AI and opera make strange bedfellows
Artificial intelligence and opera – strange bedfellows, no? One’s mascot is the tech bro, moving fast and breaking things; the other conjures the diva, the bastion of art-for-art’s-sake thinking. Although the operatic stage is perhaps not the obvious choice for exploring AI and the creative industries, it’s the subject of the Royal Opera House’s inaugural RBO/Shift festival, which is happening this week. A programme of events and performances, Shift features experimental artists and thinkers at the vanguard of AI and music, like the celebrated American composer George E Lewis and the vocal artist Harry Yeff (AKA Reeps100). Amid all the digital ink spilled pondering the impact of AI on film, books and music, what is there to say about opera?
Opera companies and composers have not missed the AI memo, and AI operas have recently been popping up around the globe. Perhaps the most striking was Semperoper Dresden’s Chasing Waterfalls in 2022, in which AI appeared as an 8m-tall singing kinetic light sculpture. Trained on the Norwegian soprano Eir Inderhaug’s voice, it generated a new aria for each performance. Elsewhere, AI has been used within the compositional process. For Lexia: An AI Opera, which is set in a future ravaged by climate catastrophe, the composer Scott Deal wrote the music with a “bespoke improvisational system” whose data set included Deal improvising on the vibraphone. AI has also pepped up the classics. Arizona Opera, for instance, synced their 2025 production of Verdi’s Aida to a “visually stunning” AI-generated video feature.
AI and opera is not as incongruous a pairing as it might initially appear. Opera has always been incredibly technological. Since at least the early-19th century, at opera’s heart has been a tension between the ineffable aesthetic governing Western art and the material reality of everything that helps to produce the spectacle – bodies, sets, clarinets, lights and wind machines. Take Wagner with his Gesamtkunstwerk (“total work of art”). At his opera house in Bayreuth, he aimed to create a sublime illusion in which music, text and staging were united as a sensational whole. But look more closely, and there was a veritable panoply of technologies – my favourite being steam machines – animating the dream. No matter how much opera might aspire to exceed the material world, it is grounded in technology.
Still, Wagner’s steam machines didn’t have their own ideas about set design. Like many who work in the creative industry, I’m uneasy – and not only because I might soon be out of a job. Humans have always been anxious about where technologies end and humans begin (Plato was concerned about the proliferation of writing, for instance). But there’s something in the transaction between the consciousness of artist and audience – and in the process of creative activity – that seems to be compromised by AI taking on generative labour. Worse than taking my job, AI might take my reason for being.
When I spoke to Jennifer Walshe, a professor of composition at the University of Oxford whose music will appear at Shift, my anxieties felt overblown. The author of 13 Ways of Looking at AI, Art, and Music, she’s an early adopter of AI who is clear sighted about its possibilities as well as its limits and contradictions.
“We’re getting lost in this sci-fi idea of AI being ‘AI’,” she told me. In 13 Ways, she describes AI as “an incredibly powerful meme”. But doesn’t it matter that AI is generative? Not especially: Walshe cites a dice game created by Mozart’s publisher after his death allowing people to create their own unique “Mozart” piece, combining fragments by chance. In lots of ways, she says, AI carries on the aleatory tradition of the avant-gard composer John Cage. So the issue isn’t that AI is generative per se, but the scale on which it functions.
Experimental opera is only a part of an industry that is comercially reliant on auditorium-filling works: Carmen, La bohème, La traviata. And there’s money to be made, as Walshe noted, from opera houses’ archives: think Abba Voyage, but with a live orchestra and AI-generated resurrections of stars – Maria Callas, Luciano Pavarotti, Enrico Caruso. In Walshe’s music, AI is a tool for finding possibilities and undreamed of noises, like a trombone that sounds like it’s turning into a drum. Or AI can provide raw material that musicians can use to jam together, riffing on the AI’s ideas to create something different. The lone creative genius model of authorship is a recent historical invention, and not without its problems. To collaborate with AI is to subvert that paradigm and acknowledge that creativity is more distributed.
Which is all very well, but it doesn’t address the pressing issues of how creators get paid today. And AI isn’t a normal collaborator. There’s a real risk that outsourcing thinking to AI – recently termed “cognitive surrender” – could impact our minds.
“AI is here to stay,” Oliver Mears, the RBO’s opera director, said recently: either we “put our heads in the sand”, or we “ride the wave of progress”. But so much discussion of AI and the arts comes from a position of disempowerment: as the composer Tarik O’Regan has observed in a Guardian article, there are the people controlling the tides (Big Tech), and the rest of us, who have only two choices: get on the Californian surfboard or drown.
“What we should want from AI is to allow us to offload some of the bad jobs, and allow us to augment some of the wonderful and creatively interesting things we want to do in any given process,” Kate Vredenburgh, who researches AI, the future of work and worker autonomy at LSE, told me. “So the question we might want to ask is: in which ways could AI support what we think is meaningful about opera?”
But good luck getting Big Tech to ask these kinds of hard questions about art. “Unlike industries where product development typically begins with an identified problem, tech sometimes inverts this process, starting with a capability and working backwards to justify it – ‘build it, and they will come’,” said Lauren Wong, a former Meta employee and Vredenburgh’s collaborator. What this means is that the specialist requirements of any given field might not be taken into account. Little money has been spent working out what opera or the arts need, or what humans need from art. People often say AI is just a tool like any other. But, if that’s the case, is it fit for purpose?
It’s a truism that technology in itself is neither good nor bad; the issue is humans. This is not reassuring. I’m pessimistic about the ability of legislators to keep moral pace with the speed and scale of AI’s development, particularly regarding the ways in which it might affect cognitive processes that require friction, like learning, or creativity itself. Given the economic incentives, I’m pessimistic about the likelihood that AI products will be attuned to the ever-hard-to-articulate value of art.
Mears is right to suggest we mustn’t put our heads in the sand, but nor should we resign ourselves. Those of us who care about what it means to be creative need to become AI-literate, and boundary-pushing initiatives like Shift might help. But I’m not holding my breath.
Emily MacGregor is the author of While the Music Lasts (Duckworth). The RBO/Shift festival runs until 7 June at the Royal Opera House, London WC2
[Further reading: Turandot, the last canonical opera]
