AIQ
AIQ
Part of the How Computers Think lesson guide. Teaching a different grade? 🌈 Explorer (5–7) · 🔧 Builder (8–10) · ⚡ Architect (15–18)
Open with the comparison cold, no build-up:
"Your smartphone has more computing power than the computers that sent Apollo 11 to the moon. But it still can't understand a joke. Why? Let's dig into how computation actually works — and where it hits its limits."
Let that sit for a second, then ask: "So if it's not a lack of power, what's actually missing?" Don't answer yet — write it on the board as the question the lesson is going to answer: power vs. understanding. Tell the class that by the end of the period they'll be able to explain, in real technical terms, exactly what a computer is doing at the hardware level and why that's a fundamentally different thing from "understanding."
This age band tends to respond well to being told plainly that the lesson is going somewhere more technical than earlier grades get — say so directly: "Younger students in this app get the light-switch version of this idea. You're getting the real explanation, with real vocabulary, because you can handle it." That framing tends to raise engagement rather than intimidate, especially with students who already think of themselves as "into computers."
Nova (the Hacker-mode mascot) walks the class through the app's three scenes; use each as a launch point for a more technical explanation than the younger bands get. This age band can handle real vocabulary — use it precisely rather than softening it.
Establish the base layer: a transistor is a tiny electrical switch with two states, on and off, which software treats as 1 and 0. On their own, single transistors don't do much — the interesting part is that they can be wired together into logic gates: AND (true only if both inputs are on), OR (true if either input is on), and NOT (flips on to off and back). Combine enough logic gates and you can build circuits that add numbers, compare values, and make decisions — which is exactly what a processor is: billions of logic gates, built from billions of transistors, doing exactly that at enormous speed. A modern smartphone chip contains on the order of ten to twenty billion transistors. Ask the class to sit with that number for a second — that's more transistors in one phone than there are humans on Earth, more than twice over. Worth being explicit that this is a manufacturing achievement, not a step toward understanding: each of those transistors is doing exactly what a single transistor has always done, switching on and off; nothing about combining more of them changes what a transistor fundamentally is, only how much can be built out of them.
Reframe the "billion calculations a second vs. ten a second" comparison in terms of what each system is actually optimized for. A CPU executes precise, unambiguous instructions at blistering speed because the problem is well-defined — add these two numbers, compare these two values. A human brain recognizing a face in a tenth of a second is solving a much messier problem — matching a fuzzy, ever-changing pattern under different lighting, angles, and expressions — which is why it took AI researchers decades of dedicated research to get computers to do it reliably at all, using a completely different approach (learning from huge numbers of example photos, covered in a later world) rather than explicit step-by-step instructions. This is a genuinely important asymmetry to name directly: tasks that feel effortless to a human (recognizing a face, understanding a sentence, catching a ball) are often the hardest to make a computer do, while tasks that feel hard to a human (multiplying eight-digit numbers, searching a billion records) are often the easiest. Researchers sometimes call this Moravec's paradox — the skills that took evolution hundreds of millions of years to build into brains turn out to be the ones that are hardest to replicate in silicon, while recent, uniquely human inventions like formal arithmetic turn out to be comparatively easy.
Connect the everyday "Input → Process → Output" idea to what's actually happening on the chip: a CPU fetches an instruction, decodes what it means, executes it on some data, and repeats — billions of times a second. Emphasize the key line from the lesson: computation, no matter how fast or how many instructions per second, is symbol manipulation without semantic understanding. The chip is moving and combining patterns of bits according to fixed rules; nothing in that process requires or produces awareness of what those bits mean. That's the real answer to the hook question.
A useful check-for-understanding question here: "If a program has a bug — say, it adds two numbers wrong — does that mean it 'made a mistake' the way a person would?" Push students toward the more precise answer: no, it followed its instructions exactly; the instructions themselves (written by a person) were wrong. The computer did not misunderstand anything, because there was never any understanding involved in the first place — only exact execution of whatever it was told, correct or not.
Run the lesson's own sorting exercise as a rapid-fire round, but push students to justify each call in one sentence instead of just voting. Examples straight from the lesson: "storing a photo as a grid of pixel values" (computation — it's numbers all the way down); "recognizing your friend's face across a crowded room" (cognition — pattern recognition your brain does in a fraction of a second, without explicit rules); "running billions of instructions per second" (computation); "actually feeling nervous before an exam" (cognition — no chip has ever been shown to have subjective experience). If a student argues that an AI system "feels" something because it produces text that sounds emotional, treat that as a genuinely good challenge rather than shutting it down — it's exactly the kind of claim later AIQ lessons on generative AI will ask them to evaluate critically.
Where students tend to get stuck: this age band is old enough to want a clean, single dividing line between "just computation" and "real intelligence," and can get frustrated that the lesson doesn't hand them one. Be upfront that this is honest, not evasive — the boundary is genuinely an open question researchers argue about, and the goal of the lesson is to give them the vocabulary (binary, logic gates, symbol manipulation) to reason about it carefully, not to hand them a settled verdict.
These four questions check whether the technical vocabulary from the activity actually landed, not just whether students can recall the lesson's overall message. If a student gets the concept right but picks the wrong option because of unfamiliar terminology, treat that as a vocabulary gap worth revisiting rather than a conceptual failure — the explanations below give you the exact distinction to re-draw.
Close with: "So the answer to today's opening question — why your phone still can't get a joke despite outclassing Apollo-era computers — is that jokes require understanding meaning, context, and intention. Computation gives you speed and precision at manipulating symbols. Those are not the same skill, and no amount of raw power turns one into the other on its own. Keep that distinction in your pocket — you'll need it again the moment we start talking about AI that seems to 'understand' language."
Extension activity — "Build a logic circuit on paper": In pairs, have students design a simple circuit using only AND, OR, and NOT gates (drawn as boxes with inputs and outputs) that lights an "output" if two conditions are BOTH true, or that flips a signal to its opposite. Have them trace through their own circuit with a partner using sample 1/0 inputs to prove it works. Then challenge them: "Using only these three gate types, what's the smallest circuit you can design to add two 1-bit numbers?" (This is a half-adder — the first building block of real CPU arithmetic, if anyone wants to look it up.)
If there's extra time: have students look up the transistor count of a chip they recognize (a recent iPhone or Android flagship processor, or a game console) and compare it to a chip from ten years ago. Ask them to connect the growth in that number to the "billions of logic gates" idea from Scene 1 — and to name one thing that number still hasn't bought us, using tonight's discussion as a starting point.