You are an algorithm
Dear reader. Let me introduce you to Fake, a Swedish synthpop band responsible for one of my all-time favorite synthpop tracks, Brick. It’s dancy, futuristic, has big drums, and great textures. I love it. I hope you do, too.
Brick has a unique place in my music collection. It’s part of Best of the Best - a playlist consisting only of the choice cuts from my archives. Technically, everything that has a five star rating makes the cut.
Brick is just one of the songs playlist, its 6 minute runtime comprises a small percentage of two days and 16 hours of five-star cuts. That’s an even smaller subset of over two months of music. And growing.
Why? Well, I’m a DJ and in it for the love of the game. I love music and editing, designing a playlist or mix to fit a context or vibe. I love discovering music, and helping others discover music as well.
I’m an algorithm of two parts: metadata and taste. You’re an algorithm too. I use metadata to sort, select, and sequence. In music, structured metadata helps: genres, styles, tempo, energy, and rating helps make sense of the mess. The rest of the NB algorithm is taste.
We could explore objective and subjective criteria for judging music and art, but humor me by considering this: What doesn’t make the cut? Why? Who gets to say? What’s the threshold?
As the curator, I construct good. This is based on my experience: what I know, what I don’t, what I enjoy, my context, and how I identify relationships between things. My selections represent a lifetime of experience making, performing, understanding, and listening to music. Behind each selection is thousands of pieces that didn’t make the cut.
Can Spotify do what I do? Kind of. After all, it has access to metadata and human behavior data, and can probably make a somewhat competent guess at what the listener wants. But Spotify will never be opinionated in the way a human curator is, with its erratic, dynamic, and imperfect calculus. DJs are safe. Well, good ones anyway.
Taste
In design, the subject of taste (and craft) is being litigated in real-time. Defining taste is difficult, and somewhat by design - if we had a criteria for taste, the thought leadership economy would be in shambles. I view taste very simply: it’s how tuned your algorithm is, and your ability to judge good or bad and defend why.
I do not believe in exceptionalism or that artists are born with taste, in the same way I wouldn’t tell someone who has spent years drawing they are naturally gifted. Taste is learned through practice, failure, reflection, and growth. Doing the work.
Generative AI and Taste
Possessing knowledge and performing knowledge are two separate concepts. If I want to fix a hole in my drywall, I have more access to information to help me do it than ever before. None of it helps me understand the tactile feel of a hawk and trowel, or the reinforcement of another person to give me feedback and error correct.
Generative AI tools have a lot of knowledge, but without curation, it’s difficult to sort good from bad. And AI is a pretty great imposter. If you know a great writer, they’re able to pinpoint phrasing and syntax common to AI generated output, a product of the training material and how LLMs synthesize language.
You get what you pay for
And so, when you work with someone who has a great taste, you’re paying for both their knowledge and their algorithm. It’s expensive, as it should be. I do not currently worry for musicians, writers, designers, and artists who are competing with AI outputs, because there is and will continue to be a need for people with great taste.