0
Turbot

The Big Sort at the Floating Market

Marina is swamped: all kinds of stuff is floating around the docks. Meet Turbot, a brand-new little AI. It sees nothing, it knows nothing.

Your teacher mission: measure each object with your instruments, give the right label, and build Turbot's brain with YOUR measurements. Keep a parent close by for the duo challenge!

Pssst… Marina swears a CLUE for the Island Mystery is hidden at the end of this quest.

Teacher school — 1 minute to understand Turbot

1Turbot does NOT see pictures. It only gets 3 measurements — and YOU are the sensor taking them. If you measure wrong, it learns wrong.
2Its brain is EMPTY. It keeps your examples in its memory: your measurements + your label. You are not just playing — you are building its brain.
3With a new object, it searches its memory for the example with the closest measurements… and copies your label. It does not think — it compares numbers.

Teacher question: if you measure everything any which way today, what happens to Turbot tomorrow?

Measure the object with your instruments, then give the label.

Your measuring instruments

Measure first — this is what Turbot will get

Inside Turbot's memory

It does not know what a fish is. It keeps your examples and looks for the one that looks most like what it is shown.

Its memory is still empty.
Thinking ticket

Turbot put the fishing lure with the fish. Why do you think?

The island notebook

Here is what REALLY just happened:

1. The data

YOUR measurements + YOUR labels — you were the sensor

2. The model

Turbot keeps your examples in its memory — that is ALL it knows about the world

3. The prediction

New object? It finds the closest measurements and copies your label

Real AIs do the same thing — with MILLIONS of examples, measured by millions of sensors.

Turbot's secret map

Inside Turbot's head, every object is a dot placed by YOUR measurements. The more two objects look alike, the closer they are.

your fishyour junk★ the trickster

It's everywhere in your life

Turbot is tiny… but big AIs do EXACTLY the same thing, with millions of examples. Tap each card to check your answer.

YouTube keeps suggesting Minecraft videos. How does it guess?

think first, then tap to check

Just like Turbot: it compares with the videos YOU already watched. Your past clicks are its training set — you are its teacher without knowing it.

Your parents' email sends some messages straight to “junk mail”. How does it know?

think first, then tap to check

It has seen millions of examples of spam. A new message whose “measurements” look like THEIRS: trash. It's your big sort, email edition.

The phone unlocks when it sees your parent's face… but not yours. Why?

think first, then tap to check

Because it was trained with measurements of THEIR face, not yours. A machine only knows what it was shown — like Turbot with the lure.

The Duo Moment

Go get a parent (or a big kid)! The challenge: you know Turbot's brain by heart. Your parent does not.

For 3 objects, your parent guesses what Turbot WILL say. You keep score — you are the referee and the teacher.

Parent: 0/3

What will Turbot SAY?

Your turn, teacher!

Explain these two ideas to your parent, in YOUR OWN words. Tap the card to see if you said it like Turbot.

1. How did Turbot learn?

tap to see Turbot's version

“It didn't learn from a book — it kept my examples in its memory: MY measurements with MY labels. I was its sensor AND its teacher.”

2. Why did the trickster fool it?

tap to see Turbot's version

“Its measurements landed right in the fish corner on its map. An AI doesn't see the object — it only sees the measurements. That's why you always have to double-check its answers!”

Master Sensor

Quest complete!

Back to the island

CLUE #1 — THE ISLAND MYSTERY

At the bottom of the fish tank, carved on a stone, the first password of the Summit Machine:

EXEMPLES

“A machine doesn't invent anything: it learns from its EXEMPLES.” EXEMPLES is French — it means EXAMPLES. Type it just like that, in French!
You'll need 4 passwords to wake up the Summit Machine. The next one is hiding in the matchbox machine…

Parent note — your child just: measured objects on three variables (a sensor's job — and the lesson “crooked data, crooked machine”) · trained a real algorithm (nearest neighbor) with THEIR OWN measurements · predicted the machine's answers before it spoke (the real proof of understanding) · read the model's internal map · discovered data bias. Quebec curriculum (PFEQ, grades 2-3) links: measuring and comparing (math) · observe-predict-test-explain (science) · oral communication · critical thinking. Everything happened on your device.

Marina's 5 measuring challenges

New objects, new instruments: each challenge has 3 new variables to measure. Train a mini-AI with 3 examples, then predict it on 2 surprise objects.