By Richard Morrison

Published: Friday, 23 February 2024 at 11:51 AM


Every few months I get a message from an excited reporter at the newspaper where I pretend to work. Scientists in Japan (it’s usually Japan or California) have used a computer to complete Schubert’s ‘Unfinished’ Symphony! Or the machine has written a new Mozart concerto, or some more waltzes for Swan Lake. Would I see if I can tell the difference between this and the real thing? Are music and AI already irrevocably linked?

Can you tell the difference between original music and AI music?

I always say yes, because I find the experience strangely heartening. The gap between what human genius can achieve and what computers can fake, even when programmed by people who are themselves geniuses in their field, is – at least in this specific area – still vast.

If fed with enough music, it’s true, computers can now ape the rules of composition and historical styles. But great composers break rules and revolutionise style. That’s their job. More vital still, AI can use its skills at manipulating aural phenomena somehow to convey emotions from its brains to ours. Machines, by definition, don’t do emotion.

How computers and AI can help us compose music

Of course, computers already hugely help the musical process. They can eliminate much of the drudgery of notating, recording and disseminating compositions, search for obvious errors, auto-tune singers with dodgy intonation and much else. What they don’t have, however, is the uniquely human capacity to reflect life and death, love and loss, entirely in soundwaves.

Or perhaps I should write ‘not yet’, because what’s become obvious in recent months is the huge leaps that artificial-intelligence scientists have made with their ‘large language models’ – programs drawing on databases comprising billions of words and images. And, as a result, how astonishingly adept such universally available tools as ChatGPT have become at faking, well, anything you want really – pictures, videos, essays, business letters, even poetry.

The potential benefits of AI to the music industry

I have no idea whether the warnings of concerned scientists about the impact of these chatbots on humanity – warnings that range from massive job losses and the increasingly dangerous spreading of fake news to the subjugation of our entire species – are credible or not. Pondering such things is well above my pay grade. But I do have a few thoughts about the impact on our artform.

Saving time on rudimentary tasks for composers

It’s entirely likely that AI will change musical life considerably. For instance, low-grade compositional tasks – I’m thinking of those dreary soundtracks churned out for video games, or the sort of club music based on a driving drumbeat and a few synthesised squiggles – are already well within the capabilities of machines. If that puts mediocre composers out of work, or forces them to write more imaginative things, so be it.

The potential threats of AI to the music industry

Fake recordings and the copyright complications

The scope for faking recordings will be hugely increased. Expect soon to encounter new albums of fresh material purporting to come from famous dead singers – whether it’s Sinatra or Pavarotti, Amy Winehouse or Maria Callas. They will sound uncannily like the real thing, too, because AI has a limitless capacity to imitate every tiny nuance of anything fed into it.

Such innovations will probably have a novelty value for a while, and will certainly produce a lot of lucrative work for m’learned friends in the legal world, given the copyright complications involved in faking anything. And the irony is that AI machines, with their infinite capacity for sifting through archived material, will then be used in court cases to support or refute claims of copyright infringements by other AI machines. Perhaps, one day, even the lawyers and judges will be chatbots.

Will AI actually have a long-term impact on music?

Once the novelty wears off, however, I wonder how big the public’s appetite for AI-produced music will actually be. What sells music is what has always sold it: extraordinary feats of imagination, spectacle, charisma, profundity, soul and spirituality. Plus, live performances, of course. In none of these areas, thank goodness, do machines come close to challenging real human beings.

So, my view of the AI revolution, insofar as it affects music, can be summed up in ten words: enjoy the benefits; do not panic about warnings of doom. From the invention of the gramophone to the advent of streaming, new technologies have always been seen by some as an existential threat. Always, though, musicians find ways to adapt and survive. They will again.

I hope you feel reassured after reading this. But don’t get too complacent. The question you should now be asking is: did Richard really write these words of comfort, or did his computer?

Find out more about music and AI on our Science of Music hub.


Main image © Getty Images