← Back to Futures
mid mixed B 4.07

The Parish of Small Models

When energy self-sufficiency meets open hardware, creative AI splinters into locally owned cultural models that reflect the memory, language, and values of specific communities instead of global recommendation systems.

Turning Point: A federation of municipalities begins funding community media grants that can only be used on locally trained models running on locally owned hardware, turning cultural infrastructure into a public utility rather than a platform feature.

Why It Starts

The era of one feed shaping everyone at once begins to fray. Neighborhood archives, regional dialect collections, church choirs, migrant radio stations, and public theaters all train their own small models to generate stories, music, posters, and video in their own voice. The result is a creative boom rich in texture and place. Yet each community also gains new tools for curating its boundaries, deciding what counts as authentic memory and who gets to speak for it.

How It Branches

  1. Open hardware and cheaper local energy make it practical for small institutions to host their own generative systems.
  2. Community groups realize that locally trained models can preserve dialect, humor, and ritual that mass platforms flatten away.
  3. Public arts funding shifts toward locally governed model infrastructure, accelerating a patchwork of distinct cultural engines.

What People Feel

At 9:30 p.m. in a coastal town hall in Jeju, a documentary editor asks the local model to score a film about winter divers using old folk rhythms from the municipal archive. The machine suggests three versions, each shaped by recordings donated by different families, and the room goes quiet as elders argue over which one sounds most like home.

The Other Side

Cultural plurality can slide into cultural enclosure. Local models may revive neglected traditions, but they can also freeze identity into a guarded canon, reward gatekeepers, and make outsiders legible only through inherited stereotypes. A world with many voices is not automatically a world with easy listening.