My daughter is 6 and has christened punk music as “pit music”, which is fucking rad. We spend hours jumping around to loud sounds and she takes immeasurable delight in pouncing and flinging her adorable sneer/smile hybrid towards me, hurling her tiny fists into the air, intent on annihilation. She is the reason I have abs. She calls this “pit dancing”.

We were listening to Andrew Paley’s recent release Scattered Light while on the road and she came up with the most insightful description during the drive: she said it sounds like “butterfly music”. I don’t know what butterflies listen to but they should definitely check this out. Everyone should.

This is not punk. I don’t know what it is. It’s simply exquisite.

Andrew Paley is a confirmed genius. This is my attempt to keep up.

Colorado Punk Rock Army: My initial viewing of the lyric video for Give Up was a remarkable experience as I am a synesthete. It was very powerful to watch scenes I had only ever witnessed behind closed eyes come to life. Can you give me a little background on Pixie, your new platform that links music to AI-generated imagery? 

Andrew Paley: I never know quite how technical to get with this stuff, but the gist is that I built the first iteration of Pixie as a final project for one of the classes I took last year as part of my PhD program. In part, it was an excuse to experiment with forms of generative AI outside the scope of my normal research. I built it on top of BigGAN, a model trained on ImageNet (a large collection of images tagged by subject) at Google’s DeepMind. The primary goal of BigGAN is to create new images based on the classes the model had “seen” during training. As example, you can ask the model to generate an image of a jellyfish, and it leverages the host of jellyfish images it’s seen to generate a novel jellyfish image based on the features it knows are associated with jellyfish. That’s oversimplified, but conceptually that’s a version of what’s going on. 

A few apps, like Artbreeder, then began to experiment with a slightly different approach to generative image creation: what if you could get models like BigGAN to “hallucinate” by requesting merged classes — not just a jellyfish, but a jellyfish-daisy-doberman-volcano-sneaker? Artbreeder also introduced a really cool UX approach to human-machine collaboration (or at least that was the first place I saw it) — the system presents you with various generated images, and you select which ones you like to push it in new directions for additional image generation. 

So, with that backdrop, I borrowed ideas from apps like Artbreeder and other open-source projects and created an end-to-end platform for the creation of music visualizations based on humans directing which imagery they like — creating “scripts” based on collections of images tied to each part of a song. With that imagery script in place, the system runs audio analysis on a given song file, and generates animations by tweening around and across the classes (and parameters) associated with those bookmarked images. To be sure, the system isn’t tweening between pixels in the images to create that shapeshifting, but generating slight variations from the model for each frame of the animation. The result is what you see in the Give Up video.

While leading the design team at Narrative Science, the group crafted work that garnered 12 patents. Do you have any intention to file patents on Pixie? 

No, not at all. Pixie relies on a few open source libraries and I would never want to do anything but contribute back to that if anything. My work at Narrative Science was in a very different domain — I was working on the design and development of systems for the automatic generation of narratives from data (aka “machines that can write”). Generally, I tend to prefer the democratization of access to technology and the open-sourcing of systems across the board — I just don’t always have a say in such things when working at a company.

You were a student at Northwestern on a Knight Foundation scholarship while seeking a master’s degree in journalism. Have you any interest in releasing a publication? 

My relationship with journalism is primarily through the lens of technology’s role in the future of information. I love writing, and I think reporters that go out and develop sources and write stories will always be core (and I read their work daily). That said, part of the reason I got into the Narrative Science stuff was because I saw that sort of technology as a core component of the future of journalism. We’ve collectively spent the last few decades working to make the world machine-readable in a variety of ways, and a result of that is that there are countless stories — old and new and ongoing — locked up in machine representations. I think one future role for journalism is in exploring how to make that world human-readable and comprehensible again. So yeah, I don’t plan on releasing a publication (unless you count my future dissertation, I suppose) — but my work in AI at both NS and now associated with my PhD continues to have at least a few toes in what I would consider a facet of journalism’s future. In fact, one of the primary projects I’ve worked on over the past year is a platform aimed at democratizing access to information associated with the US federal courts by automating the role of a data scientist and allowing non-technical people to just ask questions about what’s going on.

I read that you used to frequent a spot in Vermont – 242 Main – this sounds like it was a space that helped form you as a teenager. Do you have a particularly favored memory from 242 Main? 

Oh there are so many it’s a little overwhelming. That place was something of a second home for me from middle school on. I got to see so many great bands that would stopover en-route between Boston and Montreal to play on a two-foot stage in a small room, which was a pretty world-opening set of experiences for a 13-year-old from the woods of Vermont. Bane and AFI and Texas is the Reason and The Weakerthans and Elliott and Cave In and countless others. It also fostered such an amazing local scene of bands in this tiny city in this tiny state — Drowningman and Common Ground and The Cancer Conspiracy and many others, right up to today with bands like Rough Francis — that I was so lucky to grow up into and be a part of. And, as a part of all that, it taught me about the connection between art and activism. So yeah, one standout memory? Really hard to say. But in the aggregate, it was home to what became my extended family and was the springboard into much of what I’ve done ever since.

242 Main started in the 1980’s, it was the longest-running all ages club until it closed in 2016. Bernie (Sanders), the then-mayor of Burlington, was instrumental in scaling the spot. It was a youth-focused art space which was run like a club but accessible to the locals. These clubs are found commonly in Germany; here’s a room with a PA and a foosball table where there is no pressure to sell 50 or even 20 tickets.

I’ve been a Bernie fan since I can remember.

We played some shows with Bernie.

It’s pretty surreal to get done playing and then hand the mic to Bernie.

Do you have a favorite venue to play? What about a venue at which you’d like to perform?

242 Main will probably always be my favorite — kind of hard to beat. Now that I’m in Chicago, I love the Empty Bottle, Beat Kitchen, and Liar’s Club — and the Vic Theatre was a blast when we played it with Tiger Army a while back. Beyond all that, I’ve been lucky to get to play some awesome venues over the years both with The Static Age and solo — CBGB, Irving Plaza in NYC, Emo’s in Austin, Black Cat in DC, Molotow in Hamburg, Control Club in Bucharest, and Antiknock in Tokyo all come to mind. I really loved the place we played in St Petersburg a few years ago too — MOD had an awesome rooftop venue. Oh, also Colorado spots! I’ve had a blast playing both Hi-Dive and the Ogden Theatre. How’s that for refusing to narrow it down at all?

As for venues I’d like to play but haven’t yet, certainly Lincoln Hall in Chicago would be up there — that place has an amazing soundsystem and I love its atmosphere.

When did you first play CBGB? Who else played that gig? 

We first played CBGB during our first out-of-Vermont shows of any kind in May of 2002. It was a short run of 2-3 shows with Drowningman (who were also from Vermont and let us new kids tag along for the first few dates of their lengthy US tour). We weren’t originally part of the line up at CBGB, but we managed to jump on the show last minute thanks to Simon from Drowningman and also Rich Hall (whose birthday celebration it was). The other bands besides Drowningman included Thursday, Most Precious Blood, Unearth, Nora and Every Time I Die.

We then played it a second time in 2004 or 2005 during CMJ as a part of Tarantulas Records showcase with The Explosion, Death From Above 1979, Panthers and some others.

The recently released Caroline split with Days N Daze has such a relatable refrain in “the world’s on fire and I feel fine”. I’m curious about the backing vocals- did Whitney Flynn sing on this track? What was the inspiration behind this track? 

The backing vocals were actually done by Liza Ohm, a mutual friend (and bandmate) of Kay Petersen, who mixed all the songs on Scattered Light. The connection with Days N Daze came later when they did an awesome cover of Caroline for a compilation associated with The Fest 19 (that got postponed from October 2020 to October 2021, for obvious reasons), and then we decided to do a split 7” of the two versions together. Those guys have been great and I’m looking forward to meeting them in person on the other side of this pandemic. The split was mixed by Kay Petersen and mastered by Jurik Maretzki in Hamburg, Germany.

As for the song’s inspiration, I guess it’s ultimately about distraction and despair in the face of an increasingly overwhelming world of constant information and the loneliness of sort of never being alone. As a part of that, it’s also about our relationship to the bleeding edge of human history and how our relationship to everything is increasingly mediated by screens through experiences designed to enthrall and entangle in increasingly sophisticated ways. It’s about the frustration of seeing so much and being able to do so little, which I think is a heightened feeling during the Trump years. My generation got to grow up with the internet and all its promise, and now these tools that were supposed to democratize information and distribute power are being used to poison the former and consolidate the latter. What do you do if you’re a 20-something or 30-something in the face of all that? There are plenty of apps that would love for you to joylessly doom scroll past their advertisements while waiting for the polar ice caps to melt or snake oil salesman who will say anything to get your attention just long enough to hawk their products. I guess Caroline is a song about that, with a dash of hope thrown in — there’s no grand plan, the world is whatever we make it, either for better or worse.

“Remember me fondly or never at all”, from Give Up, is such a compelling phrase; it sounds like a call to arms in some manner. I find it to be both a demand for action and a promise to surrender. Why did you choose to release a lyric video for this song? 

Well, the genesis for that video came out of another area of my life — I’ve been doing AI research for a PhD, and I had started to play around with generative models just to see what I could create. We had already released videos for other songs from the album — Caroline and One Match Fire — at that point, but wanted to release at least a couple more singles leading into the release of Scattered Light, so I took it as the opportunity to meld the two sides of my creative life — music and generative AI — and Give Up was the first song I experimented with. I thought that the song’s texture might be a good fit for the visual style, and I wanted to highlight the song as I’d actually been carrying pieces of it around for a while — the original version was demoed back in maybe 2014, and I came back to it during the recording for this album. 

And I guess in some sense that song sums up a key idea of Scattered Light as a whole: giving up isn’t always about losing ground, sometimes it’s about creating the space for whatever’s next. It hurts, but a controlled burn makes the space for new growth — the trick is to not let it flare into a wildfire.

Caroline Andrew Paley / Days N Daze Split

Give Up / Pixie Lyric Video

Scattered Lights / Bandcamp

The Static Age / Facebook

Krista Brewer / 2021