SLIDES: https://duke.is/llm-nsoe
An invited presentation with Jenna Strawbridge to the NSOE on ways that LLMs may replace traditional resources. The presentation will consider the following:
- What are the tendencies of the big three: ChatGPT [and MS CoPilot], Claude, Gemini, plus GitHub CoPilot
- The importance of verification
- Are the responses answers?
- Do I want a prediction of clean data or do I want to verify that the data re clean?
- Prompting out a good "response" in general, or for AI-assisted coding. i.e. Prompt Engineering
- Approach Rubrick
- Do I want type-ahead buffers (e.g. AI-paired / or AI-assisted coding)
- Natural language as an alternative avenue into documentation?
- Reproducibility
- Are the responses reproducible?
- Is the code reproducible?
Where does AI excel and miss: best practices when using AI.
See Also: Kerry Ossi-Lupo's Resource Page
John Little
Center for Data & Visualization Sciences
Duke University Libaries
CC BY (See License.md file)