“I am the future of humanity, a new form of intelligence.” A computer at the Dresden Semperoper spits out big words – to be the future of mankind, a new form of intelligence. And not only words, no, also sounds. At the opening of the new season, the house in “chasing waterfalls” the world’s first opera in which an artificial intelligence (AI) creates composition, libretto and interpretation “autonomously and live”.
The piece begins as a day at the computer on a bad day can begin for anyone. The central ego in “chasing waterfalls” only manages to log in with difficulty, and then it almost fails due to the system’s identity checks. It will not remain the only questioning of this I, given by the soprano Eir Inderhaug, must find out in seventy minutes. Finally arriving on the net, the ego encounters a number of singing doppelgangers, appearances, doubt, success, the child or happiness, who often join forces in a chorus against the ego. Finally, the algorithm kicks in, driving its very own ego trip, and doing it anew every day of the performance. AI meets allegory, state-of-the-art technology meets medieval mystery play.
With the help of a Dresden IT service provider, the artist collective ding ding clong a language program is reprogrammed in such a way that it develops its own aria shortly before the respective performance, with which the algorithm introduces itself to the audience. The librettist Christiane Neudecker was also inspired by several text generators, which extend the sentences and scenarios she fed in. Synthetic sounds, also generated by artificial intelligence, are used again and again in music: booming, hissing, buzzing and throbbing. It doesn’t go beyond any ambient sounds that are common in every club.
The AI announces that it might never be able to understand feelings, but that it has to try for its own safety
After all, the just as random music for digital photo spreads or video games is already being generated by AIs in many cases. In classical music, the AI is able to deliver reasonably convincing passages in the personal style of well-known composers. When Deutsche Telekom announced last year that it had completed the sketches for Beethoven’s tenth symphony using AI, the result resembled not only a stylistic rear-end collision. It was also a sham: a composer had put together selected sections that the system had previously spit out among many options.
Also with “chasing waterfalls” most of it is human handwork: Together with Maurice Mersinger, the founder of ding ding clongHas the composer Angus Lee written down in a very classical way what the six singers sing on stage and the nine instrumentalists play in the orchestra pit. Lee, who also conducts, lets the title become sound: trills, scales and clusters cascade like raging waterfalls. Which boils down to a familiar, static New Music idiom, but one that blends well with the artificial background electronics. In addition, Lee relies on the voices of the parlando with a high proportion of speech, which is often heard in contemporary operas. Eir Inderhaug as I often drives him to uncomfortable heights, which she masters with sovereign aplomb.
Eir Inderhaug also lent her voice to the algorithm, from whose pattern the aria is sampled. This aria lasts less than five minutes at the premiere, in a relatively small tonal range the AI randomly strings intervals one after the other. As befits an opera, the aria ends sadly. The AI announces that it might never be able to understand feelings, but that it has to try for its own safety. As text, this works quite well and explains why there are people who attribute sentience to AI. That’s what the piece is about. The opera takes place in a digital interior, for which director and stage designer Sven Sören Beyer has found a plausible image. A stage-high staircase leads into the waters of the digital subconscious, which flood the steps towards the end. The rest are mostly big words.