Trevor Murphy
Instructional Technology Specialist,
Williams College
Keep Your Data Close and Your AI Closer: Local AI in the Academy
Tuesday, April 01, 2025 | 8:30AM–9:15AM ET | Room 550 AB
Session Type:
Breakout Session
Delivery Format:
Presentation/Panel Session
While cloud AI is popular, it introduces challenges with data privacy, data ownership, and other complex considerations. One way Williams College has leveraged AI is by using a selection of local AI tools for different teaching and research needs: from transcription to document research and accessibility to supporting students becoming critical experimenters with LLMs rather than just users. In this session, we explore how three specific local AI tools and use cases have contributed to our ongoing research, learning, and discussion around AI on campus in a way that goes beyond much of the current (and we argue, limiting) focus on LLMs as tied to cloud-based applications/architecture: (1) OpenVINO, a generative AI plugin for the open-source application Audacity that enables fully local transcription and translation of audio and video files used to make subtitle files for films in different languages; (2) OpenWebUI + Ollam, a desktop application and service that enables fully local, retrieval-augmented generation (RAG) analysis of private archival document collections; and (3) EphBot, a Raspberry Pi loaded with the open source OLMo LLM designed for a plug-and-play local Wi-Fi chatbot that provides users full access to adjust and experiment with LLM settings and system prompts. EphBot is packaged to be borrowed at the Williams College library circulation desk.
