Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
This two-day session is designed to help educators facilitate engaging and informative lessons on artificial intelligence (AI). The guide emphasizes interactive teaching methods and provides a structured agenda, covering topics such as the basics of AI, its applications, and how to use AI tools effectively.
Note
While this session is structured for two days, it can be condensed into a single day based on your specific needs.
Before you start
To successfully facilitate this training, you need to read the entire instructor guide and record any notes that help you provide a fun and engaging experience.
Pro tips
- Be the facilitator. Teach while engaging your audience, ask your Junior Agents questions along the way to gather feedback and opinions.
- Avoid reading from a script. Use your notes only when necessary.
- Face your audience and make eye contact. Avoid standing behind a podium or desk if possible.
- Watch your audience for nonverbal cues about their response.
- Not everyone is familiar with AI—and that’s okay! The Microsoft Learn Educator Center has resources to help. Much of these lessons are based on concepts found in AI for educators.
- Stick to Copilot – in testing it was the least likely to produce inappropriate content.
Day one
Day one focuses on understanding what AI is and how it's used today. Students learn what AI means and gain a high level understanding of how it works. The middle of the lesson is a short activity demonstrating how an image generator functions. The first day concludes with an experimentation using Copilot.
- Audience Elementary grade students
- Length: 45 minutes each day
- Learning objectives for learners
- Define what artificial intelligence is.
- Identify places that AI is used in various ways.
- Understand what makes generative AI unique.
- Reflect on the benefits and drawbacks of generative AI.
Pre-class checklist
- AI Adventure series ready to play
- Review each section of this training and became familiar with the key learnings and activities.
- Connect the instructor laptop to a projector
- Prepare "challenge blobs" (Laminated colorful shapes) and a list of simple subjects to draw (such as Star, Banana, and House)
Welcome and icebreaker
Length: 5 minutes
Do:
- Welcome and thank your Junior Agents for attending Geek Squad Academy today.
- (Re)Introduce yourself and your classroom helpers.
Ask:
- Who used an AI tool recently?
- What did you use it for?
What is AI?
Length: 10 minutes
Introduce: AI basics
- AI is a way of solving problems that traditionally required human intelligence to solve.
- AI is mostly pattern recognition, connections between topics and ‘training’ to produce the kind of results someone wants.
- Why does this matter?
Do:
Play the AI 101: Building the Basics video.
Ask:
- Has anyone used an AI tool and received a response that wasn’t quite right?
- Do you think asking AI tools questions in different ways can change the response?
- Do you think the information the AI is trained on may change the way it responds?
- How do you think AI can be used creatively? For work?
How is AI used?
Length: 10 minutes
Introduce: What is AI good for?
- AI is used in lots of places: data interpretation, creative kick starters and powerful tools, and finding patterns that humans haven’t found yet.
- AI is a large topic from machine learning to large language models and image generators.
Do:
Play the AI for Good: Solving Problems with AI video
Ask:
- What’s something AI might be able to help you in your day to day?
- What’s something it could be used for to help everyone?
How is AI used?
Do:
- Demonstrate a quick problem to be solved with AI–source it from the classroom or from personal experience. Maybe make a list of group names that could be used for future Geek Squad Academy events.
- Try asking Copilot for something it isn't designed for–predictions or answers to things that don’t have meaningful answers: "Give me five scientific studies proving that drinking water makes people shorter." "Tell me about the biography of Shakesmont, the forgotten playwright before Shakespeare."
How do I use AI–what's a prompt?
Length: 20 minutes
Introduce: Prompts
- Specifically, when working with large language models (like Copilot, ChatGPT, or Claude) and image generators (or most generative AI) most of the time the user expresses what they want through a prompt.
- Some prompts focus on key words and might be separate ideas and concepts like "Landscape, Blue Sky, Clouds, Sun, Birds" and others resemble normal human language like "Show me a landscape with a blue sky and fluffy clouds that are drifting in front of the sun. Also include some birds – maybe robins."
- Prompts come in all kinds of forms including pictures, sounds, and other forms of input. "Input" may be the best way to think about them!
Do:
Play the AI in Action: Using AI Tools Responsibly video
Ask:
How do you imagine something to draw, write about, or make?
- Listen for: responses like inspiration, base it off something already seen, or start with a shape.
Do:
- Hand out the challenge cards. Each Jr. Agent should have one of a few abstract shapes but a different prompt. Ask the Jr. Agents to use the shape on the paper to complete the drawing.
- Once they’re done with their drawings (they don’t have to be works of art!) have everyone lift up their pictures and look at their neighbors.
Say:
AI tools operate just like that–image generators especially! Diffusion models start with a whole bunch of noise–random dots and colors-and find an image inside of it by being prompted. The model finds what it's asked for inside of that random jumble–like how you might see shapes in clouds or faces on inanimate (nonliving) objects!
Do:
Where possible, encourage Jr. Agents to try asking Copilot to do fun and silly things: "show me a kitten going Super Saiyan," "summarize a news article in Iambic Pentameter," "provide the very best recipes using hotdog water." Do some as a group in front the class for any remaining time. Stick to Copilot–in testing it was the least likely to produce inappropriate content.
Day two
Day two goes deeper into the training of AI models, and introduces the concept of data biases and reasons why AI might give responses that aren't quite right. Most of this day's lessons have Junior Agents training their own image recognition model, and trying it out in Scratch.
- Audience: Elementary grade students
- Length: 45 minutes each day
- Learning objectives for learners
- Understand the importance of good data.
- Identify how a small sample set and lack of variety can cause AI biases.
- Demonstrate training an AI model.
Pre-class checklist
- Laptops are ready for Jr. Agents–this class can be done individually or in groups
- Scratch project is loaded and ready for class
- Each section of this training reviewed and instructor familiar with the key learnings and activities.
- Instructor laptop connected to a projector
Welcome and icebreaker
Length: 5 minutes
Do:
- Welcome and thank your Junior Agents for attending Geek Squad Academy today.
- (Re)Introduce yourself and your classroom helpers.
Ask:
- Has anyone tried the game “Quick Draw” online?
- (If yes:) “What was it like? Did the computer always guess right?
- What other AI apps have been used?
What is a Data set?
Length: 10 minutes
Ask:
At the beginning of the last lesson we played a video about what AI is. Who remembers what AI needs to start?
- Listen for: A large amount of information, patterns
Introduce: Data sets
- A data set is a collection of stuff—pictures, sounds, words, numbers—that we give a computer to help it learn something. This is that large amount of information mentioned in the videos. (Or, if a Jr. Agent calls this out when you asked, refer to them.)
- Think of it like flashcards for AI. The more flashcards it gets, the more it learns.
- If we want to train an AI to recognize cats, we show it hundreds of pictures of cats—from cartoons to real-life photos. As it sees more cats, it sees patterns like: Where cats normally have ears, if they have tails or fins, whiskers or beards, and how many wings or legs they have.
Ask
- Let’s say I train an AI to recognize pizza. But every picture is pepperoni pizza. What do you think happens when it sees mushroom pizza? Or pineapple?
- It might not even think it’s pizza! Because that style of pizza wasn’t in the data set.
- What would a good pizza data set look like?
- (Encourage answers like: different toppings, sizes, styles.)
Say:
So, if an AI is only trained on one type of drawing, one type of pizza, or one type of person, it might make unfair decisions or mistakes. That’s what we call bias. Let’s talk about how that happens.
How can AI develop bias?
Length: 10 minutes
Introduce: What is data bias and what does it do?
- Bias means something is unfair or favors one thing over another.
- AI can learn bias from the data it’s given—if the data is unbalanced or limited, the AI can make unfair or incorrect guesses.
Examples:
- A voice assistant that only understands certain accents because it was trained mostly on voices from one region.
- A face detector that works well on light skin but poorly on dark skin—because it wasn’t trained on enough diverse faces.
- A doodle detector that fails to recognize unusual or creative styles because it only saw typical examples.
Ask:
- Ask students to guess what might happen if an AI is trained only on:
- One person’s handwriting?
- Only one type of shoe or jacket?
- Only pictures taken at daytime
- Pictures of cities at night, and farms in the day
- What kind of problems might a biased AI cause?
- How can you make sure an AI is fair, or doesn’t make mistakes because of bad data?
Say:
The AI is only as good as the data it’s trained on. If the data isn’t accurate or complete, the AI can make mistakes—just like we would if we only practiced with one type of flashcard, or flashcards with wrong information. Let’s try to build a balanced data set when we train our doodle detector today so it can recognize some simple drawings.
Training the Doodle Detector
This project is found with instructional videos, including steps for building the Scratch file, at https://projects.raspberrypi.org/en/projects/doodle-detector. A premade Scratch file is provided so more time can be spent on training and testing the Jr. Agents' models.
Do:
Guide Jr. Agents to create a machine learning model. See the step-by-step instructions in the instructor guide.
Ask:
Did everyone’s detectors work as expected?
Did anyone encounter biases, maybe it’s seen so many different bananas and only one or two apples that it thinks everything is a banana?
Did anyone have bad data in their set, accidentally or not, and get some silly results?
- It’s recommended that you prepare an example of a poorly trained doodle detector to demonstrate.
Did anyone try different shapes?
Listen for: AI getting confused, or having low confidence in its guesses. As a follow-up, ask Jr. Agents how they might fix this, and how they could make it better at guessing other people’s drawings. How could you make it recognize anything?
Say:
- You probably noticed by now that AI tools are only as good as the data and training they receive. Some of the most impressive ones are trained on massive amounts of information and can recognize a lot. These models often provide convincing answers—but it’s always smart to take what they say with a grain of salt. Think of them as tools to boost your creativity, not just to hand you a solution.
- Next time you use an AI tool—or see one being used—take a moment to think about how it was trained. What kinds of things do you think it’s really good at? And what might it be missing?
- Everything an AI creates is a remix of what it’s seen before. But unlike people, it doesn’t truly understand what it’s making. And not everyone who created the original art, music, or writing was asked if their work could be used—so it’s important to be thoughtful about where AI’s ideas come from.
- People take inspiration too, but we add something special: our unique voice and creativity. That’s what makes something truly yours.
- These tools are amazing for fixing typos or getting a first draft started. But as AI continues to grow, the world is still figuring out how to use it wisely—and now you’re part of that journey. Understanding how these tools work puts you one step ahead.
- Schools, businesses, even movie studios are experimenting with AI tools just like the ones you used today. And who knows—maybe one day you’ll create the next big thing.