AI Fundamentals 3 hours
Workshop 1: AI Foundation
In this workshop you'll learn the fundamentals of working with AI chat tools. You'll discover how language models work and learn effective prompting with the 4 building blocks: Task, Context, Role, and Format.
Learning objectives
- Explain how AI language models work, including hallucinations (2% error rate)
- Write effective prompts using the 4 building blocks
- Critically evaluate and validate AI-generated output
Prompts for Workshop 1
Helper prompts for creating and improving effective prompts.
Task to prompt
Turn my task into a clear prompt with: - A specific instruction - Relevant context - The desired result Ask 2 questions first if you need more information. My task: [describe in 1 sentence what you want to do]
Improve a prompt with the 4 building blocks
Download workshop 1 materials
Deep dive
Tips for after the workshop
- 1 Practice daily: Try to complete at least one task with AI every day.
- 2 Always verify: Never blindly trust AI output. Check facts and sources yourself.
- 3 Iterate your prompts: A good prompt rarely comes on the first try. Refine based on output.
- 4 Let AI interview you: Ask "What questions do you have before we start?" AI will then ask for the context it needs.
- 5 Specify your audience: Say "write for an English-speaking audience". AI models are trained with an Anglocentric bias.
Frequently asked questions
Why does AI sometimes give wrong answers?
AI language models 'hallucinate' about 2% of the time: they generate text that sounds plausible but is factually incorrect.
This happens because AI is not a database that looks up facts, but a probabilistic model that predicts the most likely next word. That's why verification is always necessary.
This happens because AI is not a database that looks up facts, but a probabilistic model that predicts the most likely next word. That's why verification is always necessary.
What's the difference between 'create' and 'write' in a prompt?
Word choice strongly influences output. 'Create a report' often produces a different structure than 'Write a report'.
Experiment with action verbs: analyze, summarize, compare, structure. Small changes to your prompt can make a big difference.
Experiment with action verbs: analyze, summarize, compare, structure. Small changes to your prompt can make a big difference.
How do I know if I can trust AI output?
Use the VAC-check:
Verifiable: check facts, numbers, and names against reliable sources
Accurate: check for calculation errors and logical inconsistencies
Consistent: compare with your own knowledge: would a colleague in the field say the same?
AI is ~98% reliable, but always check critical output yourself.
Verifiable: check facts, numbers, and names against reliable sources
Accurate: check for calculation errors and logical inconsistencies
Consistent: compare with your own knowledge: would a colleague in the field say the same?
AI is ~98% reliable, but always check critical output yourself.
How do I make my prompts better?
Two principles: iterate and be specific.
Don't expect the first prompt to be perfect: refine based on output. And add context: who is the audience, what is the goal, what format do you want?
The more specific your question, the more relevant the answer.
Don't expect the first prompt to be perfect: refine based on output. And add context: who is the audience, what is the goal, what format do you want?
The more specific your question, the more relevant the answer.
Reference pages
Read more about Prompting and Hallucinations on the docs site.