Years ago, I wrote a piece called “Lurch” (actually, I got as far as implementing a study for the piece). In it, a DSP chip runs self-modifying code starting from whatever’s in the contents of it’s memory in a sandbox, provided with an external audio input (hooked to an AM radio) and an internal one (hooked to a simple sinewave oscillator). It evolves according to a “goodness” measure provided by a MIDI continuous controller value. A performer sits and drives a mod wheel up and down in response to how well (s)he likes what the hardware is spitting out.
In performance, it usually starts with no audio out, then manages to spit out a few clicks, and, encouraged by the performer, maybe manages to output a munged version of the input signals over time. There’s a recording of one performance here which amazingly spat out the internal oscillator signal practically from the start.
Anyway: I ran into this on CDM today, and it reminded me of this. A performer/builder constructing sound from scratch; the tension of the performance has to do with whether or not the performer will be able to coax out the sound on the fly by constructing the hardware. She’s going for a known result, unlike Lurch which always comes up with something different (when it works at all – early versions blew up the DSP chip in a literal quick puff of smoke by generating illegal instructions).
I still remember one of the responses to the recorded performance: someone wasn’t allowed to play it on the family’s new CD player, because the Dad thought it’d blow up the speakers. Rock’n’roll!