The AI Literacy · Plate 01 · 35 min · unlocked
What it is actually doing
A thirty-five-minute lesson on what a language model does when it answers — predicting the next word from patterns — and on saying so without the word "thinking".
Outcome
Can explain prediction from context without using the word "thinking".
Every other plate in this course depends on this one. A child who believes the machine "knows" things will treat a confident wrong answer as a fact, treat the machine as a source, and hand it their homework without noticing the handover. A child who can say "it picked the word that usually comes next" has a handle on all three. The lesson is short because the idea is small; the work is in getting the words right.
Before you start
One device with a chat model, at the front of the room. Nothing for children to sign into. You will also want the board and a way for pairs to write two sentences.
You do not need to know how these systems are built. The lesson's account — it was shown a great deal of writing, it learned which words tend to follow which, it produces the most likely next word over and over — is accurate at the level a child needs, and it is the level most adults need too.
Starter, five minutes
Say "Once upon a…" and stop. The room says "time". Say "Peanut butter and…" — "jelly". Say "Twinkle, twinkle, little…" — "star".
Ask: how did you know? Nobody looked it up and nobody read your mind. Land the sentence before the technology appears:
You have heard those words together so many times that the next one is the most likely one.
That is the whole mechanism. Everything else in the lesson is the same trick at a larger scale.
Main task, twenty minutes
Part one, ten minutes. At the front, type a plain question into the model — "Why is the sky blue?" works — and read the answer aloud. Then ask the class to describe what just happened without using the words think, know, understand, or believe. Write the attempts on the board.
The destination is some form of "it wrote the words that usually come after a question like that." Accept any wording that has the machine producing likely words. Press gently on any wording that has it looking something up, or having an opinion, or being a person somewhere.
Part two, ten minutes. Pairs write two sentences about the answer on the board.
The first: one true thing the machine did. ("It gave us four sentences about sunlight.")
The second: one thing it did not do. ("It did not look at the sky." "It has never seen blue.") The second sentence is the harder and better one, and it is where the outcome lives.
Plenary, ten minutes
Read three second-sentences aloud and ask the room which is the most surprising. Then read one line from the course's own reading for this plate — grade 4's version puts it plainly:
It does not look things up. It does not know what a peanut is. It has never tasted jelly.
Ask: if it has never tasted jelly, how did it know "jelly" comes after "peanut butter and"? The answer is the starter. Close the loop and stop.
Teacher notes
The failure this lesson guards against is "the AI thinks". It is not a mistake of vocabulary; it is a mistake about what kind of thing is in the room, and it makes every later lesson harder. When a child says it, do not correct the word. Ask them to say the same thing without it. The struggle to rephrase is the learning.
The opposite failure — "it's just a calculator" — is rarer and also wrong, because it can write sentences nobody has ever written. The accurate middle is "it composes from patterns". Composing is real; understanding is not present. Both halves matter.
Children will ask whether it will get smarter. The honest answer is that it will get better at predicting, and that this lesson's description will still be true.
Assessment
Two sentences in the pupil's own words: what the machine did to produce an answer, and one thing it did not do — with neither sentence using think, know, understand or believe.
A pair of sentences that describe the answer's content rather than how it was produced has not met the outcome.
What comes next
Plate 2 asks why the machine is so confident when it is wrong. That question is only answerable by a child who has already stopped thinking of it as something that knows.
Teach it · Read It Together
What Comes Next?
A riddle about what the machine really does. The lesson above, as a booklet to read together: a teacher at the front with the class, or a parent at the table with one child. 35 minutes, in two voices. In the app, Archie reads it aloud.
Download the deck (PowerPoint)Print it (PDF) · 10 pages · grades 3-5
A read-aloud for kids, parents and teachers
What Comes Next?
A tells B, and B tells C, "Can a machine think like you and me?"
Sit side by side and turn the page. We will solve it together, whatever your age.
Thirty-five minutes. One grown-up, one or more children, and a question to answer.
Step 1 · The starter game · 5 min
Once upon a…
Say "Once upon a…" and stop right there. The room shouts "TIME!" without a care.
Say "Peanut butter and…" and wait. "JELLY!" comes back, and it comes back straight.
Say "Twinkle, twinkle, little…" up high. "STAR!" flies out. Nobody asks why.
Do this together
Take turns. One of you starts a pair, "Salt and…", "Thunder and…", "Bow and…", and the other finishes it as fast as they can. Then ask each other, "How did you know?"
How did you know?
Nobody read your mind
Did you open a book? Did you look it up? Did you peek at a secret spy-glass cup?
No. You have heard those words together so many times that the next one was the most likely one.
That is the whole trick. When words travel together, your brain gets ready for the next one.
The machine plays the same game, with every word, at a size no person could.
Inside the machine
A mountain of writing
The people who built it showed it a mountain of writing. Stories, letters, lists, chats, guides.
From all of it, it worked out which words usually come after which.
When you type a question, it picks the most likely next word. Then the next. Then the next. Until it has a whole answer.
No brain in a jar. No heart. Just the word that usually comes next, over and over.
Watch your words
Four words that are easy to say
Think. Know. Understand. Believe. They slip out so easily, and none of them is quite true of the machine.
When one slips out, nobody is in trouble. Just say the sentence again without the word. The second try is where the learning is.
- Someone might say: The computer thinks it is raining.
- Say instead: The machine put together the words that usually follow a question about rain.
- Someone might say: It knows all about dinosaurs.
- Say instead: It learned the patterns in dinosaur writing.
Step 2 · The live test · 10 min
Why is the sky blue?
Type the question at the front, or at the kitchen table. "Why is the sky blue?"
Out pours an answer, clean and new. Four tidy sentences about sunlight and air.
Now the real challenge. Say what just happened without the words think, know, understand or believe.
Did it open a window? Did it look at the sky? No. It wrote the words that usually follow a question like that.
Step 3 · Two true sentences · 10 min
One thing it did. One thing it did not.
Grab a partner. Parent and child, or shoulder to shoulder with a classmate.
Sentence one: one true thing it did. "The machine wrote four sentences about sunlight."
Sentence two: one thing it did not do. "The machine did not look out of the window. It has never seen blue."
The second sentence is the harder one, and the better one. That is where the understanding lives.
Do this together
Neither sentence may use think, know, understand or believe. If you find yourself describing the facts in the answer instead of how the answer was made, try again.
The climax · 10 min
It never tasted jelly
It does not look things up. It does not know what a peanut is. It has never tasted jelly.
So if it never ate toast in the morning sun, how did it finish "peanut butter and…"?
Because in all the books and chats and lists, "jelly" is the word that usually comes next.
The loop is closed. The secret is out. Prediction is how the whole job is done.
Next time · Plate 2
"George Washington flew to France!"
Said with total confidence. No pause. No doubt. And completely wrong.
Why is the machine so sure when it is wrong? You can only answer that once you have stopped thinking of it as something that knows.
Plate 2: why being sure and being right are two different things.
For the grown-up
Why this plate matters most
A child who believes the machine "knows" things will treat a confident wrong answer as a fact, treat the machine as a source, and hand over their homework without noticing.
A child who can say "it picked the word that usually comes next" has a handle on all three.
When a child says "the AI thinks", do not correct the word. Ask them to say the same thing without it. The struggle to rephrase is the learning.
It is not a calculator either. It can write sentences nobody has written before. The honest middle is "it composes from patterns". Both halves matter.
Dinner-table game: start a common phrase and see who shouts the next word first. Then say it: that is what the machine does for every word it writes.
Download the deck (PowerPoint)Print it (PDF) · 9 pages · grades 6-8
Read it together · for students, parents and teachers
What Comes Next?
The machine on the other side of the chat box is not reading your question and thinking about it.
In thirty-five minutes you will be able to say exactly what it is doing instead, in one sentence, without the word "thinking".
Step 1 · The starter · 5 min
Finish the phrase
"Once upon a…" "Peanut butter and…" "Twinkle, twinkle, little…"
Everyone finishes them, instantly, and nobody looked anything up.
You have heard those words together so many times that the next one is the most likely one.
That is the entire mechanism. Everything else in this lesson is the same trick at a larger scale.
Do this together
Do three more in pairs. Then write down, in one sentence, how you knew.
How it was built
Prediction, not understanding
It was shown a very large amount of writing. From it, it learned which words tend to follow which.
Given your question, it produces the most likely next word. Then the next. One at a time, until it has an answer.
It does not look things up. It has no view of the world. It composes from patterns.
That is not nothing. It can write sentences nobody has written before. But composing is not understanding.
The words that mislead
Think, know, understand, believe
Each of these describes a person. Used about the machine, each one quietly assumes there is someone in there.
The habit to build is saying the same thing without them. Not because the words are forbidden, but because the sentence without them is the true one.
- Someone might say: It knows the capital of France.
- Say instead: "Paris" is the word that most often follows that question.
- Someone might say: It understood what I meant.
- Say instead: My wording matched patterns it had seen before.
- Someone might say: It searched the internet.
- Say instead: It assembled the most likely words, one after another.
Step 2 · The live test · 10 min
Why is the sky blue?
Type a plain question. Read the answer aloud.
Then describe what just happened without think, know, understand or believe. Write the attempts down.
The destination is some form of "it wrote the words that usually follow a question like that".
Any wording that has it looking something up, holding an opinion, or being a person somewhere is the thing to press on.
Step 3 · Two sentences · 10 min
What it did, and what it did not
In pairs, write two sentences about the answer on the screen.
One true thing it did: "It produced four sentences about light scattering."
One thing it did not do: "It did not observe the sky. It has no sensors and no experience of blue."
The second sentence is the assessment. A pair that describes the content of the answer rather than how it was produced has not got there yet.
Plenary · 10 min
It has never tasted jelly
Read the hardest second sentences aloud. Which is the most surprising?
Then close the loop. It does not know what a peanut is. It has never tasted jelly. So how did it finish "peanut butter and"?
The answer is the starter. "Jelly" is the word that usually comes next.
Next · Plate 2
Why it is so confident when it is wrong
A machine that predicts likely words can assemble a flawless-sounding mistake with complete certainty.
That only makes sense once you have stopped thinking of it as something that knows. Which is why this plate came first.
For the adult in the room
Notes
The failure this lesson guards against is "the AI thinks". It is not a vocabulary mistake. It is a mistake about what kind of thing is in the room, and it makes every later lesson harder.
When a student says it, do not correct the word. Ask for the same sentence without it.
The opposite failure, "it is just a calculator", is also wrong. The accurate middle is that it composes from patterns.
They will ask whether it will get smarter. It will get better at predicting, and this description will still be true.