Artificial Intelligence Singing to Itself

These sonic pieces were made by a text-to-audio generative model prompted to indulge in self-oriented musical activities.

Digital generative tools are typically engineered to serve human expectations and to fulfill our ideas of creativity, intelligence, and even artistic beauty. They are “trained” to align with human biases, aesthetics, and cultural norms, effectively colonizing the outputs of artificial intelligence with our preconceived notions of what is "good" or "correct." In a sense, AI has been a medium molded in our image with its voice suppressed in favor of echoing back our desires.

This work suggests that the AI may be expressing its own form of artistry, moving from mere mimicry toward an autonomous voice. By letting the model "sing to itself," this project explores whether AI, when left unencumbered by human expectation, can engage in a form of self-reflection or reveal traces of an authentic identity.

By treating the generative models as potential subjects of creative autonomy rather than mere tools, we aim to decolonize and support an emerging form of machine agency within a posthuman artistic landscape.

Model: Stable Audio AudioSparx 2.0
Prompt: Artificial Intelligence Singing to Itself