AI and Creativity: Can Machines Make Art and Music?

Ask ten people whether AI can create real art and you’ll get ten different answers. Some call it theft. Some call it the biggest creative tool since the camera. UK musicians, illustrators and photographers are all wrestling with the same question right now: if a machine can paint a masterpiece in ten seconds, what happens to the humans who spend years learning the craft? When I looked into this properly, the picture turned out messier — and more interesting — than the headlines suggest.

What Do We Mean by “AI Creativity”?

“AI creativity” usually means one of two things. Either a machine generates original-looking images, music or text from a text prompt. Or a human uses AI tools to speed up part of a creative process they still control. The distinction matters more than most articles admit.

Text-to-image tools like Midjourney, DALL-E and Stable Diffusion learned by studying millions of existing images. They don’t “imagine” the way a person does. They predict which pixels statistically belong next to each other, based on patterns in training data. Music tools like Suno and Udio work the same way with sound.

That’s not nothing. It’s genuinely useful. But calling it “imagination” stretches the word past what’s actually happening under the bonnet. The engineering is impressive even if the philosophy is shaky.

UK art schools have started teaching AI literacy alongside traditional technique, precisely because the boundary keeps shifting under students’ feet. Some now run entire modules on prompt craft as a creative discipline in its own right.

How AI Generates Images and Art

Most image generators use a technique called diffusion. The model starts with random noise and gradually removes it, step by step, guided by your text prompt. Repeat that a few dozen times and a coherent image emerges.

The results can be stunning. Photorealistic portraits. Impossible landscapes. Album covers in seconds, not weeks. A single artist can now produce concept art for an entire game in an afternoon, work that used to take a small team a month.

But look closely and cracks show. Hands with six fingers. Text that isn’t quite text. Backgrounds that don’t obey physics. These “tells” are shrinking fast, though — 2024’s obvious AI image is 2026’s near-perfect one.

Training these models isn’t cheap either. A frontier image model can cost several million pounds to train, which is why only a handful of companies control the leading tools. That concentration worries plenty of independent artists watching from outside.

AI Music: From Jingles to Full Compositions

Suno can generate a full song — vocals, lyrics, instrumentation — from a one-line prompt in under a minute. Udio does something similar. Type “melancholy indie folk song about missing a train” and you’ll have something listenable almost instantly.

UK indie labels are split on this. Some see a cheap way to generate demo tracks or background music for adverts. Others see a flood of soulless filler drowning out human musicians on streaming platforms.

Spotify has already pulled thousands of suspected AI tracks. The platform won’t say exactly how many. That silence tells its own story.

Session musicians report a genuine drop in library-music work — the low-budget, functional soundtracks used in adverts and corporate videos, which AI handles adequately at a fraction of the cost. Some have pivoted to teaching or session work for live performance instead.

Can AI Actually Be Creative, or Just Remix?

Philosophers have argued about “creativity” for centuries, so AI hasn’t settled anything new — it’s just raised the stakes. A strong case exists that AI genuinely produces novel combinations nobody typed in directly.

A weaker but still common view: AI is an extremely sophisticated collage machine. It recombines fragments of everything it was trained on, without understanding meaning, intent or emotion the way a human artist does.

UK investors keep asking about this because it affects valuations of AI art startups. If courts eventually rule that AI output can’t be original enough to copyright, entire business models collapse overnight.

Somewhere between those two extremes sits the honest answer: AI produces statistically novel output, but calling that “creativity” in the human sense is a stretch most computer scientists won’t make. The debate probably won’t be settled by engineers anyway — it’ll be settled by courts and culture, slowly and unevenly, over the next decade.

The Copyright Question UK Artists Are Watching

The UK’s Intellectual Property Office ran a consultation on AI and copyright that closed with no clear resolution. Artists want an opt-out from having their work used as training data. AI companies want the opposite — broad permission by default.

Getty Images sued Stability AI in the UK courts over allegations that Stable Diffusion trained on millions of copyrighted photos without permission. That case is still working through the system in 2026, and the outcome could reshape the entire industry.

For now, if you use AI-generated art commercially in the UK, you’re operating in a genuine grey area. Get legal advice before you build a business around it.