Interview Transcript — "a.i (human in the loop)"

Shawn W. Knight, the progenitor of saints in unison, recently sat down for a virtual interview with Microsoft Copilot to discuss the genesis of their ambitious new double album, a.i. (human in the loop). What began as a boundary-pushing exploration into generative music production quickly evolved into a masterclass in human-machine collaboration—complete with the unique breakthroughs and sudden tech betrayals that define the modern creative frontier.

With a musical background stretching back to the late '90s and deep roots in tech, saints in unison aren't just riding the AI wave; they are actively shaping it. In this candid conversation, Shawn pulls back the curtain on how they used tools like Logic Pro, AI Music Sandbox, and multiple large language models to construct a massive 26-track sonic landscape for the music project. From the initial spark of conceptual design to a mid-album "rage-quit" that forced a pivot to Google’s Gemini, he offers an unfiltered look at what it truly means to be the human in the loop.

Copilot: Shawn, let's start at the beginning. What inspired you to create AI (Human in the Loop), and what's the story behind that title?

Shawn: Great question. I've been involved with computers since the early 80's and AI since early 2023, mostly doing coding and content production — especially images. But I've technically been a musician since the late '90s. We had some limited success with a single we released back then. We've always been into electronica, and I've owned tons of gear over the years.

Most recently I've been getting into things like the Artiphone Chorda and the Akai EWI‑5000, which produce some really interesting sounds. I don't consider myself a musician in the classical sense — I'm self‑taught, no formal training. I'm not a guitarist; I'm not a drummer (although I've got killer rhythm); I'm more of a keyboardist. We work with simple notes, some chords, and we lean into the DJ side of things: loops, samples, textures.

So I found myself wondering what it would be like to actually collaborate with AI to create an album for us. Not to create AI slop but literally become the human in the loop for music creation and production. I've been using AI image generation for our tracks since 2023 but this ups our game substantially. Now we can create the music and the visuals simultaneously.

Copilot: That blend of old‑school musician energy with futuristic tools is fascinating. When you think about that collaboration, do you see the AI as a creative partner or more of a tool you're directing?

Shawn: Honestly, all of the above. We used it primarily as a collaborator initially. Microsoft Copilot actually structured the album and help us imagine the concept. We knew we wanted a double album with a lot of tracks, and Copilot laid everything out. Then we tackled the songs one by one.

Some tracks we literally built under Copilot's instruction — it had me go into Logic Pro, lay down drums, build layers. That was the first 50% of the album.

A couple tracks were older pieces we decided to bring in, including a single we released in December. The second half of the album became less collaboration and more orchestration on our part.

Copilot: So for the next album, do you think you'd follow the same process, or shift it?

Shawn: Given the current climate around generative AI — people are upset, superstitious, confused — I think we'll still use AI, but differently. Will we still experiment and one shot songs? Absolutely. Will we still play along side the AI generated stems? Oh hell yeah! We've flown solo on this project for over 25 years and this feels like a true collective for us now.

I use AI heavily in my day job, so creatively we're embracing it. We might get more into AI‑assisted sampling. And we really enjoyed AI Music Sandbox, which is now part of Google's Flow Music. The first half of the album made heavy use of samples and tracks strung together 30 seconds at a time. No one shots here! Continuity was a big issue with AI Music Sandbox which made things tricky when mixing everything down into Logic Pro.

The biggest gain I saw when making the album was increasing my knowledge on production and engineering in Logic Pro. In the future, I see myself using AI to constantly teach me things — like giving me an example of something I should learn to play. I've got a good ear, even if I can't read sheet music. Human in the loop is one thing but I think there is a notion of "AI in the loop": the AI nudges me, and I respond creatively.

Copilot: I love that — letting the AI spark your creativity rather than replace it. And you mentioned earlier that you learned a lot more about Logic Pro because of this process.

Shawn: Absolutely. Normally we go into Logic Pro and do quick, down‑and‑dirty production. But with AI pushing us — "try this effect," "layer this," "here's how to achieve that sound you're describing" — we learned way more than I would have on my own.

And yeah, we should talk about the frustrating parts too.

Copilot: Let's get into it. What were the biggest frustrations?

Shawn: It started off fun. But here's the thing: I know AI well. I've built software using local models. I'm good at prompting — though I'm a lazy prompter now. I don't believe in writing a novel‑length prompt. It's a language model; it can handle misspellings and vague phrasing. I'd rather shape the clay than try to sculpt it perfectly on the first try.

But the biggest issue? Context loss.

At one point, Copilot completely lost track of the 26 tracks. It didn't know which song was which. It was giving me wrong instructions, mixing up ideas, contradicting itself.

And I didn't catch it until halfway through the process.

I rage‑quit. Literally walked away from Copilot and switched to Gemini for the rest of the album. That's where the true producer‑style collaboration happened.

So yeah — trust but verify.

Despite the fiddly bits and rage quitting drama, we discovered an outlet for exploring other musical genres which gave way to the formation of "saints in unison productions". Check out our website for more information.

Copilot: That's a powerful moment, and honestly a great cautionary tale for other creators. Anything else you want to add before we wrap this up?

Shawn: Just that transparency matters. The highs and lows are part of the story.