You are here

Leader: Soft Machines

Soft Machines

Machine learning is a ticking time bomb underneath the arts. AI tools such as DALL‑E and Stable Diffusion create visual artwork that is sometimes uncanny and often frighteningly convincing. ChatGPT will generate short‑form text that is distinguishable from real human writing only by its perfect spelling. It can even code.

The challenges are greater with audio and video content, but they’re not fundamentally different. It’s only a matter of time before AI can produce music that ‘passes’. And it won’t be created using layers of overdubs, by mucking about with MIDI, or in any other way we can observe and make sense of. It will simply emerge fully formed in response to our polite inquiries.

Does this mean the end for music as a human career? No. Music differs from writing, painting and photography in that live performance is fundamental to it. An artificial intelligence is pretty much the definitive studio artist. The parts of our job that involve playing ‘Mr Brightside’ to a handful of uninterested punters in the Dog & Duck on a Friday night are safe — for now.

An AI trained entirely on symphonic classical music would never be able to come up with rock & roll, and vice versa.

More fundamentally, the creativity of machine learning is essentially synthetic. The scope of an AI is set by the body of work on which it’s trained. An AI trained entirely on symphonic classical music would never be able to come up with rock & roll, and vice versa. Machine learning is fantastic for creating pastiches and, as such, is a real threat to composers of library music, for example. But it’s debatable whether it could anticipate or drive the emergence of new musical genres, at least without extensive human input.

And this is something that the alarmist view of machine learning tends to overlook. What you get from an AI is highly dependent on what you put in. Crafting the prompts that will deliver what you’re looking for is a different skill set from writing and recording music the old‑fashioned way, but it’s a skill nonetheless. Is it optimistic to think that people who’ve spent their lives working in music might be able to interact with musical AI more effectively than others? And that this could open up a new income stream for some? Time will tell.

Before AI music itself becomes big, though, it’s a safe bet that you’ll see plenty of AI‑generated writing about music. And in a world where anyone can create a website stuffed full of artificially created ‘reviews’ or ‘tutorials’, the value of an established and trusted source like SOS will be even more apparent. Machine learning can do amazing things, but it can’t get a new synthesizer or microphone out of its box, plug it in and put it through its paces. And as long as we are still making music the human way, human experience really matters!

Sam Inglis Editor In Chief