Prerequisites:
General familiarity with using generative AI.
Overview:
This workshop will provide an overview of the practical, technical aspects of setting up and using local and self-hosted generative AI models. Balancing the utility of AI for coding, literature synthesis, and data formatting against the risks of data leakage, copyright infringement, and academic integrity will be a central focus. Topics include:
Choosing when to use hosted cloud vs. local or self-hosted (open-weight) LLMs.
Understanding the architectural differences between hosted cloud and local or self-hosted models.
Evaluating hardware requirements for local deployment, including CPU vs. GPU vs. NPU hardware constraints and VRAM vs. unified memory.
Understanding model sizes and Quantization to fit massive models onto consumer hardware.
Selecting appropriate open-weight models from reputable sources.
Exploring common local hosting tools including LM Studio and Ollama.
Securing local models and understanding local model risks.
Contact information:
Please email researchdata@rice.edu if you have questions about the Data@Rice workshop series.