Alt-Pop Trickster 1010Benja Is Not Scared to Admit He’s Using Generative AI
“If you have it broken down as organic versus AI, the AI is gonna win our world. The good guy doesn’t win this one.”
When I reach 1010Benja on the phone, he’s just blocked someone who blew up at him online because of his use of generative AI. “He was going crazy,” Benja says of the irate listener. “He said I was destroying art.” Benja sighs. “I just wish he could understand it from the perspective of an artist: I was born into a world where art was destroyed, and yet it persists.”
The Kansas City, Missouri musician has received several disgruntled messages since he released his latest EP, Time Has Nothing to Do With What You Choose, last month. To make the EP, Benja and a few of his longtime friends wrote and recorded original, organic vocals, beats, and instrumentation, and then used gen-AI tools like Suno to augment, subvert, and generally fuck with those recordings. The results are often too odd—and, to be honest, too genuinely mesmerizing—to simply be dismissed as “slop.” Instead, this music comes from a liminal creative space that artists are just beginning to explore, and that listeners are only starting to wrap their heads around.
“Semiramis’ Dream,” a highlight from the EP, is a cloudburst of drum’n’bass, hyperpop, and shapeshifting vocals. Some of the singing comes courtesy of an uncanny children’s choir; then there’s Benja’s all-too-human voice at the center of the song, pleading lines like, “Kissing you makes me feel clean!” It’s the first song created with gen AI to capture my imagination, and I’m afraid it’s not the last.
To be clear, this is not to say I’m a full-throated AI convert—I’m well aware of how companies like Suno scraped millions of copyrighted songs (including some of Benja’s, by the way) to make their models. I am unnerved by the technology’s disastrous environmental impact. I know that a tsunami of 100-percent gen AI tracks are being uploaded to many streaming services every day, and that many tech companies would love to populate their products with submissive bot music instead of songs made by human artists who demand to be paid fairly. But I also feel like it’s my responsibility as a music journalist to contend with this technology, in all the shapes it takes. Which is why I wanted to talk to Benja.
Part gonzo pop prodigy, part goofball trickster, part aesthetic philosopher, Benja is a dad in his late 30s who’s been singing and playing guitar for most of his life. First in apocalypse-themed church productions and punk bands while growing up in Tulsa, Oklahoma. Then on subway platforms in New York City, where he endured a period of homelessness. For the past decade or so, he’s toiled as an unclassifiable eccentric in a music industry that likes to place artists in neat, marketable boxes. He’s released interstellar ballads, heartfelt folk, avant-rap, explicit vocal jazz, and lighter-worthy pop-rock. He’s got the type of voice most singers would kill for—clear and crisp and soulful—but it often felt like he was subverting his own innate marketability by putting his vocals against these proudly uncool, ambiguously winking musical backdrops.
Benja himself admits as much. “Part of why I’ve struggled in the business is because no one wants to give me the bag because they know I kind of want to make a joke out of it,” he says. “But that’s so important to my art that I haven’t been able to resist it in order to play the game. Humor and terror and existential dread and love and eroticism all live together for me.”
After releasing music under a few different labels—including British tastemaker Turks, home to FKA Twigs and the xx—he found himself without a deal last year. “I figured, if I’m gonna be broke anyway, I’m just gonna rough it and make exactly the art I want,” he says. It’s then that he started really diving into the world of generative AI with a new project called Penis Angel.
As the name suggests, Penis Angel was originally conceived as a joke. (He remembers the first thing he tried to make with gen AI was a “Doja Cat parody song about worshipping Satan. It wasn’t very good—the joke or the gen.”) But as he and his friends kept working with the rapidly advancing tech—making it spit out nonsensical Rat Pack toe-tappers, extremely graphic rap, and over-the-top cock (!) rock—he started to think of Penis Angel as a seriously ridiculous satire of a pop industry that lost its humanity a long time ago.
“Everyone knows there’s been a genuine insincerity going on in popular music for decades,” he says. “No disrespect to Britney Spears or the Backstreet Boys, but AI has been what art has been for a minute. What’s the difference between these songwriters teams and AI? In fact, emotionally speaking, on one level you’re creating more of a direct violence to the artists themselves by putting them in these structures—this sort of human centipede where the artists still don’t get paid. I understand the implications of AI being trained on artists’ work, but none of it’s going to stop. If you have it broken down as organic versus AI, the AI is gonna win our world. The good guy doesn’t win this one.”
The way Benja sees it, using gen AI isn’t a cash grab or even a shortcut. It’s a way to fight back against an industry that’s often left him stressed and skint, and a way to realize the insane musical dreams coursing through his skull. “Don’t get me wrong, I understand the stigmas and the confusion. I’m not a nihilist, and I don’t want the planet to be destroyed either, but let’s check our hypocrisy,” he says. “Artists are so often boiled down to just a product, so in my generative projects I was able to take the devil’s advocate role—not really an advocate, but more just take the devil’s role—and make this program slave for me, and churn out unlimited options, in a way that maybe a label director might have viewed me in the past.”
For a while, he kept his experiments with gen AI within the Penis Angel expanded universe and didn’t use it for his work as 1010Benja. But when the Penis Angel records failed to gain much traction, and he started making gen AI-assisted tracks that weren’t as blatantly parodic, he decided to release this hybrid music under his better-known alias. So far, he’s put out five 1010Benja tracks that were created with gen AI: this spring’s glitchy freakout “Y.T.YN,” along with the four songs from the recent EP. “This is shit that I’m in love with, that my kids asked to hear every day,” he says. “At this point, I’m making things that I’m emotionally invested in.”
Below, Benja goes into detail about how he and his collaborators, Patrick Holder, Sigeppi Muse, and Nick Lastly, made “Semiramis’ Dream,” and offers more thoughts about the stigmas and possibilities associated with generative AI.
Gain access to and everything we do at Hearing Things—including podcasts, album reviews, and more—by signing up for a free 30-day trial.
I want to drill down on “Semiramis’ Dream” because this song really struck me. Did you use Suno to make it?
1010Benja: Yeah, that’s the main one we like to use. But Suno is not what people clearly think it is. It’s very hard to ride that bull. It doesn’t do what you want. You have to put a bunch of hallucinations together and try to triangulate them to get what you want.
Gen-ing that’s done A to B is like those freaking fruit videos where the banana is abusing the strawberry, so she cheats with the plum. [laughs] But if you want to bring art out of it and have it not just make you uncomfortable, it takes hard work. Depending on what the song calls for, there’s a few approaches we take. Sometimes we use it more for layers within a song, like a vocal performance. Sometimes we do the whole song organically, and then run it through Suno.
For “Semiramis,” it was back, forth, and then back again: My friend Patrick Holder did the beat. I recorded the vocals. Then we fed Suno the elements, settled on something, brought it out, edited it on Ableton, and then put it back into Suno. So when you look in Deezer, it marks the song as AI-generated. We’re not hiding it. We don’t feel any shame about it.
Was the original beat for “Semiramis” generated by AI?
No. But what you’re hearing in the end, all the complicated edits, that’s not all us. You can hear the maximalism in the track, which is very distinct from my typical minimalism. At the same time, there is an essential quality that carries through, because it’s the same artist.
For “Semiramis,” if the vocals are based on your vocals and the beat was made on its own, why did you even use gen AI at all?
We were curious what it would sound like. Giving an artist who cares about what they’re making a restrictive tool that requires perfect dictation and a little bit of technological know-how isn’t giving them a leg up. Like, I can sing. [laughs] That matters to me very much. The only reason I’m using gen at all is because my visions are currently far beyond my means. People have been asking me, “Did you use gen AI for this EP?” I’m like, [sarcastically] “No, I hired a professional children’s choir.” What are we talking about? I haven’t paid rent last month and I won’t pay rent this month. I don’t know where the fuck I’m gonna be.
If we can’t figure out better stuff to do and we’re burned out, we’ll just start throwing shit into gen and see if it’s funny. Then, when we heard “Semiramis” back, we were like, Oh god. Because of all of these stigmas against AI, the initial feeling was to do a non-AI version—maybe I could close a deal, and then get a budget and hire a band, and then recreate it from scratch, like that weird virtue signaling everyone’s doing. But that just felt like a waste of time. Obviously, in a perfect world, we can get some funding and be able to take this type of stuff on tour and play it for people.
People’s perspective about this is fundamentally anti-artistic, and it’s wrong. It’s not our responsibility to keep greedy companies from being greedy, and we’re not going to be able to. People protesting the use of generative AI isn’t going to end it. It’s here. We’re under control. So what can you do? I’m gonna make sincere work that still speaks for the ethos I’m trying to embody, and I’m gonna be agnostic about tools, like I always have. It doesn’t mean that I’m Team Peter Thiel or something. I just think the stigmas about this stuff don’t apply to me or anyone who’s using the tool in a way that I view to be ethically correct.
How would you draw that line between ethical and unethical AI usage?
My intentions are pure. My connection with using the tech arises purely out of my love for hearing the possibilities, and I think sharing and making music for those reasons could never really be unethical. I’m not naive enough to think that that erases the terrestrial effects that potentially parasitic business practices around AI may be causing, but I don’t know if an artist is necessarily complicit in whatever mishandling by the developers of these tools.
The best we can do in this messy world is to try to do things that resonate on that pre-physical plane where music lives. That’s where change happens in society first: It’s in the ethos, in and from the heart and mind, and the point of an artist is to stimulate those muscles and take whatever punishment comes with it if need be. It’s up to the artist to draw that line for themselves. I’m not judge nor jury. I’m doing my best to do right, the best way that occurs to me, given my specific gifts and interests.
So if you had a budget to work with, you would hire people to recreate these songs organically?
Ideally I’d like to have the option. But there’s a lot of logistics there, so I will probably bite off what I can chew, and expand little by little. I’m an idea guy, I can sing, and I have really immediate, natural musical talent, but that’s not the god I pray to. The goal is to be able to have the absolute fluidity to hire anyone I want. Anytime you need to hire a bassist, a keyboardist, and a drummer, you’re paying more than I’ve made in the last year. If you bring in a full choir and an arranger, it’s gonna be a pretty penny, but whatever. Once I can justify its usage for this or that piece of work, that’s the goal. And I probably will still continue to use whatever technology comes out agnostically, unless I feel that the use of some tool is actually antithetical to art. It’s just a crazy context these days. I feel like people who aren’t involved in music don’t really understand how much the well has dried up.
What kind of prompts are you using in Suno?
I cannot outlay the exact prompts, because I’m convinced no one is thinking of how to use this stuff right, and I don’t want to give people the cheat codes. But I will say that the point with how I prompt in general is absolute fucking chaos. There’s very little censorship in the process. Aside from adjusting the writing pretty consistently, how I prompt styles is I just keep stacking. And I don’t proofread, which helps a lot.
I played around with Suno for an essay I wrote earlier this year, just trying to get an idea of its capabilities, and it always arrived at this really generic sound. It seems difficult to use it to make something stranger.
It’s gonna do what it wants, so you have to trick it. It censors a lot of profanity and racial epithets and sexual language, all of which is freely in my music. And it makes super-generic voices, so I tend to blend the gen voice with my voice, so it has this uncanny quality to it. It can sometimes take a few hours of gen-ing to start to get it to really fucking rock out and do some weird shit. So I just keep pushing at it, like I have with everything, until my will manifests. [laughs]
How long did “Semiramis” take to make?
It took about a day-and-a-half, all day, all night, all morning.
How many times did you put elements of the song or the full song through Suno to arrive at the final version?
Probably a few hundred. We generally do up to a thousand. It depends on whether or not we’re trying to extract a full gen-style song with that built-in irony or we’re trying to use layers. If we’re using layers, we typically have to gen a lot more. You can feed it just the layers you want, but it’ll still try to turn it into a full song. So you end up having to root through and dig the stems out. When we do that, we sometimes have to do even more gens, because we’re using 70 layers on one song, chopping up words and shit.
Do you ever think, This would actually be much easier to do without AI?
No, because we do that all the time. That’s not a mystery to us. We’ve worked with gen, but that doesn’t comprise a lot of how we work, really. Before we’ve had coffee, we’ve physically made music. In any given five-minute conversation between us, we probably will have sang 10 times. We’re so obsessively musical, and using AI became a place to release some of the tension. Because sometimes you’ll go into the studio and have a problem with your keyboard, and you’ll be there seven hours later and have put nothing together. There’s a lot of anxiety and trauma that wells up from that, and it’s why most people start to become jaded with music, and then they quit.
We just felt that it would be fun and interesting at first, that it could be a part of a conceptual art project, and that’s kind of where we’re leaving it. Anyone who thinks our main goal is to make this our project and trick everyone, that’s some luddite perspective that’s been fed to people. We’re still broke, and we haven’t made anything off of it, and that’s that. We’re making what we want to make and trying to survive these changing times that have no manual.
Are you using generative AI for lyrics?
We write the lyrics. I write most of them, but when we have rap-type songs, I usually work with my friends, because we play off of each other really well.
What does “Semiramis’ Dream” have to do with Semiramis, the ancient Mesopotamian queen whose story has been reinterpreted for centuries?
The reference comes from an old anti-Catholic book I found called The Two Babylons; I’m not anti-Catholic, I’m not religious at all, and I have no prejudices, just to be clear. But I like to swim in people’s ideas and hear people out. This book confabulates a few social figures, and it paints Semiramis as the first Jezebel queen, biblically speaking. In the book, which I’m told has a lot of cultural and historical inaccuracies, Semiramis installs all of these different precepts in Babylonian society, like usury and subjugation and idol worship. So yeah, we’re living the dream.
Could you put a percentage on how much of “Semiramis’ Dream” is AI-generated?
I think that’s the wrong way to think about it, but also, I don’t know, dude. We wrote the core rhythm and arrangements, the core melodies, and the core words. But it wasn’t even initially a jungle type of song, so a lot of that we prompted in, pulled from Splice, mixed up. The way we get these diverse results is by actually making music. [laughs] It’s just a new technology. If you get an Akai drum machine and think it’s going to give you the best rap beats of all time, it’s not. You have to know how to do it.
Are you planning to use generative AI on the 1010 album you’re working on now?
That’s not how we’re making the record, but it will likely have kids’ choirs where we record the layers and then turn them into kids. I’m heavy into children’s choirs right now. The choir is one of the first sounds I ever heard. I grew up singing, or lazily lip-syncing, in choirs that my father and other pastors led and arranged, and I don’t know if there’s a purer sound in music than the singing of children. When I can afford to record a children’s choir ethically and without too much constraint on the kids, it will be really amazing music. I’d love to get my sons involved. They are beautiful singers.
