“Resonances of historical artifacts in the present” – Chris Salter on his multimedia production “Zangezi”

Chris, with “Zangezi” you used a historical literary text and transformed it into a performance art piece, why did you do that?

Chris Salter: I’m always looking at the connection between the world I’m in and the world that I was not in. While I’m very interested in historical artistic works, I’m not interested in reconstructions, recreations or reenactments. I’m interested in the resonances of historical artifacts in the present.

Velimir Khlebnikov was wandering around with torn shoes in bombed out Ukraine and Russia with his archive carried on his back in a sack while he was writing this transcendental poetry, which is of course very deeply about his historical moment (1917s), and this is something that’s so beyond our kind of imagination and comfort, and the comfortable world we live in as Europeans. We constantly say that we’re in such an unstable period, and this is what also drove me to that text again, although I knew about it for many, many years.

What is the kind of calamity of history at the intersection of the end of the First World War, the end of the Russian civil war, and the Russian revolution which completely transformed the society for good or for bad. In fact, it was Lunacharsky, Lenin’s culture minister, who saved “Zangezi” (published in 1922 a few days after Khlebnikov died) from being turned into waste paper by a commercial printer because the printer was only paid “on credit” by Khebnikov’s friends. Otherwise, “Zangezi” would have been lost forever.

So, in our current situation, you grapple with that historical text, and it was never our intent to try to historically situate it in that historical moment (the Russian revolution), because we don’t live in that historical moment anymore. Also, what is that historical moment anyway?

Do you think a digital space is particularly suited to this layered story?

Chris: Yes! I mean, first of all, “Zangezi” is a text that is unstageable in many ways. This text needs a kind of visual acoustic world, because it is not a dramatic text. But there is also a very strong narrative, and: “Ein Geschichte wird aus Wörtern gebaut wie aus Baueinheiten Gebäude.”

This is the introduction of the play. Khlebnikov says, “A story is made of words like a building is made of construction units. Each individual unit forms a certain set of these structures. The first order narrative is built out of these blocks.” It is architectonic in its structure, a building system. But also very interesting in terms of thinking of language in units, just like a Large Language Model in AI also does. Language is basically broken down into tokens, into smaller units to be analyzed. And this is completely different than language as text on a page.

What we struggle with in staging “Zangezi” is that it is not a traditional theater work. It is told in a series of “planes” – fragments.

You didn’t want to deconstruct the original framework?

Chris: If you deconstruct the original framework, the text would make even less sense than it makes now. If you deconstruct Khlebnikov’s already fragmented structure, the audience will have no idea what’s going on.

Is sound art central to the work?

Chris: Yes, it’s very dense sonically, but also visually.

The two performers, Judith Rosmair, a very well known actress from the “deutsche Sprechtheater.” And Audrey Chen is a “Stimmkünstlerin”. That was the idea to try to mix these different styles together.

And the birds that appear in the work speak onomatopoetically. So, Audrey speaks and then the birds are speaking in weird noises that sound like some primitive language. And then the “plane” with the Gods has them speaking in tongues. The sounds that Khlebnikov wrote down, “Mara-roma, biba-bull! Uks, kuks, el !”, are the sounds of both old Russian words, and then words that become sound, like the rolling of thunder.

In fact, “Zangezi” is almost like a radio play. It was clear from the start that in order to tell this story, we had to marshal all these different media to do that. It really is about using these different kinds of technical media to create a strong atmosphere which can hold up and support Khlebnikov’s crazy experiments with language.

Before I came to Zurich, in Canada where I was professor and artist running a research lab called xmodal, we were very much focused on the anthropology of the senses, trying to understand what sensory effects are in different media and producing very abstract, non-narrative installations which toured around the world. Now I’m coming back to narratives, but we have to find new ways to work with narratives and the technical world we live is also very non-narrative world.

The programme was trying to grapple with these crazy Klebnikov lyrics … trying to force them into a 4/4 system, a regulated, statistically averaging system.

To go back to that sound, did you say that this was AI created sound?

Chris: Pieces of it, yes. Remco Schuurbiers from CTM in Berlin and who served as artistic producer, sat at my house and we spent nine hours with one particular passage from “Zangezi” sticking that into Suno (an LLM-based music generation tool) and I think he generated a hundred songs with a bizarre set of prompts. Many of them were impressive but absolutely dreadful artistically.

Suno is very problematic with artists, because it is an LLM driven composition system which is based on the work of thousands of sound artists, musicians, you name it. You write a prompt and then you pick a setting for “weirdness” and whatever style, and then it generates an entire song. It is quite unbelievable. What was interesting is that the programme was trying to grapple with this crazy Khlebnikov poetry, because he switches between rhyme and verse, and then suddenly it is a totally weird and unique style. The system is trying to force them into a 4/4 system, a regulated, statistically averaging system. And so we used two songs that were quite over the top.

It is interesting from a historical standpoint of looking at all of this, because I don’t know Russian. So, of course, I’m relying on people who know Russian to tell me things about the text. And also, of course, relying on just the basics of machine translation systems to try to understand. Comparing the French, the English, the German, and the Russian translations of Zangezi was super interesting, because Paul Schmitt, who had translated it into English, he was a US-based Russian scholar and actor. And he had a commission from the Dia Art Foundation in the 80s to translate all of Khlebnikov’s work into English. Otherwise, there would be no English translation.

What was interesting with these AI generated songs that we used in the production is that the LLMs in Suno encountered this human poet who already thought like a language model. You might say that “Zangezi” is like one full length machine hallucination!

But using these big data-driven LLMs is problematic. For example, there is a huge problem with archiving, because you can’t repeat anything. We used the same prompt and settings two days later, because we had that only one version and we wanted to change it. And Suno gives you the opportunity to download the stems, thankfully so you can then remix it. Otherwise, you just have the mix. But with the same prompt and settings, you can never get back to the original. If you try to repeat it, you will get a completely different thing. So two days later, I said, okay, it would be great if we could do it, try to do it in English and get the same song. Absolutely impossible. It kept generating things that sounded like Rammstein event with the same prompt, with the exact same sentence, because the data set has been updated or the model has changed.

How does this impact the working process?

Chris:

This is a huge problem, because what do you archive? You archive the prompt at that moment that generated that thing, because 3 hours later the data set may have changed. It is a black box. Where is the archival knowledge in the contemporary transformation? This is a huge question in historical frameworks, because obviously we are assuming that the archive exists out of the artifacts of historical processes.

When you’re trying to reconstruct artworks, especially technological artworks, you have instructions about code that might have been written 15 or 20 years ago and then you try to run it or at least reconstruct it on a similar operating system or some kind of software emulator. But the AI systems are absolutely opaque to what knowledge you are capturing. Because the prompt is just human convenience for a language-based system to just start the process. To the machine, ultimately this is just starting input in human language. For the machine its ultimately irrelveant because computers turn language into massive matrices of numbers to represent language.

How do you go on from something that we understand is human, human-generated – which is, of course, symbolic – and that then gets turned into other symbols, which are mathematical symbols, which have no relationship to the original at all?

How do you go on from something that we understand is human, human-generated, which is, of course, symbolic, that then gets turned into other symbols, which are mathematical symbols, which have no relationship to the original at all. Which is why no one can understand really, even scientists, like why did this group of neurons focuses on some particular feature in a massive training set, when it wasn’t told to do that. That is the nature of the complexity of the model. Interestingly, in the process of working on “Zangezi” these questions arose. They are obviously not in the original text, but we try to make sense of Khlebnikov’s manipulation of language in 1922 in our present. At the same time, there are also very current themes in “Zangezi” – particular ecological and planetary.

Where does that come from?

Chris: There is already with the Russian futurists this tension between understanding what the natural world was, and then the world that was destroying that. So that is why they were very much against the Italian futurists’ brutality of machines and war.

For the Russian futurists, a short-lived movement, there was this utopian belief. And there is very interesting work from an Indian historian who is an expert of Russian futurism in Berkeley, Harsha Ram, he wrote something very interesting about the planetary vision of the futurists.

And what hold this project together?

Chris: There is this belief that there is a kind of universality that links people together that is not dividing them through culture or language. That is why the Futurists and specifically, Khlebnikov wanted – a language of sounds that was no longer about its semantic framing. I thought, in relation to this, let’s bring then different artists on to the project from around the world to contribute their own interpretations of “Zangezi” from African, Asian, South American perpsetives, so we have a multi-perspectival view of this text. In other words, a very interesting puzzle as a historical artifact, deeply embedded in the practices that we took on in trying to realize “Zangezi” 104 years after its publication.

Thanks very much, Chris!

Curious for more? Then check out the website of the Immersive Art Space at ZHdK.