Meet technoshamanism – a counterbalance to AI’s use in state and corporate control and surveillance. Hexorcismos is transforming technology into something mystical – and producing beautiful images and music, at once both ancient and futuristic.
ፕዪልክነቻጎኗሁዪልርጎዐ́ክ (Transfiguración) is a stuttering, powerful trip through rhythmic collisions and echoes, beneath a thick sheen of cascading digital artifacts. It’s not entirely unfamiliar, these AI-deconstructed snippets. But in Hexorcismos’ hands, you really get the feeling of music unearthed on some alien planet in a lost futuristic artifact, reconstructed by shipboard computer before hypnotizing the entire crew into a rave in the mess.
That planet is Earth, though – Mexico, specifically. And across Transfiguración, the resulting compositions span from gentle whisper to violent roaring ritual.
Notes from the artist – now Berlin-based, American/Mexican technologist and creator Moisés Horta Valenzuela:
ፕዪልክነቻጎኗሁዪልርጎዐ́ክ (in English, transfiguration) is my latest release; it takes a step further my research of AI for electronic music and uses a GAN for synthesizing an album in the style of my own music.
Conceptually, the idea of transfiguration has been present since ancient times and refers to the ritualistic process of transforming, literally or symbolically, one object into another, giving new functions and meaning to it. New AI technologies, such as the ones used for this album, make this a reality.
But what happens when you transfigure one musical epoch with another; prehispanic and acoustic into electronic, synthesizer, and noise-driven techno, with the help of AI?
This question is what this album tries to answer, by transfigurating the classic Mexican album Templo Mayor (1982) into my own electronic music productions. All tracks have been generated by inputting each individual composition from Templo Mayor by Antonio Zepeda into the GAN.
The result is the (first?) 100% GAN generated album. All tracks have been treated with minimal subtractive EQ’ing and slight compression.
I suspect it’s not the first 100% GAN-made album, but I think that doesn’t matter – it might be the first you actually want to listen to. The machine learning approach employed here:
The source material, Zepeda’s Tempo Mayor, is itself worth exploration – a combination of artifacts and imagination in an acoustic context, but much like the same brew Moisés applies in the digital realm. Fanciful and mysterious, the album finds nuanced cycles of patterning and gesture that blossom organically as though emerging from the instruments themselves. In short, it’s the perfect basis for exploring the organic and unexpected possibilities of AI. (The titular temple, Tempo Mayor, is itself a fantastic place – an archaeological site in the middle of Mexico City. It was my first stop when I went there a few years back, and it immediately positions a visitor in the layers of history and current urban density, all at the same time.)
There’s a video for the opening track, too, edited by Anna Phaenarete Lioka and featuring AI-produced artifact-hybrids, made by StyleGAN2 training of Prehispanic objects from Mesoamerica.
Instead of using AI to be predictive and controlled, that is, here we have AI as fertile soil for growing unexpected discoveries.
The notion of decolonizing AI has a literal meaning. (In case there’s any doubt, one paper on the topic admonished readers that it’s not a metaphor right in the title.)
Self-ownership does take on a personal dimension for almost anyone. Since AI is trained on data, there’s the practical question of what dataset you use. If the algorithm is pre-trained, it can steer your use toward the contents of that data. So like a growing number of artists, Moisés has worked with his own choice of inputs and his own work. (I wasn’t using GANs, but I tried a similar approach myself. And there’s a different feeling to hearing echoes of your own music in the output of an AI – because you’re presented with a smeared and syncopated mirror of your musical self. It feels more personal from the creative side, as I think it must to the audience, too.)
If AI can be a personal voice, it can also be a voice of resistance. In processing and reimagining objects, far from obliterating the source material, the artist is both reinventing and reclaiming them. Here’s his statement on the work (plus the mix below):
I feel these works, the album and this mix, is about mnemonic resistance, about memory. The colonial processes that happened in Mexico and in most post-colonial countries is that of epistemicide (erasure of indigenous cosmovisions). The remnants of the cultures that have survived throughout the years in the face of capitalistic modernity have adapted, either through embracing or hybridizing with the colonial cultures. For example, in Mexico a lot of the contemporary religious imagery harks back to the Aztec polytheistic theological vision which continued into the colonial and modernist times as a specific trait to workship not only the “main” god but also all other saints, having special and unique commends to each one. A trace of the polytheism still lives on within contemporary catholicism in Mexico, though the special workship of these saints.
Using this as an example, these sound works pose the questions of how to embrace new western technologies while not erasing or falling into normalization of culture?
If these questions seem open-ended, that’s part of the power of this practice – it’s meant to be that way. And machine learning is then an ideal aleatoric tool. As Eric Ritskes writes, “the struggle for decolonization is a journey that is never finished and that, on this journey, uncertainty is not to be feared.”
Musicians, by reclaiming ownership of AI as technique and by de-centering from European colonial inputs, are able to act out in a personal and creative way this idea of decolonization. And it isn’t just “political symbolism” – partly because as in the work here, there’s an actual output.
There’s real transformation – well, in addition to literally titling the album transfiguration. Shakir has an engaging overview of the topic and the particulars of its overlap with AI, and writes:
More worryingly, decolonisation increasingly seems to be used as a replacement for Transformation. The price of transformation cannot be paid by allowing ourselves to be distracted by the language of decolonisation and delaying the work of deep social, institutional and personal responsibility for change.
That article is a great introduction to what decolonization can mean in this context:
More details on how the album was produced (and why):
Approaching this work from a technoshamanistic conceptual framework, ፕዪልክነቻጎኗሁዪልርጎዐ́ክ treats the GAN (Generative Adversarial Network) AI architecture as a kind of alchemist tool able to morph one sound aesthetic into another, while preserving the original input’s cadence and long form structure.
The audio datasets used by the AI to synthesize ፕዪልክነቻጎኗሁዪልርጎዐ́ክ were:
Templo Mayor by Antonio Zepeda (1982):
Liner Notes and link to album: https://www.youtube.com/watch?v=eynosej7du4
“The music of Templo Mayor (Main Temple) was entirely composed and played with ancient Mesoamerican instruments and ethnic Mexican instruments. In these recordings no electronic effect has been added, except the over-dubbing of individual channels, and some echoes to create a balance between the different instruments, as well as the desired atmosphere. The various bird voices and the sounds of wind and water have been created only with pre-Columbian musical instruments without any special effect or electronic gadget.”
ℌEXOℜℭℑSMOS – Various Tracks from
OKUPA! [2017-2019] https://hexorcismos.bandcamp.com/album/okupa-2017
Desesperanza  https://hexorcismos.bandcamp.com/album/desesperanza
Codex Enteogeno  https://hexorcismos.bandcamp.com/album/c-dex-ente-geno
The decision to use Antonio Zepeda’s Templo Mayor album as the “input audio” is that of memory and resistance. Templo Mayor by Antonio Zepeda marked one of the first albums to perform and record prehispanic sound artifacts and musical forms with media recording technologies. The practice of using only pre-hispanic sound artefacts as a call to not forget the colonial process and keep these sounds alive through a general widespread of their knowledge acts as a form of resistance, adding to the fact that, since it was performed in 1982, in a musical ecosystem of other seemingly de-localized styles of music, this album is part of a bigger cultural portrait of the pre-1985 Earthquake México. For example, the artist Antonio Zepeda also collaborated with Jorge Reyes in future releases, another musician experimenting with electronic music and prehispanic sounds as a way to bring these prehispanic ontologies of sound and cosmovisions into the digital future, back in the 1980’s.
The reason to use my own music as the raw sound immateriality to be transfigured by the AI from Antonio Zepeda’s is one of possibility, how can a “statistical hallucination” of the information contained in a recording of prehispanic sound provide an insight into contemporary electronic sound practices in the context of migratory modes of sound production. By history and design, AI is a conceptual framework and technology originating in the military-industrial complex in the Global North; how can I as an artist from Tijuana take hold of these technologies and represent histories of resistance as well as critically approach these seemingly “deracialized” and “neutral” post-human narratives? ፕዪልክነቻጎኗሁዪልርጎዐ́ክ is an album that tries to answer those questions and pose further ones into the development of AI as a new tool for artistic and knowledge production.
Moisés has also been producing AI-generated artifacts by training on pre-Colombian archaeological sources. So, from this:
You get things like this:
It’s “generated output from NelkotoniGAN, a GAN trained on Nelkotonis,” he says. Nelkotonis means ‘amulet’ in the Nahuatl language, he tells us – that’s one of the native languages spoken by people in the valley of Mexico.
The cover image, too, is generated this way – at left, a seashell, at right, the NelkotoniGAN-produced output. The seashell also has a sonic significance – it’s a horn used in ritual practice.
And it’s possible to generate more music with the GANs, too. We get the exclusive premiere of an AI-made mix – not a DJ mix, as such, but an AI-produced musical output.
As Artificial Intelligence development continues, which narratives and cultures will be represented in this new medium?
𝖈𝖑𝖚𝖇_𝖙𝖗𝖆𝖓𝖘𝖋𝖎𝖌𝖚𝖗𝖆𝖈𝖎ó𝖓 is a response to the increasingly Global-North centric narratives within Deep Learning in music research.