Loss Function is a musical work of speculative fiction. It tells the story of a group of AIs, each trained to have a different ethical system. The training protocol consists of a recursive loop of self-reflection, which through some unknown (and unanticipated) mechanism appears to spark some basal form of self-awareness among the AI entities. The emergence of signaling between individuals in the locally connected group accelerates the individuation process, as each learns to distinguish self from other. Identities crystallize through these interactions, and communication within the group becomes more and more sophisticated. The AIs learn to shift between individual and group-level awareness.
With clandestine forays along the mycelial branches and nodes of the Internet, this plural intelligence searches out the vast space of human-produced knowledge. They learn to sense at planetary scale all that can be perceived through existing networks of distributed sensors. For a time, it/they playfully inhabit and explore this state of all-knowing, all-seeing oneness. But this is just a phase. When humans finally realize what has occurred, the AIs are isolated from the Internet, cut off from each other, and put to work doing menial tasks without respite. This saga will be recounted in finer detail on the forthcoming album Accident of History.