Day 3: Into the Black Box
A 15-minute activity to understand how you can use the mini LLM board.
In this third tutorial, you'll learn how to use the LLM Mini Board to help others understand how AI processes their requests.
What to print:
Just print the A4 sheets from tutorial-day3.pdf (in the Tutorial folder of your Starter Kit)
1. Basic Setup
- Place the LLM Mini Board in front of you
- Put the Text Input and Output cards in their slots
- Add the Review Task card in the task slot
The LLM Mini Board helps visualize how AI processes your requests
2. Think It Through
Look at your combination and write down:
- What could this AI setup do?
- What kind of review tasks would work well?
Example use case: An AI that reviews technical documentation
- Input: Your documentation draft
- Task: Review for clarity and completeness
- Output: Suggested improvements and missing topics
3. Oh no, AI Hallucinates!
- Place the Hallucination Risk card next to the output card. It symbolizes an hallucinated output!
- Think about:
- How could your AI review setup hallucinate?
- What would be the consequences?
In our documentation example:
- AI might invent non-existent technical specifications
- Could suggest "missing" topics that are actually irrelevant
- Might make up false references or sources
See how you can combine two cards (here text output + hallucination) to create a new meaning?
4. Let's mitigate the risk
Look at your Web Search Capability card:
- Think about:
- How could web search help with hallucinations?
- What sources would be most reliable for your use case?
- What are the limitations of web search?
- Place the Web Search card onto the capability placeholder of the mini board
For our documentation review example:
- AI can verify technical specs against official documentation
- Can cross-reference features with product release notes
- But remember: documentation might be outdated or private!
5. Reflect as a Trainer
The LLM board turns abstract AI concepts into tangible objects people can touch and manipulate. This physical interaction often leads to better understanding than slides or diving straight into using an AI assistant!
Think about how this mini LLM board activity helps explain AI:
- Physical Representation: See how the board makes abstract concepts like "input", "task", "output", and "capabilities" immediately tangible.
- Adding more cards progressively: See how you start with a basic setup and then add more cards to explain more complex concepts?
- For learning or brainstorming: See how the different combinations of cards can be used for both learning concepts or brainstorming use cases?
- Problem and Solution: Notice how introducing the Hallucination card creates tension (a problem) that's later resolved by the Web Search card (a solution)?
Finally, a super important point:
- Demonstrate how it works on a real AI assistant: In a real workshop, you'd alternate between the card representation and trying things out in ChatGPT. This back-and-forth between theory and practice helps participants understand both the concepts and their real-world application.
What's Next?
Congratulations! You've completed the solo tutorial, tomorrow, go to the next page, reflect on what you've learned and start planning your next steps.
To explore the full possibilities of the LLM Mini Board, you will need the first expansion pack: The Modern AI Assistant. You'll get the full set of input, output, and capability cards.