A rehearsal is a repetition that points somewhere else. It happens before the event, in place of it, or in case it ever comes. Most of what I make is a rehearsal in this sense: a system set up to perform the same action again and again, for an audience, a future or an outcome that may never arrive.
I call this cybernetic rehearsal. In cybernetics, a system rehearses in order to gain control: it acts, measures the result, corrects itself and acts again until its behaviour holds. My works borrow that loop and hold back its pay-off. French has a single word, répétition, for both repetition and rehearsal. The works below sit in the gap between the two, and each one asks a different question of it.
01
Without a performance
I began with two machines. In Sisyphus1, six small robot arms build brick arches and an industrial arm demolishes them. Then the small arms begin again. The loop has no opening night, so it is all rehearsal. The question it left me with was not whether the effort is futile, but what repetition produces when the outcome is fixed. I have written about the work twice, in an art paper for SIGGRAPH 2024 and in an article with Shan Wong for Leonardo2. The second proposes what we call dialectical robotics: sustained opposition between machines generates adaptation and difference, not only waste. It never reaches a performance, but it changes as it repeats.
02
As work
The robots in Sisyphus work for no one. Most repetition does not. Someone performs it so that a system can run, and machine learning is where that becomes industrial. To train an agent is to make it attempt the same task millions of times until a behaviour holds. In The Farm3, made with Shan Wong, agents learn to climb a sheer wall, fall and begin again across 108 screens, while a counter records every trial. The work sets this training beside the human labour it resembles and relies on: migrant construction work, and the outsourced data work that trains AI. In each case the rehearsal is extracted and the one who rehearses is kept out of view.
03
As maintenance
What does that work cost the one who does it? The stone steps of Hong Kong Island were cut from granite by Hakka stonemasons. Granite is acidic, so the ferns and moss on the steps cannot grow on the stone; they live in the joints, in the mortar the masons left behind. Connected Moss4 sets twelve screens into a steel staircase and runs them as a neural cellular automaton, each cell updating from its neighbours alone. Like the moss, it does not develop. Crushed or scraped away, it grows back to yesterday’s height. This is homeostasis, the state cybernetics was built to produce, and here it describes the climber’s knees: deplete, recover, reset, repeat.
04
After the event
A rehearsal is meant to lead to something. Across Hong Kong, where I live, I keep finding loops that have outlived what they were rehearsing for. Ghost Logistics5 places reinforcement-learning agents inside 3D reconstructions of the city’s vacant shops, moving goods through interiors where trade has stopped. In the room stand sixteen flatbed trolleys, and one motorised trolley loops against obstacles it cannot pass. The agents move without friction while the trolley stays trapped. Between them runs a line from the coolie to the software agent: the body that carries is replaced, but the carrying goes on.
05
For a future
If a loop can outlive its purpose, it can also be asked to rehearse a new one. Ding Lab6, made with Shan Wong as Foreseen Agency, ran on one of Hong Kong’s trams. We spent three months with the engineers, mechanics and drivers whose daily maintenance keeps the system running, and then built Tram0, a cyborg tram that tells its own life story and remembers the older trams before it. The city usually offers its infrastructure two endings, preservation or abolition. Tram0 rehearses a third, in which the tram evolves. This is what I mean by speculation: a rehearsal for a future that is not yet decided, staged with the people whose repeated work the system already depends on.
06
Method
I work by building systems that run. Machines, simulations and agents repeat in public, so that the repetition itself can be watched and questioned. When a work enters a real place, as with the trams, it begins with fieldwork among the people who keep that place running. I teach media art theory and practice at the Academy of Visual Arts, Hong Kong Baptist University, and co-founded SATA, Society Art Technology Asia, an artist-led platform that questions Western-centred accounts of technology.
What repeated effort holds a system up, and what would it take to see it?