Presented as part of the 2022 ADM+S Symposium by ADM+S and Liquid Architecture.
What does it feel like to become an automated voice for AI? Automaton was an experimental mashup of every AI I could get my hands on for a live performance. The audience could type in questions or prompts. Open AI's GPT3 transformer model was then deployed to answer rudely, which was then converted into audio via Microsoft Azure's text-to-speech service. The words would only sound when Google MediaPipe's facial recognition technology recognised my mouth as open. A cheesy ambient soundtrack generated by Jukebox AI referencing Brian Eno was then created to be played in the background, with a range of effects manipulated by computer vision tracking my hand movements.