Learning from Examples
Why more examples usually means a smarter AI.
Imagine you want to teach a robot to sort your Halloween candy into two groups: chocolate and not-chocolate. You could write a giant list of rules, OR you could just show it hundreds of candies and tell it which group each one belongs to. AI learns by seeing examples — millions and millions of them!
Training Data: The Food AI Eats
The examples that AI learns from are called training data. Think of training data like the food you feed to your brain. If you eat healthy, balanced food, your body works well. If your training data is good, balanced, and varied, the AI works well too!
Let's say you want to teach an AI to recognize pictures of apples. You'd show it thousands of apple pictures — red apples, green apples, big apples, tiny ones, perfect ones, bruised ones. The more different kinds of apples it sees, the better it gets at recognizing ANY apple.
Now imagine you only showed it pictures of perfect red apples from the grocery store. It might not recognize a small green apple from someone's backyard! That's because its training data was too narrow — it didn't include enough variety.
The QUANTITY and VARIETY of training data matters! More examples + more diversity = a smarter, fairer AI.
How AI "Practices"
When AI is learning, it doesn't just look at examples once. It looks at them over and over, making tiny adjustments each time it gets something wrong. It's like practicing a sport — the more you practice, the better your brain gets at it.
Every time the AI gets an example wrong, it adjusts its thinking a little. After millions of adjustments, it gets really good! This process of adjusting is called training, and the AI that results is called a model.
The whole training process can take weeks of computer time and uses enormous amounts of electricity. Building a good AI is a huge amount of work — which is why it's so important that the training data is good quality.
Wow! Some AI language models (the kind that can write stories and answer questions) have learned from more text than a person could read in thousands of lifetimes! That's billions of web pages, books, and articles.
🏫 The Unfair Program
A school district decided to use an AI to help pick students for a special advanced program. The program had been around for 20 years, and over that time, most of the students selected came from one wealthy neighborhood in town.
The AI was trained on 20 years of data about which students had been picked before. So it learned a pattern: students from that neighborhood tend to get selected.
But here's the problem — students from other neighborhoods were just as smart! They just hadn't been selected as often in the past because of other unfair reasons.
Now the AI was repeating the same unfair pattern, and students from other neighborhoods were being skipped over again. It wasn't the AI's fault exactly — it just learned from data that was already unfair.
🔑 CCR Connection — Think, Create, and Be Responsible!
When AI makes decisions that affect people, ask: Where did it learn? Was that fair and balanced?
If you're creating something with AI, feeding it good, varied examples helps it do a better job!
If you notice AI being unfair, it's okay to say something. Fair AI is everyone's responsibility.
When the Examples Are Wrong
Bias — the word for when AI is unfair without meaning to be.
What if your teacher only taught you about one country in social studies, and then you were tested on ALL the countries? You'd probably only know about that one country! When AI only learns from certain types of examples, it can miss a lot — and that's called bias.
What Is AI Bias?
Bias is when something is unfairly slanted toward one group of people or one way of thinking. AI bias happens when the training data wasn't fair or balanced — and so the AI learned to treat some people or ideas differently than others.
Here's a real example: A teacher asked an AI image tool to make pictures of "a doctor." All the pictures showed men. When she asked for "a nurse," all the pictures showed women. The AI wasn't trying to be unfair — it just learned from old images on the internet, and many of those old images showed doctors as men and nurses as women.
This is a problem because it teaches people (especially kids!) that only certain kinds of people can be doctors or nurses — which isn't true at all! Doctors and nurses can be any gender.
AI bias usually isn't on purpose — it comes from unfair patterns in the training data. But even accidental bias can hurt real people.
Bias Can Look Different Ways
Bias shows up in lots of places. Sometimes AI doesn't know much about languages other than English because most of its training data was in English. Sometimes it doesn't recognize faces of certain ethnicities as well as others because its training photos mostly showed other groups.
Francisco in your class asked an AI for book recommendations, and it only suggested books by one type of author. That's bias in action! The AI learned from bestseller lists that may have been unfair, and now it repeats that unfairness.
The good news is that people are working hard to make AI fairer. And YOU can help by learning to notice bias and asking good questions — like "Did this AI show me all kinds of people, or only some?"
Real Story! The author of your book, Kathi Kersznowski, once asked an AI image generator to make pictures of "pilots" — and every single one was a man! But Kathi used to fly planes! AI doesn't always reflect the real world.
📚 Francisco's Book List
Francisco needed to find books for a big school project. He asked an AI assistant to recommend some great books to read.
The AI gave him a list — but when Francisco looked at it, he noticed something. Every single book on the list was written by a white male author. No women. No authors of color. No books from other cultures.
"Why didn't it suggest books by girls or people from different backgrounds?" Francisco wondered. "Where are those other voices?"
Francisco decided to ask the AI again, but this time he said: "Can you recommend books written by authors from different cultures, including women and people of color?" This time, the AI gave him a much more diverse list!
Francisco learned something important: you can push back on AI bias by asking better questions.
🔑 CCR Connection — Think, Create, and Be Responsible!
When AI shows you examples of people, check: Are all kinds of people represented, or only some?
You can improve AI's answers by asking more specific and inclusive questions!
Pointing out bias isn't being difficult — it's being responsible. Everyone deserves to see themselves represented.
AI Doesn't Think Like You
Why AI and your brain are totally different.
You've been through a lot of experiences in your life — you've felt happy, sad, excited, and scared. You have a family, friends, and memories. AI has none of that. And understanding this difference is really important!
No Feelings, No Experiences, No Common Sense
You might talk to a chatbot and it seems really friendly — like it cares about you. But here's the truth: AI doesn't actually have feelings. When a chatbot says "I'm so happy to help you!" it's not actually happy. It just learned that friendly-sounding responses make people feel good, so it uses friendly language.
You also have something called common sense — knowledge about how the world works that seems obvious to you. Like knowing that a sandwich won't start talking to you, or that if you drop something it falls down.
AI doesn't have common sense the same way you do. It can make huge mistakes on things that seem obvious to any human, because it never actually lived in the world and experienced it. It just read about the world, which is very different.
AI Patterns vs. Real Understanding
Here's a way to think about it: imagine someone who has read every recipe book ever written but has never actually eaten food. They might be able to describe flavors, write new recipes, and answer questions — but they've never actually tasted anything!
That's kind of like AI. It has processed an unbelievable amount of information, but it hasn't experienced any of it. It doesn't know what a strawberry actually TASTES like. It doesn't know what it feels like to be nervous before a big test.
This means when AI writes a story with feelings, or gives you advice about a hard situation, it's working from patterns it learned — not from actually understanding what you're going through. A good friend who knows you might give better advice than an AI, even if the AI sounds really smart.
AI is incredibly good at patterns and information — but it doesn't truly understand the world the way you do. Your brain is still amazing in ways AI can't replicate!
Think About This! You can look at a drawing your little sibling made and immediately understand it's a house, even if it looks like a lopsided box with a triangle on top. AI can struggle with things that are obvious to you! Your brain is incredible.
💬 The Homework Helper Mix-Up
Lena was feeling really sad because her best friend was moving to another city. She asked a chatbot: "My best friend is moving away and I feel terrible. What should I do?"
The chatbot gave a very organized, detailed list: "1. Stay in touch by text. 2. Plan video calls. 3. Make new friends at school. 4. Visit during school breaks."
"It's technically right," Lena thought, "but it feels like advice from a robot."
She showed it to her older brother, who just said: "That really stinks. I'm sorry. Moving friendships are hard, but the best ones survive distance."
Lena felt so much better after talking to her brother. The chatbot had information, but her brother had understanding.
AI can help with lots of things — but some things really need a human.
🔑 CCR Connection — Think, Create, and Be Responsible!
When AI gives you advice about feelings or personal situations, think: Does this actually understand what I'm going through?
AI is great for information tasks! But for creative and personal things, YOUR ideas and feelings make it better.
Some conversations — about how you feel, hard problems, or important decisions — are better with a real person who knows and cares about you.
Quiz Time! 🎯
Show what you learned — you've got this!
Amazing Work!
You finished How Does AI Learn? — Teaching a Machine! Here is your certificate and digital badge to keep.
Save your badge and share it! 🌟
Ready for the next unit? 🚀