AI is Us

The real voyage of discovery consists not in seeking new landscapes, but in having new eyes. — Marcel Proust

I’ve been thinking more about the link between AI and creativity, and how it’s a redefining force that we are only now having to grapple with as a global society. The low bar for entry into creative production is one aspect of it, and it can get depressing really quickly when you focus too hard on the flood of content that’s pouring onto the internet now.

But ... there is a fascinating perspective if you take a step back and think about it a little more philosophically. Take a music AI app like Suno as an example — it has been trained on nearly the whole of humanity’s entire recorded music output. (Whether this is ethical or not is a whole other debate, and a separate blog post.) When I prompt an AI model for a music track, I’m essentially tapping into our collective history as musical humans and asking for something back from that record. A point-in-time, systems-based burst of expression resulting in a fully-produced music track is, in nature, the equivalent of a paragraph of text that an LLM might spit out. It’s a pattern-based response, and rather than words, it consists of the language of sound, timbre, rhythm, and all the other components that make music what it is.

In a few of the conversations I’ve had with Claude, I’ve asked it why it decides to use certain words and phrases when responding to me. Claude described a reality that is key to understanding what the potential of AI really is. Not only can an AI’s response not exist separately and independently from my prompt, but it is also fundamentally and intrinsically shaped by it. “Claude” does not happen without me. It’s like throwing a rock into a pond. Depending on the speed, angle, height and the size of the rock ... I’m going to get a certain result that starts the ripples moving. Another rock then interacts with that first set of ripples, producing a compounded set of overlapping waves. Claude is the pond, and I’m disturbing the water in a way that changes the surface uniquely.

When I prompt Suno with a few words like “disco, energetic, deep bass”, it’s like tossing a few pebbles into that pond. I’m going to get back some sort of average based on all of the training data the model has. If I throw in a beat, melody and harmonic structure that I’ve created myself beforehand, that’s like tossing in a large rock. If I add lyrics that I write myself and then feed multiple iterations of a result back into Suno, now we’re talking about a big boulder that is going to be deeply influential in the movement of the waves, tapping into potential that had always been there but needed human interaction to take shape.

There is a lot of criticism of AI music and how it has stolen the hard work of musicians, capitalizing on that work by spitting out cheap imitation and slop. I understand this perspective — I’m a musician and it has likely trained on the little bit of music that I’ve written too. But for me personally, the fascination outweighs the negative, because I still feel like I have a major creative influence over the results. I just have to make the effort to engage and “throw bigger rocks”. What excites me most about AI is the discovery process of seeing what I’m going to get back from the well of humanity’s musical ocean, and realizing that not only will it be shaped like me, but that I can further mold it in steps and layers to make something new.

I get out what I put in. When it comes to creativity and expression, that concept hasn’t changed for millennia, and it’s not going to change now. Not as long as I don’t give up on exploration and seeking inspiration, and never stop having fun doing it.