Tuesday, November 17 – 10:55 am to 12:15 pm – Operations
Presenter: Matt Landry (Cambridge Public Schools)
Instead of relying on corporate, cloud-based AI platforms, what if community media centers hosted their own local AI models for internal use and to serve the public? This session explores a growing question of civic tech: running open-source Large Language Models (LLMs) entirely on local computers to create privacy-first, community-specific tools.
Many community media leaders want AI’s benefits without handing community data to big tech or leaning on distant data centers. This panel asks a practical question: can a media center run local models on its own hardware to serve both staff and the public? We’ll walk through a three-tier approach, from internal tools like grant support or archive search, to public question-answering over meeting records and community resources, to connecting residents with services. We’ll be candid about limits: hardware, hallucination risk, how much concurrent traffic one machine can handle, and who is responsible when an answer is wrong. And we’ll look honestly at the environmental argument, which is more complicated than it sounds. Attendees leave with a hardware and software starting point, a risk framework, and criteria for deciding what to build locally and what not to.

