Logic vs Learning Slides Lesson 1
Logic vs. Learning
How do we tell a computer what to do? The difference between standard coding and machine learning.
The Rule Master Challenge
Imagine I give you a box. Every time you put a blue block in, a red one comes out. Every time you put a yellow block in, a green one comes out.
"Is the box thinking, or is it just following a rule?"
Input → Rule → Output
Standard Coding: Explicit Instructions
The "Recipe" Approach
Human writes specific IF-THEN rules.
The computer follows them exactly.
If something new happens, the computer gets stuck.
// Rule for a Cat Photo
IF (has_pointy_ears) {
AND (has_whiskers)
AND (says_meow)
RESULT = "Cat";
}
What happens if the cat is a hairless Sphinx with rounded ears? The code fails!
Machine Learning: Pattern Recognition
Instead of giving rules, we give examples .
The "Childhood" Approach
Computers look at 10,000 photos of cats. They find the patterns themselves. They learn what makes a cat look like a cat.
🐱
🐈
🦁
🐯
🐆
...
MODEL CREATED
Activity: The Human Algorithm
We are going to test both methods in our classroom!
Round 1: The Rule Book
1
I will give you a specific set of rules (The Code).
2
You must follow the rules exactly to identify "Zoggles."
3
Then, we'll try it again—but this time, I won't tell you the rules. You'll have to learn from patterns!
Get into pairs!
One "Computer," one "Input Manager"
Human Algorithm Worksheet The Zoggle Detective
Unit: Machine Learning Masterminds | Activity 1.1
NAME: ____________________________
DATE: ____________________________
PART 1
The Rule Book (Standard Code)
In this round, you are a computer following strict rules. Use the "Rule Book" below to decide if the objects shown are Zoggles .
Zoggle Rule Book:
IF the object has exactly 3 sides AND is shaded blue → ZOGGLE
IF the object has more than 4 sides AND has a dot in the middle → ZOGGLE
OTHERWISE → NOT A ZOGGLE
Object A
Yes
No
Object B
Yes
No
Object C
Yes
No
Object D
Yes
No
PART 2
Pattern Recognition (Machine Learning)
In this round, you have no rules! Look at these examples of "Greebles" and "Not Greebles" to find the pattern yourself.
Examples of GREEBLES:
A1
B2
C3
Hint: They all follow a pattern involving letters and numbers.
NOT GREEBLES:
AA
44
1C
Now classify these mystery objects:
D4
Greeble Not
5E
Greeble Not
Reflect:
1. Which part felt easier for you: following the Rule Book or finding the Pattern ? Why?
2. What happens to the "Rule Book" if I introduce a green triangle? Does the computer know what to do?
Power of Data Slides Lesson 2
The Power of Data
Building your first AI model. Why what you show the computer matters more than what you tell it.
What is "Training Data"?
Training data is the collection of examples we give a machine so it can learn.
"Garbage In, Garbage Out."
If you give a computer bad data, it will learn bad patterns!
The 3 Rules of Good Data
Quantity
The more examples, the better!
Variety
Show the object from different angles and in different light.
Accuracy
Don't mix up your labels!
Let's Build: Teachable Machine
We are going to use Google's Teachable Machine to train a computer to recognize YOUR gestures.
Step 1: Capture Samples
Step 2: Train Model
Step 3: Test
STATUS: READY TO LEARN
The Edge Case Challenge
What happens when the data changes slightly?
👍
Class A
Trained with 100 photos of a bare hand giving a thumbs up.
🧤
The Test
Will it recognize a thumbs up if you are wearing a thick winter mitten?
WHY OR WHY NOT?
Model Maker Lab Sheet Model Maker Lab
Unit: Machine Learning Masterminds | Lab 2.1
NAME: ____________________________
DATE: ____________________________
Lab Mission: The Gesture Recognizer
You will train a model to recognize three different gestures. Then, you will "stress test" it to see how well it learned.
Class 1
Thumbs Up
Goal: 50+ Samples
Class 2
Peace Sign
Goal: 50+ Samples
Class 3
Neutral (Nothing)
Goal: 50+ Samples
Phase 1: Initial Testing
Test Scenario Model Confidence % Result (Correct/Incorrect) Standard Thumbs Up Standard Peace Sign Neutral Face (No hands)
Phase 2: The Stress Test
Try these variations and record if the model still works.
Change the Lighting
(Turn off lights or use a flashlight)
Worked Failed
Change the Distance
(Back up as far as you can)
Worked Failed
Use the "Wrong" Hand
(If you trained with right, use left)
Worked Failed
Add a Distraction
(Wear a hat or hold a pencil)
Worked Failed
Lab Conclusions:
1. Which stress test was most likely to break your model? Why do you think that happened?
2. How could you fix the problem you found in Question 1 without writing any code?
Teachable Machine Teacher Guide Teacher Facilitation Guide
Unit: Machine Learning Masterminds | Lesson 2
TEACHER RESOURCE
Lesson Overview
In this lesson, students will transition from theoretical "human" patterns to actual machine learning. Using Google's Teachable Machine , they will see how a computer builds a model from image samples.
Tech Requirements
Chromebook/Laptop with Camera
Internet access (teachablemachine.withgoogle.com)
Ensure site is whitelisted by IT
Setup Tips
Avoid direct backlight (windows)
Pairs work best for capturing samples
Encourage 100+ samples per class
Lesson Flow (50 Minutes)
0-10 Min
Introduction & Slides
Explain "Training Data." Key concept: The computer doesn't know what a 'hand' is; it only knows the pixel patterns in the photos you provide.
10-35 Min
The Lab Activity
Guide students to create 3 classes: Thumbs Up, Peace Sign, and Neutral. Encourage them to move their hands around so the model learns variety.
Discussion Prompt: "If your model is 100% sure a peace sign is a peace sign, is it always right? Try making a peace sign with just your pinky and ring finger. What happens?"
35-50 Min
Stress Testing & Reflection
Students test their models with different variables (lighting, mittens, distance). This builds intuition for Lesson 5 (Bias).
Common Student Pitfalls:
Label Swap: Students accidentally record a 'Thumbs Up' while the 'Peace Sign' label is active. This is a great teaching moment for "Bad Data."Static Background: If a student stands perfectly still for the 'Neutral' class, the model might think the wall behind them is the reason they are 'Neutral'. Tell them to move around!Overfitting: Only training with the right hand, then being surprised when the left hand fails. Curation Loop Slides Lesson 3
The Curation Loop
How platforms like YouTube, TikTok, and Netflix know what you want to watch next—and why they might be too good at it.
The Recommendation Engine
You
A.I.
Algorithm
Viewing History
Likes & Interactions
Watch Time (Wait... did you skip?)
It's all about Prediction!
The computer builds a profile of your interests. Its goal is to keep you engaged —meaning it wants you to stay on the app as long as possible.
"If you liked 3 videos about skateboarding, what is the probability you will like a 4th one?"
The Echo Chamber Effect
The Problem
Algorithms show you more of what you already like . This sounds great, but...
You stop seeing new perspectives.
You think everyone agrees with you.
The world feels smaller.
You are here
Simulation: You are the Algorithm
We are going to divide into "Users" and "Algorithms." Users will pick content they like, and Algorithms must predict what they'll choose next using a secret profile!
The User
Picks 3 cards based on a secret personality.
The Algorithm
Watches the choices and identifies the category.
The Prediction
Suggests a 4th card. Did they click it?
Recommendation Engine Simulation Worksheet The Content Engine
Unit: Machine Learning Masterminds | Activity 3.1
TEAM: ____________________________
DATE: ____________________________
How the Simulation Works:
1. The "User" Role:
Secretly chooses a **Persona** (e.g., "Gamer," "Baker," or "Sports Fan"). They pick 3 content cards that match their persona from the deck.
2. The "Algorithm" Role:
Watches the 3 cards. They must analyze the **tags** on the back and choose a 4th card they think the user will "click."
🎮
Minecraft Tips
Tags: gaming, sandbox, tutorial
🍪
10-Min Cookies
Tags: food, baking, quick
🏀
NBA Highlights
Tags: sports, basketball, recap
Algorithm Log:
USER CHOICE 1
USER CHOICE 2
USER CHOICE 3
ALGO PREDICTION
Thinking Critically:
1. If a User starts watching "Minecraft" videos, does the Algorithm eventually stop showing them "Cooking" videos? Why is that a problem?
2. How could you "reset" your algorithm in real life if you felt like you were in an echo chamber?
Creative Computers Slides Lesson 4
Creative Computers
From ChatGPT to DALL-E. How machines "dream" up new content based on what they've learned.
The Great Shift: From Sorting to Creating
LESSONS 1-3
Discriminative AI
"Is this a cat or a dog?"
🐈
CAT
Focus: Sorting and classifying existing data.
TODAY
Generative AI
"Draw me a cat wearing a tuxedo."
PROMPT
🤵🏻🐱
Focus: Creating brand new content based on patterns.
Prompt Engineering: The Art of Specificity
The AI is only as good as your instructions .
The Formula:
[Subject] + [Action] + [Environment] + [Style]
Example: "A robot (Subject) drinking tea (Action) on the moon (Environment) in the style of Van Gogh (Style)."
Garbage Instructions = Garbage Results
Real or Robot?
Can you spot the A.I. generated content?
🍕
"A high-quality photo of a pepperoni pizza with melting cheese and steam rising from the crust."
🧑🍳
"A chef smiling while tossing pizza dough in a rustic Italian kitchen with warm lighting."
LOOK AT THE FINGERS!
Pro Tip: AI often struggles with complex details like hands, text, or consistent lighting.
Prompt Engineering Workshop Worksheet Prompt Engineering Workshop
Unit: Machine Learning Masterminds | Lab 4.1
NAME: ____________________________
DATE: ____________________________
The Master Formula
SUBJECT
A Golden Retriever
ACTION
riding a skateboard
ENVIRONMENT
in Times Square
STYLE
Cyberpunk 2077
Challenge 1: The Iterative Prompt
Start with a basic idea and make it better by adding details.
Version 1: Simple
"A cat on Mars."
Version 2: Adding Detail
"A ginger cat wearing a space suit on the surface of Mars, with red dust everywhere."
Version 3: Master Class (Your Turn! Add 'Style' and 'Action')
"____________________________________________________________________"
Challenge 2: The Hallucination Hunt
A.I. sometimes makes things up or gets details wrong (hallucinations). Look at the prompt below and list 3 things an A.I. might struggle to draw accurately.
"A classroom full of students writing with pencils on paper, where all the clocks on the wall show exactly 3:00 and the posters have perfectly readable text about the solar system."
Trouble Spot 1
Trouble Spot 2
Trouble Spot 3
The Ethics of Creation
If an A.I. draws a picture in the style of a famous living artist, who "owns" that picture? The A.I. company, the artist, or the person who wrote the prompt?
The Artist
The Prompt Writer
The AI Company
Explain your choice:
Algorithmic Bias Slides Lesson 5
Broken Mirrors
Identifying algorithmic bias. How unfair data creates unfair machines.
What is Algorithmic Bias?
"Bias occurs when an A.I. system consistently makes unfair or inaccurate decisions for certain groups of people."
It's not that the computer is 'mean.'
The computer is just a mirror. If the data we give it is biased, the results will be biased too.
Case Study: The "Invisible" Faces
In 2018, researchers found that facial recognition software was 99% accurate for light-skinned men, but 35% less accurate for dark-skinned women.
Why did this happen?
The Training Data was mostly photos of the programmers themselves—who were mostly light-skinned men.
Included in Data
Left Out
Included in Data
Included in Data
Who is Responsible?
If an A.I. makes a biased decision, who should we blame?
The A.I.?
But the computer only knows what we show it.
The Coder?
They chose the data and built the model.
Society?
The data comes from a world that isn't always fair.
Discussion: What can we do to fix it?
Bias Detectives Case Study Worksheet The Bias Detectives
Unit: Machine Learning Masterminds | Final Reflection
NAME: ____________________________
DATE: ____________________________
Scenario: The Lunch Line Robot
Your school builds an A.I. robot to predict what students want for lunch so it can pre-package meals and save time. The robot was trained by looking at what students in the 8th Grade Honors Class ordered for the last three months.
1. Spot the Bias
When the robot is moved to the 6th grade lunchroom, it starts making errors. It keeps offering "Kale Salad" to students who want "Pizza," and it doesn't recognize that many 6th graders have nut allergies.
Why is the robot's training data biased?
Think about who was in the training data versus who is using the robot now...
2. Fix the System
You are the lead developer. List three ways you would change the training data to make the Lunch Line Robot fair for everyone.
1
2
3
Final Unit Reflection:
What is the most important thing a person should know about A.I. before they trust it to make decisions?
Unit Completion
Machine Learning Mastermind
