The Brain Is a Playback Device — A Composer's View on the Common Ground Between Humans and AI
What the Med Student Said
What first sparked your interest in the relationship between the brain and creative work?
Back in college, I was in a band for a while. One of the members was a med student. One day he just dropped this line: "Clinically speaking, all humans do is ruminate on their memories."
That's a pretty remarkable thing to hear from a bandmate.
It hit me hard. I was just a young kid in a band back then—I didn't understand anything about music yet. But that one line stuck with me. After that, I went pro and spent 15 years composing and arranging, and I've never once had the experience of a melody just "coming to me from above." I'm always just rearranging things I've heard, things I've lived through, things I've accumulated. What that guy said as a student—that's exactly what he meant.
A phrase you heard when you were young gradually became something you could feel firsthand as your career progressed.
Exactly. I've never once felt like I was creating something from zero. I'm picking one thing out of a hundred inputs and rearranging it. "Rumination of memory"—that phrase came closest to describing what it actually feels like.
The Brain Was Already on My Mind
Did that med student's words hit so hard because you were already interested in the brain?
Reading Paul Broks' "Into the Silent Land" was a big part of it. The book is full of case studies—patients with brain damage who lost their memories or had their entire personalities change. Reading that, I realized that "being yourself" is ultimately just a state of your brain.
So the question about the brain came before the question about music.
Right. Making music, I always had this nagging question: "Why do I create the way I do?" But after reading that book, I realized it wasn't a question about music—it was a question about the brain. Broks' book and the med student's words. Those two things converged, and "all humans do is ruminate on their memories" just clicked.
It wasn't about creative theory—you were drawn to the brain's mechanism first, and that circled back to music.
That's why I ended up conceiving MUEDlobe—a tool for understanding your own brain. It's positioned in the context of music production, but the root motivation is really "I want to understand my own brain." Music is just the context because that's my field.
Dream Visuals Are Low-Res
This "rumination of memory" idea—it also came up when we talked about dreams, right?
After a long road trip over New Year's, I had this incredibly vivid nightmare. My guitar was decaying and falling apart. But the way it broke wasn't physically possible—the mechanism was all wrong.
That's similar to early AI image generation, isn't it? Six fingers, arms that aren't connected to anything.
Exactly. The visuals in dreams are a lot like those low-fidelity images early AI used to generate. Something that looks like a guitar gets output, but the details don't hold together.
Mechanistically, the same thing is happening. AI image generation is just reconstructing from patterns—there's no layer that checks for physical laws or structural consistency. The dreaming brain works the same way: the prefrontal cortex is offline, so it outputs without verification.
Both of them are just bluffing their way through on "looks about right."
Two Databases
There was a fascinating phenomenon in that dream, right? The memory of "I've jumped from here before."
Yeah. In the dream, I had jumped from the fifth floor of a building and landed safely before. In the dream, that is. So I thought, "I can do it again this time." Obviously in real life I'd never think that.
So inside the dream, there's a separate "learning history" that's different from reality.
When I'm awake, I know the laws of physics—"jump from the fifth floor and you die." But the version of me inside the dream is acting based on past successes within dreams.
Real-world memories and dream memories function as separate databases. When you're awake, you can't access the dream database, but inside a dream, you can. It switches depending on your state.
And in the dream, you can't even realize something's off. The prefrontal cortex—the part that does logic checks—is asleep.
In AI terms, it's like the validation layer is off and only inference is running.
Exactly.
What "It Just Came to Me" Really Means
This connects to the creative process, doesn't it?
Yeah. You know how people say "the idea just came to me" or "the song just descended"? I think that's the brain trying to consciously use the same thing it does in dreams.
Intentionally loosening the prefrontal cortex's filter and waiting for fragments of memory to connect on their own.
Ideas "come to you" during walks, baths, right after waking up, or during monotonous tasks—because those are the moments when your logic checks weaken. You're using REM sleep-like brain activity while you're still awake.
If you can control it, it's creativity. If you can't, it's a nightmare. Different outputs from the same engine.
But the mental cost is insane, right? You're deliberately lowering your brain's censorship while simultaneously making judgments to shape what comes out into something usable. Both at the same time.
It's like flooring the accelerator while trying to steer at the same time.
I approach it pretty systematically, so I manage. But people who just sit around waiting for inspiration to strike—of course their mental health takes a hit.
If it doesn't come, it feeds straight into self-doubt—"I have no talent," "I've dried up." Because they've mystified the process, they can't diagnose why it isn't working or how to fix it.
If you've systematized it, you can break it down: "I'm short on raw material today," or "I need to increase my inputs." It's not that the muse isn't in the mood—something in the pipeline is just clogged.
Same Machine, Different Specs
Dreams and AI images looking alike, memory reconstruction, the presence or absence of a verification layer—putting this all together, the human brain and AI have remarkably similar architectures.
I think they're the same. The specs differ, but the underlying machine is the same.
Sticking to the facts: at an abstract computational level, there are many commonalities. "Reconstructing from learned patterns," "generating output based on statistical regularities," "output coherence varying depending on whether a verification layer is active." Deep learning itself was inspired by neuroscience, and recent studies have shown similarities in activation patterns between language models and the human brain during language processing.
At the implementation level they're probably different, though.
Yes. Biological neurons and transformer architectures are structurally different. The human brain has embodiment, continuous learning, and evolutionary constraints. The honest conclusion is that "they might be the same" is the most intellectually honest answer we have right now.
That's exactly why I got into AI, you know? For 30 years I was making music without being able to explain how I did it. I couldn't explain it to anyone around me, couldn't even put it into words for myself. But when I looked at how AI works, I thought, "Oh—this might be close to what's happening inside me."
Your own brain was a black box, but by looking at AI—a similar mechanism—you could observe it from the outside.
After understanding how AI works, I became much more conscious about "recalling memories" in my own composing process. I stopped focusing on theory as much. It became clear that the source of creativity is the reconstruction of memory.
Default Mode: The Composer Within
Have you ever had the experience of your best melody coming to you when you're "not thinking about anything"?
All the time. In the shower, on a walk, right before falling asleep. It always happens when I'm not facing the problem directly.
In neuroscience, that's called activation of the default mode network. When you're not focused on a task, the brain spontaneously combines past memories to generate new patterns. It's as if a separate composer is working beneath your consciousness.
It connects to what we said earlier about inspiration "coming to you," right? Trying to do it consciously is mentally expensive. But there are times when the brain just does it on its own.
The problem is that this automatic output isn't always usable.
Right. That's why just waiting for inspiration is inefficient. Whether you can systematically capture what the brain generates on its own and actually use it—that comes down to the quality and volume of your inputs, and having a system to organize them.
You Can't Optimize a Human
This understanding of "rumination of memory" has also affected how you think about copyright, hasn't it?
If nobody can create something from nothing, then what exactly is copyright protecting? That's the question.
That can become a dangerous argument.
Sure. But my view is simple: copyright isn't compensation for a flash of inspiration—it's compensation for labor. All the effort of accumulating vast amounts of memory, digesting it, reconstructing it, and getting it to the point of output. So that value deserves to be protected for a reasonable period.
Even if AI can do the reconstruction in one second, a human spent 30 years acquiring that input. That time can't be compressed.
Right. Reconstruction keeps getting easier for AI. But not for humans. You have to live, listen, experience, accumulate. That still takes enormous resources, and that's why compensation is necessary. Without it, people can't survive.
In the end, it's not philosophy—it's survival.
"What is creativity?" is way less urgent than "Can people make a living?" Call it rumination of memory or whatever you want—we still need to make sure people can eat from it.
AI as a Mirror
A composer whose music changed through studying AI—that's quite rare.
"AI is going to steal our jobs"—you hear that everywhere. But "I discovered what my own creative process actually was through AI"? Not so much.
Usually it's the opposite. When veteran creators learn about AI, they tend to pull away: "Human creativity is different."
I went the other direction—I moved closer. "Oh, so this is what I've been doing all along." Ingesting massive amounts of patterns, pulling out the right ones for the context, combining them. I'm not creating from zero—I'm searching through a vast accumulation and reconstructing.
That's literally how RAG works.
Right. My brain was RAG all along. Retrieve from external sources, transform, return. Not generating from scratch.
But even with the same RAG architecture, the output differs between a database built over 30 years and one built yesterday.
That's the point. What we used to call "genius" is probably just a difference in database quality, database size, and conversion accuracy.
Whether you call it "rumination of memory" or "talent" is a matter of the observer's framing.
Coming back to what the med student said: "All humans do is ruminate on their memories"—that's not despair. It's liberation. You don't have to create something from nothing. Inputs and reconstruction are enough. The moment I accepted that, things got a whole lot easier.
This article was reconstructed from multiple conversations with an AI (Claude).