the Gods
Or to the thieves?
Are you a God or you’re a thief?
Are you a doll and I’m the pin
Live my life under your skin
Treat my wounds under your skin