Holly Herndon Coding the Voice of the Future

Holly Herndon’s music stands at the forefront of technological possibility.

2015’s Platform, created as part of her PhD at Stanford, embraced the laptop as the ‘most intimate instrument’. For her latest album, PROTO, she built an AI ‘baby’ named Spawn to sing as part of a live ensemble, learning from these human collaborators.

You did a lot of choral singing as a kid. Did you imagine it having such a central role in PROTO?

That was probably the start of the process, even before [AI singer] Spawn. I missed the pure joy of being in a recording studio with other people, singing and working through parts together. Emoting and musicking together is something that I grew up with and was yearning for again. But I didn’t know how many tracks on the album would have the ensemble or how integrated it would be. It was a stumbling-through-the-dark process.

Did you always gravitate towards the laptop as an instrument?

I grew up playing piano and guitar, a bit of contrabass, then samplers and drum machines. A bit of everything. But I never really found my voice until I started using the computer at Mills [College, Oakland]. I felt like I wasn’t rehashing something that came before. It felt really liberating. It was a way for me to transcend the limitations of my messy fleshy body and put it in a space that made sense with digital instruments as well, instead of being a layer on top.

Why did you want an AI to sing?

One of the dominant approaches for making AI music right now is to extract the information from existing compositions or data sets and turn it into MIDI information, then train the system you’re using to automatically compose in that style. I find it pretty uninteresting because it gets us in this recursive loop of recreating what’s come before.

From the beginning we wanted to approach sound more as material, coming from a musique concrète lineage, thinking of music and all of the sound around you being abstracted from its origins and AI being the next level of abstraction. I also really wanted to focus on the voices of the people training the AI, to make audible their work. I think human labour often isn’t acknowledged in a lot of AI that we see.

This question of agency seems important to your process. You see AI as something to have its own voice, but also be guided by the humans around it.

I see the AI as a part of ourselves, a part of a community. Whoever’s training the AI is pointing it in a certain direction. I think we have this narrative of AI as something that already has its own agenda. That’s a way of decoupling our humanity from the responsibility of the complex systems we design.

People will say about the stock market, ‘the market will work itself out’. Capitalism teaches us that the market is smarter than any individual. But at the same time the market crashes every ten years and humans have to put it back on track. So it’s seeing ourselves as not separate from these complex systems that we set up, but rather as actors who can point it in different directions. As a community we encode the value systems and the ethics we share into the technology and the systems we develop.

PROTO embraces a positive view of AI and our role alongside it. Have you always had this optimistic view of technology?

It’s part of a longer process. With Platform I was thinking of the laptop as the most intimate instrument. But then with that intimacy, I noticed how I also opened myself up to vulnerabilities through the laptop, to corporate and government surveillance, for example. So there started to be some cracks in that relationship, some criticism and scepticism. Approaching AI with PROTO, it’s not rosy-eyed, ‘technology will liberate us all’ Valley-speak. There’s a critical lens to it, but I’m always trying to imbue a sense of agency. I think if we focus too much on dystopian narratives we can concede power and let the dominant corporate view take over. It’s neither purely optimistic or purely pessimistic. I tried to think about another way to use the stuff that isn’t automating humans out of necessity.

Why do you think Berlin, where you are currently based, endures as a hub for leftfield art?

Berlin’s one of those rare places where people have more time. There’s less economic pressure. But I also think it’s geopolitical, in that West Berlin was essentially an island, housed inside another country. It was a very unique geographic situation. Take the example of the military draft. If you lived in Berlin you didn’t have to do it, so it automatically attracted people who didn’t want to work within a very hierarchical structure, a lot of artists.

West Berlin was an outpost for Western democratic ideology behind the Iron Curtain, so a lot money from the United States went into funding avant-garde music, art as an expression of superiority. As problematic as that is, it funded a lot of really great work and some really interesting institutions. In some ways it lives off that legacy. It attracts new generations of people who belong to that history. But I think the economic realities of it still play a part.

How do you switch gears between academia and music?

If I spend too much time in the academy, like I find it conservative in some ways, then I yearn for some of the communal feeling around club culture. Then I get tired of that scene and I yearn for the more analytical and conceptual approach of the academy and I go back to that. So whenever I get sick of one I go to the other, but I see it as a complement to everything else.

Were you always drawn to an academic path?

I wanted to learn how to develop my own software and I was interested in the history of computer music from an academic perspective. But it’s also an economic reality too. I think for artists to have the room to really be free and fully explore without worrying about their future in this precarious time, it’s often limited to an upper-class kind of exercise. It’s almost like the dream of bohemia is now limited to an upper class, which I don’t think has always been the case.

I’ve never really been that drifting bohemian. I’ve always been really interested in those ‘out there’ things, but I also set myself up in case nobody cares about the music I’m making. PhD programmes in the States are fully funded, so I was able to transition from working full-time to being an artist full-time. Having that five years of funding was really a launch pad for me.

Sometimes people criticise me, like I analyse music too much or they see it as a snobby or elitist thing. But that was my job. It’s funny how this sense of authenticity is sometimes tied to these freedoms that are only afforded to those with certain material comforts.

What’s next?

Well, we’re figuring out how to capture the record experience, how it can mutate into a live version. We’re trying to figure out a real-time system for Spawn. Because the processing time is so long right now, it’s really hard to get a real-time system working. And [frequent collaborator] Matt [Dryhurst] and I are also looking at combining some of our ideas into a single publication. He’s teaching at NYU right now and his area of research is infrastructural music.

There’s a lot of things in the works, it’s just what do you have time to do and what pays the bills. It’s a constant juggling act.