Session 1 activities
ACTIVITIES · 27 MINFictional data only
You can do every activity here. Sign in to save your responses and add your vote to the class tally.
What does the model actually do?
Poll at the start of session 1, then discuss.
When Cape Harbor Inn's (fictional) assistant answers 'What time is check-out?' from its FAQ, what is the model doing at its core?
More tokens or fewer?
For each string, guess whether a typical English-trained tokenizer splits it into one token or several. Then check in Lab 1.
- breakfast
- Chatham
- 8:15
- desayuno incluido
- the
Price the FAQ bot
Branching. The Cape Harbor Inn FAQ is about 1,950 tokens. Each question sends the whole FAQ plus 100 tokens of instructions and question, and gets a 150-token answer. Practice prices: $1.00 per million input tokens, $4.00 per million output.
What is the cost per question?
Explain-it-back: tokens and attention
After session 1, in pairs, then post.
In three or four sentences a coworker would understand, explain what a token is and what attention lets the model do. Use one Cape Harbor Inn example.
- Token defined in plain words.
- Why token counts matter (cost, limits).
- What attention does with context.
- One concrete example.