The Difference Between a Song and a System

For most of recorded music history, a song has been treated as a finished object. It is written, recorded, mixed, mastered, and released.

AI introduces a different possibility: music that is never fully finished.

Instead of creating one definitive track, an artist could build a system that continuously generates variations. The structure, instrumentation, intensity, or mood could change depending on the listener, the environment, or what is happening in real time.

At that point, the artist is no longer only writing music. They are designing the rules that determine what the music can become.

This idea existed long before today’s AI boom. Brian Eno’s work with generative music explored systems capable of producing ongoing variation rather than one fixed composition. AI pushes that concept further by allowing music to respond dynamically to data, behavior, and context.

Games already offer an early version of this through adaptive music systems, where soundtracks change according to player actions or environments. AI could go further, generating musical material itself rather than simply moving between pre-composed sections.

That creates a deeper question about authorship.

If an artist defines the sound, boundaries, and emotional direction of a system but does not manually choose every note, what exactly are they creating?

The rules may become part of the composition.

It also changes what a “song” means. If two listeners hear different versions, are they experiencing the same work? And if the music keeps evolving, what does it mean to release or preserve a definitive version?

The music industry is still built around fixed tracks, albums, and masters. AI may eventually challenge that model by turning music from a finished object into a living system.

The most interesting question may no longer be whether AI can write a song.

It may be whether the future of music needs to be a song at all.