THE_OPER& Was a Ritualistic Study of Machine Learning—But What Did the Machine Learn? | Theater | Indy Week

Arts » Theater

THE_OPER& Was a Ritualistic Study of Machine Learning—But What Did the Machine Learn?



Going into the premiere of John Supko and Bill Seaman's THE_OPER&, I didn't know what to expect, despite having spent two weeks studying it for a preview in last week's INDY. I knew about its intricate, layered wordplay, full of references to Raymond Roussel and Marcel Duchamp, Ada Lovelace and Peter Greenaway. I knew about its creation and destruction of worlds, its vivid naturalistic videos. I knew that difference, change, and randomness were central. And I knew to expect beautiful singing from the Lorelei Ensemble.

The work had all these things and more, happening all at once. But I wasn't prepared for its sheer, overwhelming density: the singers walking around the stage in geometric patterns (including one poor soul cursed to forever pace back and forth), the lights flickering as the voice of the computer contemplated its own existence, and the strangely soothing music burbling through the seams. Four worlds were indeed created and destroyed, though I found their creation to be far more interesting. The computer projected seemingly random strings of words for the singers, which gradually coalesced into mystical incantations. Then, the computer would project an image and describe its constituent parts ("21.73% boat," "8.32% spider web," "0.34% church"), gradually learning to understand—sort of—the visual world. By the fourth cycle, it was predicting the images, often with amusing results ("very very beacon," "not unlike bridge").

What surprised me was how little actually seemed to change over the four cycles. Many small details varied and evolved, but the larger architecture felt oddly static. In each cycle, the Lorelei sang the same five minutes of music to bring about the fall. Every time, the images fell apart in roughly the same way. Every time, the aftermath of the collapse was an inundation. By the fourth cycle, that sameness overwhelmed my ability to perceive differences.

Thus the opera was less an exploration of randomness than a ritualistic study of machine learning, the process by which a computer comes to understand an aspect of the human world. What that aspect is, I'm not totally sure. With so much tightly entwined, wonderfully realized surface, it's hard to reach the depths that are clearly there.

Add a comment