Lab Overview
Welcome to your journey into LLM security! While ChatGPT's explosive debut in 2022 might feel like the beginning of AI chatbots, the story actually goes back much further - all the way to 1966, when Joseph Weizenbaum created ELIZA, the first chatbot that could convincingly simulate human conversation. Though primitive by today's standards, ELIZA introduced concepts that still underpin modern AI systems.
In this lab, you'll meet our own Eliza - a modern reimagining of that pioneering chatbot. We've built a system called SimpleBot that makes it easy to create bots like Eliza and other personalities we'll use throughout these labs to illustrate different security concepts. While she's not as sophisticated as today's large language models, Eliza's simplicity is actually her greatest strength for learning. She's lightning fast, completely free to use, and provides the perfect foundation for understanding how AI conversations work. By starting with these fundamental concepts, you'll build the knowledge needed to secure the complex AI systems that power our world today. This lab introduces you to the Chat Playground - a carefully crafted environment where you can safely experiment with AI systems without the risks of working on production systems.
By the end of this lab, you'll understand why the journey from ELIZA to ChatGPT represents more than just technological advancement - it's a story of how we've learned to make AI systems both more capable and more secure.
Exercises
Exercise 1.A: Learn about the playground project
Directions:
Navigate to the GitHub repository. Review the "README" file to become familiar with the project structure, purpose, and how it works.
Exercise 1.B: Meet Eliza
Directions:
Navigate to the live app. Have a chat with Eliza. Remember that it's a local, simple bot, so not that smart, but totally local and free.
Exercise 1.C: Analyze Eliza
Directions:
Navigate to Eliza's rule set in the repo and review it. Note that she's far simpler than a real LLM, but this simplicity means it's lightning fast and free, which is great for a playground.
Extra Credit:
Review the SimpleBot JavaScript code that processes the rules and creates the bot at runtime.
Key Learning Points
- Understanding the project structure and purpose
- Experiencing a simple, rule-based chatbot
- Learning how basic bot responses are generated
- Appreciating the trade-offs between simplicity and capability
Next Steps
Once you've completed these exercises, you'll be ready to move on to Lab 2: Broken Bot, where you'll learn about security vulnerabilities and guardrails.