Numb.station is a multi-channel radio installation that interferes with signals from people who speak like machines for audiences of people as well as people who speak like people for audiences of machines. Cold War era techniques of counter-intelligibility meet cybernetic training models.

In this work, utterances of numbers drawn from collections of archived cold-war era numbers stations and machine listening datasets are broken into very short segments and reassembled. Speech-to-text methods “listen” to these scrambled, reassembled sounds. Sounds recognized as numbers by software are broadcast using a software-defined radio (SDR) transceiver over FM radio. This recognition, however, might more accurately be understood as mis-recognition as these sounds functionally become gibberish. An interference signal broadcasts into the hollow space between these two logics. Another station overhears, unreliably archiving sound and metadata, trying to make sense of this assemblage.