ALLMUSIC: What made you wish to work with AI within the first place?
CLAIRE L. EVANS: It’s humorous fascinated with it now—why did we even wish to do that? I’ve to maintain reminding myself that 2016 was a really totally different time, that we weren’t influenced by every thing within the air now. It was on the horizon, and we needed to study it. We had a sense that it will turn out to be vital, even when we didn’t anticipate what it will turn out to be. I learn a whole lot of science fiction, and I had an concept that, at the least within the cultural consciousness, individuals could be unpacking this for a very long time. And we wanted a problem. We wanted one thing to try this was new, that was attention-grabbing, that might be creatively partaking for us and would pressure us to make one thing totally different than we might ever made earlier than.
It would additionally require a whole lot of collaboration, which was thrilling. It appeared enjoyable to do one thing that might contain asking for lots of assist. We had no concept what we have been stepping into. We thought that you possibly can “train an algorithm” by yourself music and generate extra of that music. We had each a too optimistic and too pessimistic view of AI. We thought it will be a system that might generate sounds on our behalf. Then we realized it wasn’t even attainable to make a foul tune with the know-how accessible on the time. So then the undertaking turned about effectively, how can we make one thing good? Which is far more durable.
ALLMUSIC: When did this all begin coming collectively, because the album got here out in 2019?
CLAIRE L. EVANS: The documentary began filming in 2017. We began on the report in 2016, attempting to determine how you can do it, assessing the instruments accessible on the time, which have been restricted. There have been open supply fashions within the laptop science world that you possibly can tinker with, if you happen to knew how you can code. There have been nascent music startups that stated they have been utilizing AI, but it surely was unclear what was actually occurring beneath the hood. Then there have been artists and technologists constructing their very own instruments, constructing their very own fashions. That was the panorama. So we had to determine what we needed to do, who we might collaborate with, and what instruments we might use that might enable us some measure of management over what we have been doing, although we weren’t coders, which took a very long time.
ALLMUSIC: You began fascinated with this and 2016. Now it is 2024. That’s 1,000,000 years within the tech world. Have you used any of the brand new instruments? If you have been to undergo this course of now, would it not be simpler?
CLAIRE L. EVANS: It could be quite a bit simpler, but it surely would not be as attention-grabbing. The incontrovertible fact that it was cumbersome was what made it compelling to us. The instruments accessible have been unhealthy at doing what they have been ostensibly meant to do. A text-generating mannequin would generate absolute nonsense, a sound-generating mannequin
The movie known as The Computer Accent as a result of every thing AI-generated at the moment had a sort of an accent. You might simply inform it wasn’t human—all of it had an identifiable weirdness.
would generate one thing both completely haunted or with such a low pattern fee that it sounded too lo-fi even for us. The enjoyable was taking all that chaotic materials and seeing if something attention-grabbing had by accident occurred, and attempting to restructure all of the chaos into one thing significant. Our job was to provide it construction.
Now—the instruments are so good. There’s not a whole lot of friction there. AI doesn’t make attention-grabbing errors. What we have been drawn to initially was the oddness of the fabric these fashions generated. The movie known as The Computer Accent, and we do not even discuss this within the film, however we known as it that as a result of every thing AI-generated at the moment had a sort of an accent. You might simply inform it wasn’t human—all of it had an identifiable weirdness, a wonkiness. That was charming. That’s sort of misplaced now. I’m by no means interested in any of those new instruments. They do not present any attention-grabbing texture. It’s identical to, extra human mediocrity.
ALLMUSIC: For some time I used to be utilizing DALL-E to attempt to recreate well-known album covers with prompts and simply seeing what would occur. Iconic issues like, “a yellow banana on a white background,” simply to see what it will do. It’s far more enjoyable when it is goofy and you do not actually know what you are going to get. But when it begins getting higher, it loses that bizarre/creepy issue.
CLAIRE L. EVANS: Yeah, completely. I feel we’re gonna miss that. One of our theories early on was that the “neural aesthetic” of early generative AI would turn out to be the brand new analog. That we’ll have nostalgia for the wonkiness and fucked-up weirdness of those early fashions—and I feel we already do.
ALLMUSIC: One of my overarching questions is what did you surrender and what did you acquire on this course of? Did you’re feeling such as you gave something up by letting the pc spew out a bunch of sounds which are presupposed to sound such as you?
CLAIRE L. EVANS: It undoubtedly wasn’t mirroring our course of. It was interfacing with our course of on an extremely floor degree. All the AI “knew” about us was the MIDI knowledge we fed it, which was stripped of context. We could not put an MP3 into these fashions. We had to attract out two-bar chunks from multi-tracked songs, which could have dozens of various patterns and melodies.
The AI was working on such a minute piece of us, and something it put out would essentially solely signify a fraction of who we’re.
Normally we might go into the studio with a pocket book filled with lyric concepts, a couple of melodies swimming in our heads, massive ideas, or boneheaded riffs—that might be the supply materials. We’d nonetheless have to rearrange and put it collectively, making all these decisions which are depending on context. That course of remained precisely the identical. It’s simply that the supply materials had been run by way of the blender on this high-dimensional, conceptual approach. The actual distinction was that there was rather more of it. With these fashions, even on the rudimentary scale at which they have been working again then, you’d simply hit “go” and get 40,000 phrases. So we needed to make aggressive choices about what stayed and what went. That was perhaps probably the most alienating and overwhelming a part of it, really: letting go of the sensation that one factor may be higher than the opposite.
We might give the identical corpus of textual content and notes to 100 different bands, they usually’d all make actually totally different data. The supply materials is just the start of the method. I feel that is a very good angle to take in the direction of generative art-making: you should never take the output as the top of the method. The output is simply a part of the method. You cannot simply slap your title on it and name it accomplished. I imply, that is no enjoyable. I’m not even speaking in regards to the ethics of it. It’s simply not attention-grabbing, or enjoyable, to try this. I do not assume no matter time I’m saving is value it. What would I be doing with my time if I wasn’t making music, or writing? Why not simply make the music? So yeah, I do not assume we gave up an excessive amount of, apart from flexibility, by advantage of the truth that the undertaking was so structured. Everything needed to be actually pure conceptually.
ALLMUSIC: So what have been the principles then?
CLAIRE L. EVANS: Everything needed to be generated, indirectly, from our personal again catalog. Obviously, these fashions require extra knowledge than even our 20 years as a band can present, however the prompts needed to be from our personal historical past. That was the very first thing. The second was that we could not improvise, jam, harmonize, or generate something on our personal. If we needed one thing particular for a tune, we had to determine how you can make the machine make it. Sometimes that might imply, “shit, we want a bassline that sounds like this—so let’s take this melody from this song from 10 years ago, and this melody from this other song of ours from five years ago, and run it through a latent space interpolation model, and hopefully, the model will split the difference and give us what we need.” That was ridiculously difficult, however we handled the entire album as a science experiment. None of the fashions we used might generate chords on the time. So there are virtually no chords on the album.
What else? We might transpose issues to a unique key. And we might hack issues up as a lot as we needed. We did a whole lot of collaging, taking little MIDI riffs from totally different outputs and arranging them collectively. Especially with the lyrics, that was actually attention-grabbing, as a result of we have been working these fashions at totally different temperatures. The excessive temperature output could be a lot wackier and extra verbose and filled with neologisms. The low-temperature stuff could be repetitive, punk rock fashion. Mixing the totally different temperatures inside a single tune was an attention-grabbing approach of approaching verses and choruses. We gave ourselves leeway with association, manufacturing, and efficiency. Everything’s performed stay. But by way of the supply materials, we have been actually inflexible: it needed to come from the system, and it needed to come from our personal historical past.
ALLMUSIC: So the method of creating what you have been going to select was largely instinct? Listening to issues, figuring it out, discovering what you preferred, arranging it. You used the phrase collage.
CLAIRE L. EVANS: There’s a wealthy custom in avant-garde artwork and music of doing cut-ups, doing collages. Mixing sources. David Bowie used cut-ups for his total profession to write down songs; he even had a pc system he made within the 90s that might lower up newspaper articles and recombine them. It was known as the Verbasizer. Something actually attention-grabbing occurs if you use an digital or generative course of to floor materials recombined from a unique supply. You’re pressured to interpret it, to impose which means on it.
The course of of creating which means is what artwork is all about, however that meaning-making can exist at totally different occasions, and from totally different positions. You could possibly be issuing it from your self, or you possibly can be deciphering xand projecting which means onto one thing else. In music, it usually goes each methods. You might write a tune about one thing actually particular, however then everybody that listens to it thinks it’s about one thing else. I might sing the identical tune 1000 occasions stay, and it might imply one thing totally different to me each time, relying on the place I’m in my life. That’s all the time altering. I feel it is simply actually enjoyable to play with that in a extra specific approach. To draw from these avant-garde histories and discover methods to undertaking which means onto chaos, basically.
ALLMUSIC: I’ve been making area recordings, issues like my buddies speaking at their report retailer, or a dialog taking place subsequent to me whereas I’m getting espresso. Incorporating a bizarre little quote right into a tune. Can that even be replicated by the instruments which are on the market? Pop music has a construction. Producers have formulation on how you can make a success. And that is why they’ve a number of hits. But perhaps it is bland as a result of it appeals to a quite common denominator. Working with cut-up samples, that’s one thing attention-grabbing to me. Can machine studying replicate that? What would that sound like?
CLAIRE L. EVANS:
A pc can definitely replicate that on a proper degree. If you prepare a machine studying mannequin on a bunch of area recordings, and collage-based experimental music, it will generate extra stuff that feels like that. But will probably be issued from nothing, completely divorced from context. It can by no means replicate the second if you recorded, on the planet, your mates speaking within the report retailer.
ALLMUSIC:
Lots of innovation comes from the restrictions of the method. Even the report you made, you have been restricted by the know-how that was accessible on the time, and if you happen to have been to do it at the moment, it in all probability would not be
the identical. A band like Beat Happening would not sound the identical in a flowery studio. Early hip-hop can also be primarily based on this limitation of course of. I ponder if these items could possibly be replicated with machine studying?
CLAIRE L. EVANS: 100% agree with that. Again, the machine studying methods might proceed to iterate primarily based on present varieties, which had emerged from limitations and constraints, but it surely couldn’t create new varieties from those self same constraints. All good artwork comes from limitations, and is made by people who find themselves questing past their means. That second of striving past your capacities or your instruments, attempting to do one thing that transcends your state of affairs, or what you’ve gotten entry to. That’s probably the most lovely human gesture. When you’ve gotten every thing, although, you don’t have anything.
ALLMUSIC: It’s talked about within the documentary about how Chain Tripping did not get a whole lot of press. Did individuals not perceive what you have been doing with the report so far as the idea? Do you’re feeling like individuals weren’t totally prepared to reply to it, and if it got here out at the moment, perhaps the tradition could be prepared?
CLAIRE L. EVANS: It’s all the time exhausting to know. When the press ignores one thing it’s like, is it me? Or is it the thought? We did undergo an actual shitstorm within the press that in all probability bought us blacklisted from protection on some music web sites. I do not assume individuals have been then, or are actually, tremendous eager to see us as conceptual artists, regardless that we’ve been doing bizarre tasks for 20 years.
That second of striving past your capacities or your instruments, attempting to do one thing that transcends your state of affairs, or what you’ve gotten entry to. That’s probably the most lovely human gesture. When you’ve gotten every thing, although, you don’t have anything.
Our residual public picture is as an indie-pop, indie sleaze band. I assume artists are all the time attempting to flee their personas. It is what it’s. I’m completely blissful making the artwork I prefer to make. The folks that prefer it prefer it. That’s all I actually care about.
It was in all probability too quickly, regardless that we felt prefer it was too late on the time. We thought we might missed the window. It’s not like individuals weren’t speaking about AI. Artists have been partaking with the topic, they usually had a lot clearer narratives round it than we did. We selected a course of that was attention-grabbing to us, but it surely was not an excellent compelling course of narratively. And the capability that the music press has to speak this topic is restricted. Because the nearer you get to AI, the extra it’s simply difficult, costly, boring math. That’s not as enjoyable as saying that an AI is your new bandmate, your clone, or coming to your job.
I do not know when a very good time is. The movie is out now, and I’ve been struggling to determine how you can discuss it, as a result of I nonetheless really feel prefer it’s too quickly, in a approach. Now the movie is a time capsule of a second in time—not way back, however 1,000,000 years in AI years. The outdated DFA Records motto was “too old to be new, too new to be classic.” I feel that is sort of the place we are actually. I’m to see how individuals reply to it. I’ll say that much more persons are fascinated about speaking about it with me now. I get inquiries on a regular basis, and I really feel like I’m like a veteran of some outdated AI guard.
ALLMUSIC: How do you’re feeling in regards to the report now?
CLAIRE L. EVANS: I really like the report. I feel it is the very best music we have ever made. Partially as a result of it liberated us from our personal persona. I feel we had this concept that we needed to be a sure sort of band, earlier than that report. We needed to make sure sorts of songs for sure contexts, to keep up a repute as a poppy enjoyable electropop social gathering band. Dropping a bizarre burner album with a whole lot of gradual, formally experimental songs—and I feel it is fairly restrained, too, extra delicate than something we have ever made earlier than—was so releasing.
ALLMUSIC: The truth that you’ve been a band for thus lengthy is a testomony to one thing for positive.
CLAIRE L. EVANS: People within the AI music world, again after we have been making Chain Tripping, would say stuff to us like, “with AI, you’ll have a fourth bandmate that never drinks, never parties, never gets into trouble, and always agrees with what you say.” That’s by no means what I discovered attention-grabbing. What I discovered attention-grabbing was that we had this interface that might generate concepts, and we felt completely high quality dismissing these concepts, as a result of there was no ego concerned.
And as a result of we have been all working in the direction of making an album inside this conceptual framework, we have been solely fascinated about discovering the concepts that might most swimsuit the tune. Normally if you work with a gaggle of individuals, as I’m positive you realize, individuals get hooked up to no matter they dropped at the desk, they usually can have a tough time letting go of issues, even after they don’t serve the top. But we had this mannequin issuing concepts and we might simply be like, “nope, nope, nope, nope, nope—okay, that one’s good.” All with out feeling like we needed to coddle anybody. It really made us higher collaborators as a result of we have been all united in working in the direction of a typical purpose, if that is smart.
ALLMUSIC: Right: you’ll be able to’t harm the pc’s emotions. One factor that drives me loopy is the oversimplification of complicated concepts. My feeling on the present state of AI is that “AI” is getting used as a blanket time period for every thing now. Before it was generally “the cloud,” or “the algorithm.”
CLAIRE L. EVANS: Right. There’s not only one AI. There are 1,000,000 totally different fashions, with totally different coaching knowledge, totally different scopes, totally different powers wielding them to totally different ends. What are we really speaking about?
ALLMUSIC: I take into consideration the dearth of and lack of humanity with it. You nonetheless should make your choices. It has to do with the intent and course of and resolution making. What is misplaced if you let a machine do all of it for you? Versus utilizing it as a software, prefer it’s a calculator, a drum machine, or no matter.
CLAIRE L. EVANS:
I feel we’re in a proof of idea stage within the tradition proper now. People are doing AI tasks simply to point out that they are often accomplished. Someone might use AI, say, to generate a comic book e-book. But in the end, that gained’t be as attention-grabbing as an artist utilizing AI to generate a comic book, then taking a panel from that comedian and portray it, or utilizing that as a immediate to do one thing else, say, write a novel, or a screenplay. What I imply is that there must be one thing else within the daisy chain. I feel we’ll get to that time, the place artists are integrating AI instruments into a bigger course of, they usually will not consider it as something totally different from utilizing a synthesizer or utilizing Photoshop. It gained’t all be in regards to the software. That’s my hope. But that sort of reasoned, considerate integration of a brand new software into a bigger artistic imaginative and prescient is gradual work, and this know-how is shifting in a short time. I’m unsure we’ll make it in time. Every time I open the web it’s filled with generated dreck, generated language and imagery, thrown into the works by hustlers and hucksters. We’re all competing with a lot quantity, attempting to pierce the sign.