What AI Actually Is (and What It Isn’t)
Before you can direct AI with confidence, you need to see what is under the hood. This module opens the machine, shows where bias gets in and how it gets repaired, introduces the people who build these tools, and previews the creation studio you will use in Session 2.
Complete this module before Session 2. The concepts live here, so the live time belongs to making. On Wednesday, July 15, you will train a real AI model and create real lesson media with your cohort. Arrive with the ideas already in place, and every live minute goes to building.
By the end of this module you will explain in plain language what AI is and is not, spot where bias enters and how it is repaired, know who builds the major AI tools and what each is good at, and arrive ready to create: in Session 2 we train a small AI model together and then use AI design tools to make an illustrated story you can use with your class.
What You Will Be Able to Do
Five outcomes, one per lesson. The knowledge check at the end verifies them.
Explain the Pattern Machine
Describe how AI turns training examples into predictions, and why confident output is not the same as correct output.
Explain Bias In, Bias Out
Show where bias enters an AI system, who it makes invisible, and why the repair is better data, not better math.
Name the Makers, Match the Tool
Identify the teams behind the major AI tools and sort tasks into generative and predictive work.
Say What California Expects
Summarize the state’s direction on AI literacy, guardrails, and the place of professional judgment.
Preview the Creation Studio
Know the tools we will use together in Session 2 and arrive with a story seed ready to grow.
The Pattern Machine
AI is not a mind. It is not magic. It is ones and zeros, trained on examples, guessing what comes next.
Training Data, Patterns, Prediction
Every AI tool you will ever use runs on the same three-step engine. Once you see it, the mystery goes away and the confidence comes in.
Here is the whole recipe. First, the system is fed training data: millions of examples of writing, images, or sounds. Second, it finds patterns in those examples: which words tend to follow which, which shapes tend to appear together. Third, when you ask for something, it makes a prediction: the most likely next word, the most likely pixel, based on everything it has seen.
Think of a student who has read ten thousand book reports but never opened a single book. Ask for a book report and they will hand you one that looks right: the opening line, the summary, the closing opinion, all in the familiar shape. They are not lying. They are pattern-matching. That is what AI does, at enormous scale and speed.
Notice what is missing from the recipe: understanding. The machine does not know what a book is. It knows what book reports look like. That difference is the key to everything else in this module.
Activity: Be the Machine
Four short writing samples. Sort each one into Pattern A or Pattern B using only what you can see on the surface: how it opens, how it is shaped, how it ends. Do not worry about what any of it means.
You just did what AI does.
Create the AI, Then Review It
You just did the machine’s job by hand. Now train the real thing. In this interactive experience you will build an AI model, watch it learn a bias nobody typed in, and repair it yourself.
Work through it right here, or open it full screen. It is narrated, so turn your sound on. It is self-paced, and everything you see in it comes back in Session 2.
Confident Is Not the Same as Correct
Sometimes the pattern machine produces something fluent, polished, and wrong. That is called a hallucination, and it is not a glitch. It is the recipe working as designed.
The machine’s job is to produce the most likely-looking answer, not the most true one. If your question reaches past its training, it does not stop and say “I do not know.” It keeps predicting, and the result reads just as smoothly as a real answer. A made-up citation looks like a real citation. A wrong date arrives inside a confident sentence.
This is why the context skills from Module 1 matter. The machine only knows patterns; you supply the truth about your students, your standards, and your voice. When you give AI real context and review every output, you close the gap between fluent and correct. You direct, it amplifies, and you own every final call.
Bias In, Bias Out
A pattern machine can only learn from what it is fed. Whoever chooses the data shapes what students see.
Scenario: Train the Library Model
You are training a reading-recommendation model for a school library. It will suggest the next book to every student who checks one out. The model itself is neutral math. The training data is your choice. Pick a data stack and see what happens.
Choose a data stack above. You can try all three.
What the Model Learned
The model now recommends one band of books to everyone. It suggests the same chapter series to the youngest reader and the oldest, because that is all it has ever seen succeed. Readers above and below that band get suggestions that never fit, and they slowly stop asking.
What the Model Learned
The model now mirrors one neighborhood’s tastes, languages, and shelves. Students from that community get recommendations that fit. Everyone else gets a menu built for someone else, and each suggestion nudges the whole collection further toward what the model already knows.
What the Model Learned
The model now draws from checkouts across grades, languages, genres, and interests. Its suggestions spread wide: a graphic novel here, a bilingual picture book there, a science title for the student who reads nothing but space. It still makes mistakes, and it still needs a human eye. But its starting point is fair.
The Repair
Notice what changed between the three models. Not the algorithm. The data.
The fix for a biased model is not better math. It is better data. All three library models ran identical code; the only difference was what they were fed. That means bias in AI is not a technical mystery reserved for engineers. It is a curation problem, and curation is something educators already understand deeply. Whoever chooses the data shapes what students see, which means the person choosing needs to be someone who sees all the students.
This is not hypothetical. AI systems trained on lopsided data have misjudged the very students who most need to be seen accurately: flagging the wrong kids, overlooking the right ones, and recommending resources that fit some learners while erasing others. When a model’s training data leaves students out, its decisions leave them out too, at scale and with a straight face.
In Session 2 you will train a real model yourself, break it with lopsided examples, and repair it with balanced ones. Reading about bias is one thing. Watching your own model fail and then fixing it with better data is what makes the lesson stick.
Meet the Makers
These tools are built by teams of people, each with strengths. Knowing who makes what helps you match the tool to the task.
Who Builds the Major AI Tools
Four teams you will hear about again and again. No rankings here, just what each is known for.
Google DeepMind
The team behind Gemini. Strong at working inside Docs, Gmail, and school Google accounts, which is why many districts start here.
OpenAI
The team behind ChatGPT and custom GPTs. Known for conversational drafting and a wide range of custom assistants built on its platform.
Anthropic
The team behind Claude. Known for careful, long-form work: reading long documents closely and writing with steady attention to nuance.
xAI
The team behind Grok. Known for fast, conversational answers connected to current online discussion.
Activity: Two Kinds of AI
Every AI task falls into one of two families. Generative AI creates new content. Predictive AI sorts what exists and forecasts what comes next. Drag each task to its family, then check your answers.
Drag each chip into the right column.
Skilled educators move between AI tools the way a carpenter moves between saws: the question is never “which brand do I love,” it is “which tool fits this cut.” Use your district’s approved tool for work, every time. Take your curiosity anywhere you like on your own time. Fluency across tools is the skill. Loyalty to one logo is not.
The World Is Moving (and California Is Ready)
Six numbers tell the story of where education and AI stand right now. Tap each card to reveal one.
Six Numbers, Six Signals
Tap each card to reveal the number behind it and what it means for your work.
What California Expects
The state has already set its direction. Here it is in plain language.
The California Direction, Unpacked
Three pieces make up the state’s current direction on AI in schools. Open each one.
AB 2876 · AI Literacy in the Curriculum
AB 2876 requires AI literacy to be woven into California’s curriculum frameworks. Students will learn how AI works, where its limits are, and how to question its output, not as a separate tech elective but inside the subjects they already study.
SB 1288 · A Working Group for Safe Use
SB 1288 created a working group that guides the safe use of AI in California schools. Its job is to study how AI is being used, identify risks, and shape guidance so schools do not each have to invent their own rules from scratch.
State Guidance · Judgment, Privacy, Equity
California’s guidance puts human judgment first: AI can inform decisions about students, but educators make them. It emphasizes privacy protection, keeping student information out of tools that were not built to hold it. And it names equity of access, so AI narrows opportunity gaps instead of widening them.
You are expected to build AI literacy, because the state is writing it into what students learn. You are protected by clear guardrails, because the working group and state guidance carry that weight so individual educators do not. Your professional judgment stays in charge, because every piece of state direction puts the human decision above the machine’s suggestion.
Your Creation Studio
In Session 2 we review Teachable Machine together, then create. These are the tools on the workbench.
The Tools We Will Use Together
You do not need to master any of these before Wednesday. Just know what each one does, so when we build together you recognize the workbench.
Google Teachable Machine
Train a real AI model in minutes, right in the browser. No code, no installs, works on Chromebooks. Your students can use it too, which makes it a bridge from your learning to theirs.
Gemini Image Creation
Describe a scene in plain language and get an illustration back. This is how the pictures in our story get made: you describe, it draws, you decide what stays.
Canva
Design polished pages and slides fast, with AI features built in. When the story needs to look like a book instead of a document, this is where that happens.
NotebookLM Audio
Turns your sources into a spoken conversation between AI voices. For listening learners, your written materials become something they can hear.
Google Vids & Slides
The assembly table. Images, narration, and text come together into something you can play for your class the same week.
teachablemachine.withgoogle.com · train a model in your browser
gemini.google.com · image creation and drafting
canva.com/education · design tools for educators
notebooklm.google.com · sources into audio conversation
We Will Make a Story Together
Session 2 ends with the cohort creating an illustrated, narrated mini-story you can actually use with your class. Not a demo you watch. A thing you make, page by page, with the tools above. Along the way you will also draft a real lesson plan with your AI, built around a topic you already teach.
Your part is simple: bring a story seed. A character, a setting, and one thing you want students to learn. A curious tide-pool crab and the water cycle. A new student and the first week of school. A grandmother’s garden and fractions. Small seed, real lesson. We grow it together on Wednesday.
Show What You Know, Then Bring Your Seed
Four questions, then a short prep list. Complete both and you are ready for July 15.
Get Ready for Session 2
Four small things before Wednesday, July 15 at 1:00 PM Pacific. With them in hand, your whole live session goes to making.
One Minute of Reflection
Think about your teaching. If AI could help you create one piece of learning media, illustrated, narrated, ready for your class, what would it be? Write it here for yourself.
See You Wednesday, July 15
1:00 to 2:30 PM Pacific · live on Zoom. Bring your story seed and your lesson topic. You will train a real AI model with your own hands, then make an illustrated, narrated story together with your cohort, ready for your class that week.
Review This Module Again