Working with LLMs Workshop

6 hours of instruction

Introduces participants the core concepts and tools to use Large Language Models (LLMs) in practice. Participants will learn the concepts underlying LLMs and what they can — and cannot — do.

PREREQUISITES

Participants should be comfortable programming in some language commonly used in data science (e.g., Python, R, SQL, etc.). Some prior exposure to machine learning is useful but not mandatory.

LEARNING OBJECTIVES

1. Describe the current landscape of LLMs in use (e.g., ChatGPT, Bard, etc.)
2. Define the meaning of “training data” in the context of LLMs (& possible legal controversies)
3. Explain and use common LLM inference-time parameters (e.g., Temperature, Top-P, Top-K, etc.) to tune LLM outputs
4. Generate SQL queries using an LLM from colloquial language
5. Generate code (possibly in unfamiliar programming languages)
6. Parse LLM output carefully to identify possible errors or “hallucinations”