The Science of Creativity and 40 Year Old Obsession

© Steven Boykey Sidley

A very long time ago, at the dawn of technological time (or so it seems from this remove), I was a naïve and enthusiastic young man and the first Apple was yet to be assembled. I was at Wits University in Johannesburg then, trying to learn about computer science, wading through the mud of a room-sized IBM 370, using punched cards and a language called Fortran, while trying to play jazz saxophone on the side, and, with limited success, trying to meet girls.

And into my word arrived an eccentric professor from England, David Brown, who offered a course on Artificial Intelligence, a subject so abstruse that my scientifically-literate father burst out laughing when he heard about it. It turns out to have been the first such course on the continent.

I swallowed it whole. It seemed to me to be the most important subject in the world. The analysis and deconstruction of human creativity. What could possibly be more absorbing?

And so began a merging of interests. What was happening when I played a jazz solo on my sax (I was, at this point, improving markedly as a player). Why was I choosing the notes I did? Why did some phrases sound fine, and others boring? What determined a coherent improvised journey? Why were some soloists like Coltrane or Bird consistently brilliant, and what comprised their genius? What was dissonance, what was consonance, what was surprising and what was pallid? How did composition work, and in how many ways could harmony be draped over a melody?

And how was emotion extracted from all of this stuff?

I turned to AI to try to figure this out. The details of my inchoate thoughts and approaches to the problem are less interesting than the following — Dr. David Brown loved what I was up to, and encouraged the effort, and co-wrote the only academic paper I have ever had my name attached to, which was published years later. It was given the arcane, dense and self-regarding title — ‘The Expression of Aesthetic Principles as Syntactic Structures and Heuristic Preferences and Constraints in a Computer Program That Composes Jazz Improvisations’

Life moved on, I went to graduate school, continued to play sax, continued in tech, thought about AI and creativity a lot, but that was basically it.

Until now.

A couple of months ago I listened to the podcast called Mindscape, from one of my heroes, physicist Sean Carroll. His interviewees were David Rosen and Scott Miles from a company called Secret Chord Laboratories.

And what this startup company is doing is at the centre of my 40-year old passion — using AI and neuroscience to strip the veil of mystery from creativity, specifically music. A bunch of young, smart scientists thinking about the same things that so exercised me back then.

I thought fuggit, I’ve got to make contact. Have a chat. No goal in mind. I foraged Dave Rosen’s email address and basically wrote — l am the OG. Here’s the first paper ever written on AI and jazz. All the best with your company.

Dave contacted me immediately. A Zoom call followed with Dave and Scott. Exchanging ideas and approaches and stories for too brief a time. The founders were generous with their time. Their addition of neuroscience and machine learning to their toolsets expanded their palette well beyond my paltry 1970’s crayon box. If their company were around 40 years ago, that’s where I would be.

So Secret Chord Laboratories. I don’t pretend to have gone as deep as I would have liked during the short time we talked. But as feedstock for insight they have input hundreds of thousands of popular songs, carefully taxonomized across various criteria. They have applied the learnings of neuroscience and AI and ML to extract hints and guides and rules as to what works, what doesn’t, what is pleasant, what is surprising, what is likely to find an audience in a given demographic or genre.

And most important, using the cement of causality, why that is so.

Why are they doing this? Well, Secret Chord Laboratories is a commercial venture, so profit is clearly part of the mix. But in talking to the founders there is clearly more at play. These guys are explorers. Why do we create? For whom do we do it? Where are boundaries between genre and artist and song? How can science cast its curious eye on the process? And how can we make the world a slightly better place in doing so?

And, at least for me, are science and the arts really separate magisteria? Perhaps, as I suspect, creativity is merely an elegant expression of scientific absolutes. A debate which rages continuously.

The company has been engaged by some of the largest music companies in the world for assistance in product selection, filtering, prediction, quality, and audience appeal.

Heady stuff indeed, and from my short interaction with them, the right questions, the right tools, the right team and the right stuff.

May they slay.