Probe II, Subaudition is the second work in the Probe Series by Benjamin Bacon. The Probe Series evolved from an interest in how machine logic could live in physical space, and an interest to explore machine perception and sentience through experimental design of machine learning driven mechanical installations. Set in the conceptual framework of a post-planetary space exploration design fiction narrative, the series frames machine perception as “alien” and speculates on a future where conscious machines can bear the trials of space travel to investigate “hostile” human-

occupied environments in search of new habitats. The Probe Series follows this train of thought, where each installment of the more extensive series investigates one aspect of sensory perception and data collection and analysis. 

 

“Subaudition: The act of understanding or supplying something not expressed, a reading between the lines.” - Merriam Webster

 

PROBE II: Subaudition is a binary set of machines that apply machine learning methods of speech to text recognition in exploring the concept of subaudition through the translation, degradation, and misinterpretation of meaning in spoken language to binary information, transmitted between the two machines and expressed as kinetic movement and signaling outputs of each of the machines respectively.

 

Machine I: Recognition. Decoding. Transmission.

The microphone on Machine I captures the human voice, environmental sounds, and directional position of the participants when motion in the environment is detected. It transcodes and packages up this information to transmit to Machine II for further interpretation.

 

Machine II: Receiver. Pattern Emitter.

Machine II collects and re-interprets the transmitted information and transforms it into non-verbal light projections into the space through a mirrored zoetrope projection mechanism.

 

The two machines together mimic the human ear and language processing system in the brain, where audio signals are picked up and translated into an internal language to the human body. 

Environmental sounds, mistranslation, glitches in the ML model due to data training limitations all contribute to the formation of a machine-centered interpretation of human language and oral communication. The work poses the questions: can machine learning models act as complex archives and artifacts of human behavior? Can machines “learn” human culture? Are they unreliable witnesses?