🎯 Lecture Goal

Introduction to the course and understanding the two types of LLMs — Base LLMs vs Instruction-tuned LLMs.

📝 Key Concepts

🤖 Two Types of LLMs

1. Base LLM

• Trained to predict the next word based on large amounts of text data (internet, books, etc.)
• Completes text in a generative way — not necessarily answering your question
• Example: Prompt → "What is the capital of France?"
  • Base LLM might respond with more quiz questions about France, not the actual answer

2. Instruction-Tuned LLM

• Trained to follow instructions and give direct, helpful answers
• Built on top of a Base LLM, then fine-tuned with instruction/response pairs
• Further refined using RLHF (Reinforcement Learning from Human Feedback)
• Designed to be helpful, honest, and harmless — less toxic output
• Example: Prompt → "What is the capital of France?"
  • Instruction-tuned LLM responds: "The capital of France is Paris."

🏆 Which one to use?

For most practical applications → use Instruction-Tuned LLMs. They are easier to use, safer, and more aligned with what you actually want.

💡 Prompting Tips from this Lecture

• Think of prompting an LLM like giving instructions to a smart person who doesn't know your specific task