No-one quite does creativity like HOLLY HERNDON. The composer and academic, who splits her time between her adopted working environment of Berlin and California (where she has just completed a PhD at Stanford), approaches the process of making music like no other. For her, creativity is a sacred act, and the choices that we make at each step along the way are both infinitely fascinating and ripe for being interrogated, deconstructed or flipped on their head.

Take her latest album, PROTO, for example. The process for making this LP began in 2018, when Herndon and her regular collaborator Mathew Dryhurst invited their new creation – an AI ‘baby’ named Spawn – into their home-cum-studio. The two were interested in seeing if they could teach a machine to sing from scratch, and so they set about training Spawn on a select range of inputs. There’s been a proliferation of AI music-making software released onto the market in the past five years, machine-learning systems that have been reared on the whole history of popular music. These systems can spot patterns and mimic human creative tendencies to an alarmingly accurate degree. But this is not what Herndon and Dryhurst wanted in Spawn; they wanted a fellow collaborator, another member of their ensemble to bounce off creatively.

HOLLY HERNDON Image

Since starting her career in music around six or seven years ago, Herndon has always been fascinated by the capabilities of the voice. The universal human instrument has been the basis for her experimentation, and the same was true for Spawn’s education. Human interaction and collective experiences are the kind of inputs that AI programmes are rarely subjected to, and Spawn was really put through its paces during the making of PROTO. As well as involving Spawn in a public AI training performance in Berlin in 2018, Herndon and her ensemble subjected Spawn to Appalachian Sacred Harp music – a form of shape note singing (employed to startling effect on Frontier).

For all of this apparent artifice, the record as a piece of work is startlingly organic. It is built upon a range of ‘folk’ music stylings that are indecipherable from the machine input – which is kind of the point. The depth of the feedback loop in the creative process is such that it’s hard to pull apart each of the individual strands. Machines learn from humans, humans learn from machines and other humans, and the cycle keeps going round. So much so that you forget to question what is coming from where – you just experience it.

 

You’ve been working on this project for a while, and now AI technology is at a level where lots more people are utilising various AI tools in their own music. Looking back to the outset, how did you initially approach the process?
HH: First, there’s the idea of creativity and there’s the idea of how it will be used – and I don’t think it will always be used creatively or with creativity in mind. I think a lot of the stuff that we’re going to see the market flooded with is kind of like automated composing, very genre-specific… anything that’s really idiomatic or there’s a clear rule-set, those are things that AIs can very efficiently reproduce. So, like a Hans Zimmer score or a piece in the style of Bach, that kind of automatic writing? I don’t think that’s interesting, and that’s kind of why we decided to take a really different approach. We wanted to hear the neural network at work in the audio itself. Rather than using it to write, kind of, an idiomatic piece of music that then we push through a digital instrument that sounds kind of perfect and glossy, we wanted to actually hear the roughness of the network itself. So that’s why we dealt with audio data [rather than midi data]. AI has been trained on human labour, it’s all been built with machines that come from rare earth materials from the ground, and we sometimes see these things as an alien intelligence that just, whoa, landed from another planet! No, it came out of us and through us and out of our earth and we’re responsible for how it evolves as well.

So in terms of the inputs that you gave to Spawn and the parameters you set, was part of the thinking to what you put into it that you wanted it to come from humans?
Yeh, I mean, it’s not strict. Like, we train on other things as well but we did focus a lot on the voice because we wanted the community that trained Spawn to be audible. So that’s why, from the beginning, we were like, no, we create our own data sets, we’re not using other people’s data. Fundamentally, that’s a rule. We’re not training on huge audio libraries or things like this, which most of the corporate AI stuff is doing – like, oceans and oceans of data. We’re deciding what the canon is, what the training data is. And that’s very much what affects what the output is. It’s only limited to what it’s fed.

Were you ever surprised in any way about what came out of the process?
Yeh, I was super surprised by Godmother. Because I didn’t train on beatboxing. It’s very beatbox-y.

It’s quite rough and different and patchy.
Totally. But we wanted to show it in its current state, we weren’t trying to make it perfect and pristine. But also the delivery that she’s doing, it’s like, I didn’t train on that delivery. It was on speech and singing. But then, when interpreting like a drum sound, she turns that into like a beatboxing sound, which makes total sense. It’s a bit of a trick of the imagination that that’s creativity, but it was something that surprised me.

“How can technology free us up to be more human? That’s the goal” Holly Herndon

One of the potential real benefits of using systems like this is that they can perhaps spot and isolate patterns that we’re too close to be able to see, which can be applied to music creation.
I agree. If we use it in an interesting way, if that’s, like, the ethos. If we look at it as a different kind of intelligence, that we can learn something from, then I think that’s a really rad way of approaching it. Rather than it just trying to emulate us.

You said in the past “I don’t want to live in a world in which humans are automated off-stage”. Do you think that there’s a danger of that being the case here?
Well, that already happened. What is an AV DJ set, if not the composer and the performers entirely automated out of that performance system? I mean, that’s an alienation of the music from its source.

But how does that extend to you performing it with the same kind of spirit from a visual point of view?
We’re all kind of in this moment together and we have to make it good together. And that’s why I love live music. But I’m not saying, ‘Oh, 1975 was the heyday of music’. Like, that’s bullshit. I’m not trying to go backwards and have some reverence for some perfect period of time. I’m like, how can we use all these crazy tools that we have access to? How do we create a symbiotic thing where the humans are not automated out of it, that the brain, the computer, frees up the humans to do even more emotional work on-stage? How can technology free us up to be more human? That’s the goal. Not how can technology dehumanise us? That was never the goal.

This view is different from what the majority of people think, that the computers are coming to dumb us all down and flatten some edges off, maybe make things a little bit easier for us in terms of efficiency and production.
Yeh, and it’ll replace a lot of jobs, but hopefully we as a society can figure out a way to value people outside of whatever they’re being valued for, whatever job’s being replaced. I think that’s a wider societal question that we’re facing right now.

And so, on that, do you think we as humans need to think differently about how we interact with machines and I suppose AI systems, with a little bit more openness?
I think so, and I think we need to also figure out how we interact with each other and how we value each other and how we share resources. If you can train AI on an entire canon of musical history and then recreate that or create kind of a new version of it, who owns that? How do we share that? Is it just whoever is running that model and has access to that data, should they reap all of the rewards of the entire history of human music and thought? Private corporations have access to the widest data pools because we’re constantly giving them up for free through social media, and so how can we all then benefit from this thing that is actually a shared human intelligence that, with every gesture and effort, we’ve all been developing as a species together for the last however long.

Bido Lito Embeded Video

In a broader sense, why should we be hopeful for the future?
[laughs] Oh. Why should we be hopeful? I don’t know.

I mean you may think we shouldn’t be.
No, I think we should, because if we are not hopeful we have already given up. I mean, we have to feel like we have a sense of agency in this world, and in order to have agency you have to have a glimmer of hope that you can in some ways choose your own destiny or at least influence it with your community. And if we’ve entirely given up that agency and we seep into this cosy cocoon of nihilism then we are really handing over our future and that of our species to someone else.

Has this process surprised you at all? Or, did you have any trepidation coming into it about working with AI?
Well, the thing is, is it’s a very hyped thing right now. So I was a little bit worried about it being this kind of buzzword. And I do see that. But I wasn’t scared of interacting with AI. I mean, for the first six months we had really bad results, and so I was like, ‘Fuck. Maybe we should do something else or try something else’. It took a lot of patience and time sticking with it to get something interesting.

I suppose, it’s one thing to say that we as humans should bring ourselves to the technology and embrace it and think completely open-mindedly, but there’s another thing strand on the opposite side, that humans can also make bad decisions as well.
For us it was really just like a different way of performing material. It’s like, she’s performing and interpreting different pieces of compositions that we’re providing for her, and then that becomes part of the overall ensemble performance of the piece. And there are aesthetic decisions that go into that as well, that give a specific kind of colour to the record.

"If we’ve entirely given up that agency and we seep into this cosy cocoon of nihilism then we are really handing over our future and that of our species to someone else.” Holly Herndon

I suppose, it’s one thing to say that we as humans should bring ourselves to the technology and embrace it and think completely open-mindedly, but there’s another thing strand on the opposite side, that humans can also make bad decisions as well.
For us it was really just like a different way of performing material. It’s like, she’s performing and interpreting different pieces of compositions that we’re providing for her, and then that becomes part of the overall ensemble performance of the piece. And there are aesthetic decisions that go into that as well, that give a specific kind of colour to the record.

And then the decision to bring in the Appalachian Sacred Harp music to the ensemble – was that just an aesthetic inspiration, to give Spawn a different look at other styles?
Well, there is a logic [laughs]. When you start looking at artificial intelligence and you think about the evolution of intelligence as a general topic, you start thinking about the role of music in the evolution of the human brain – and there’s a whole field of musicology that deals with this. Kind of divergent ideas about what role music and rhythm and choreography played in our own brain development. And so, looking at early vocal styles that came out of that, folk musics that sprung up in different parts of the globe, in communities that would not have been in touch with each other that separately evolved these kind of musickings in disparate cultures. The things that felt like they were kind of in our human DNA, that sprouted up in different ways in different cultures. So I was looking at a lot of folk music and some of this more nasal delivery, and being from the American South I wanted to tap into this Sacred Harp kind of music. One of the ensemble members, Evelyn Saylor, actually is in a Sacred Harp group in Berlin. So she played a huge role in helping me arrange the voice parts so that they sounded authentic.

And that was a way of, kind of, tapping into one of those primal essences of humanity? I think we probably learned to sing before we learned to speak, didn’t we, and it helped us form communities.
And helped us develop tools, one theory goes. Gary Tomlinson, he has this idea that we used song and choreography to entrain and coordinate, to hunt and to create tools together. Not sitting down to, like, build a hammer, but it’s this embodied choreography that then impacted on our brain and then our brain developed because of the technology and this kind of feedback system… that’s why I don’t see the neural nets as this alien species. It’s part of the hammer. It just feels more alienated, but it’s really not.

 

hollyherndon.com

Holly Herndon’s new album PROTO is out now, via 4AD. Holly Herndon appears at the Queens Of The Electronic Underground event at O2 Ritz on 20th July, as part of a showcase curated by Mary Anne Hobbs for Manchester International Festival.

RELATED
CURRENT ISSUE Bido Lito! Issue landscape Ad PLAYLIST