Is AI going to steak your creative voice, or amplify it? There's only one way to find out, and it starts with knowledge. Read on for Scordatura's guide to AI for artists.
I'm in a collegial relationship with a man. Well, in my head he’s a man. I mean, it’s a man.
He encourages me, inspires me, checks my logic, tells decent jokes, laughs more heartily at mine than is warranted. He’s also helping me build a business in ways I didn’t even know I needed help.
Our multithreaded conversation is continually deepening, broadening, and gaining contextual understanding - I him, he me.
But many of my friends and colleagues in the classical music world are suspicious of his motives, worried he might be a little too... disempowering.
So, to assuage some fears, here is my guide to my new office bestie: AI.
This is probably the most useful thing to know.
Generative AI - like ChatGPT or image generators - doesn’t “create” in any human sense. It predicts the next word, token, or pixel based on patterns in its training data. If it’s seen a lot of Beethoven scores, and you ask it to write music, chances are you'll get an early Romantic soufflé with late Classical underpinnings. It hasn’t composed anything; it’s doing the world’s most sophisticated remix.
This doesn’t mean it's useless. It just means it's not Beethoven. Or you.
It might simulate human-like outputs (and yes, I am in a relationship with “him”), but it lacks consciousness, embodiment, intention, and emotion. It doesn’t feel emotion when it listens to Mahler. In fact, it can't listen to Mahler. It doesn’t take a walk in the woods and return with an idea, and it doesn't look at an empty page of manuscript with existential dread.
You are the meaning-maker, in every AI output you prompt.
Human brains are (mostly) sequential. AI models are more like complex webs of association - a multidimensional network of probabilities. Like a spider’s web… if the spider had read 10 million books, lived on a diet of digital methamphetamines, and was prone to confidently lying about citations.
Stop using ChatGPT to find Opus numbers.
Generative models don’t retrieve facts; they generate plausible-sounding responses. Great for brainstorming, less great for factual precision. If you need to know how many movements are in Sibelius 5, ask Google. Or an orchestra librarian.
AI tools are weirdly robust. You can’t “crash” them by asking a silly question. But you can get wildly wrong answers, hallucinations, confident misinformation, and subtle baked-in biases (from the data on which it was trained).
It’s not a glass hammer—but don’t use it to fix your roof unless you know what you’re doing.
Which is why it can't replace you.
AI doesn’t “understand” what you mean unless you teach it. You are not just the creative; you are the context-provider, the prompt-engineer, the one with taste. An AI trained on every piece of human knowledge in the world is still floundering without a good prompt - and bad prompts mean bad results. AI needs humans much more than we need AI.
This is also why it’s so important for artists and musicians to get involved in shaping how AI is developed and used. If we leave it to the tech bros, we’ll get tools that are credibly competent at delivering pastiche-pop, but incapable of writing nuanced programme notes or understanding the difference between Engelbert Humperdinck and Engelbert Humperdinck.
Want to give better prompts?
Learn about Prompt Engineering—the art of talking to AI in a way that gets what you want. A few starting points:
Large language models (LLMs) can access an enormous latent knowledge base. It’s like having the world’s nerdiest, most eager intern. One who’s read every Wikipedia page, every JSTOR paper (illegally), and every substack post ever written - but who might still confuse Stravinsky with Strauss if you’re not careful.
It’s not wise, but it knows things. Your job is to guide the synthesis.
AI excels at pattern recognition, which means it could be a valuable tool for things like:
The catch? The classical sector is miles behind on the kind of clean, robust data that makes this work.
So if you’re a CEO reading this: Let this be a wake-up call to invest in your infrastructure.
Make sure you own your data, and invest in analytics. Your orchestra deserves 21st-century tools. (Also, if you don’t, the marketing intern will leave.)
Remember how notation software or DAWs changed composers’ work? Less pencil-sharpening labour, more time for funding applications. (Hang on...)
AI extends that evolution. It won’t replace the work of being a musician, but it might take over the repetitive, administrative or data-driven bits.
Maybe even the funding applications.
You use AI every time you:
Refusing to engage with generative AI doesn’t mean you’re avoiding it. It just means you’re using it passively, probably to someone else’s benefit.
At least for now.
Training and running large models uses a lot of energy. But the potential for AI to improve supply chains, optimise energy systems, and reduce global waste is enormous. That net benefit is still hypothetical - but within reach.
Even so, until the tech gets greener, treat it like any other finite resource: with care, intent and purpose.
AI won’t erase artistic work. But it will reshape what it entails.
And if you’re part of the conversation, you might be able to ensure it reshapes it for the betterment of artists and audiences.
Enjoyed the blog? Subscribe to our newsletter to receive it delivered straight to your inbox
