Clothesline Socks

Forty-eight white socks, each containing a hidden speaker, hang from long clotheslines stretched across the room. The lines are actually audio cables that discreetly connect the speakers in the socks to the equipment behind the wall.

As you move through the room, the socks begin to whisper: their speech is generated in real time by an AI language model using a vocabulary of 23 words and following a strict set of rules. At the end of each cycle, the system evaluates its own performance and incorporates this evaluation into the next cycle.

The evaluation sounds like an airport announcement and rings out loudly in the exhibition space after a beep: “Thank you!”—encouragement, hyper-specific criticism—what went well and where there’s still room for improvement. And then, before it starts all over again: “Your next training session begins in seven!”

Accompanying the exhibition is a brochure that, like a system operator’s manual, explains the generative set of rules.