AI 101

What can AI do today?

It's already part of your life. Even in places you haven't noticed.

Four nested rings.

AI is an umbrella concept. Each ring is a subset of the one before it; the further inward you go, the more specialised it becomes. Click the rings to explore.

So how do experts define artificial intelligence?

Surprisingly, there is no single clear definition. "Artificial intelligence" is a suitcase word: everyone packs something different inside. And once a technology starts working, we stop calling it "AI", spell-checkers were once considered AI, now they're mundane. That's why experts look for two characteristics instead of a single definition.

The more a system possesses both of these properties, the more inclined we are to call it "artificial intelligence". Neither is an on/off switch, both are a spectrum. You'll test this spectrum yourself just below.

Is this AI?

Six everyday scenarios. For each one, make your call: clearly AI, debatable, or not at all? Keep the autonomy and adaptability criteria in mind. Choose your answer, then see the reasoning.

The AI map.

Let's put the six scenarios onto a single map. The horizontal axis shows autonomy, the vertical axis shows adaptability. The further toward the top right, the more inclined we are to call a system "AI". Tap a point to see where it stands.

Notice: the points don't cluster in the corners, they're scattered everywhere. "AI" is not a sharp boundary; it's a region that slides across two axes. A thermostat and an autonomous vehicle are on the same map, but in completely different places.

How does deep learning work inside?

Deep learning gets its "depth" from its layers. Let's run a small neural network that recognises a handwritten digit together. Use the "advance signal" button to move through layer by layer; hover over nodes to inspect their connections.

In real networks there are millions of connections across these layers, in large language models, billions. But the principle is the same: each layer learns a slightly more abstract feature from the output of the previous one. Nobody tells the network "what is an edge"; it figures it out from examples. That's precisely what the "learning" in machine learning means.

What does a large language model actually do?

The core task of models like ChatGPT and Claude looks surprisingly simple: predict the most likely next piece. Complete the sentence below together with the model.

Which sentence needs verifying?

Because the model speaks by probability, it fails most often exactly where it looks most certain: figures, names, dates and sources. Below is a realistic output a teacher received. Click the statements you would have to verify before using it.

Keep this as a rule: every figure, name, date and source the model produces gets verified; conceptual and pedagogical statements pass through your professional judgement instead. We come back to this reflex in detail in Module 4.

What's the situation in education?

Promising or concerning? Both at once. Usage is spreading fast but training isn't keeping up.

The picture says this: students are already using it, while nearly half of teachers have received no training on the subject. That gap is the reason this course exists. One more comparison: in 2026 the AI in education market is valued at $8 billion.

Myth or fact?

These sparked the liveliest debates in our workshops. Give your own answer first, then flip the card.

AI takes over tasks, not your job.

A common fear: "Will AI replace teachers?" A more precise question: which tasks does it take over? A large portion of a teacher's week is spent on lesson planning, marking, reports and rubrics, work not directly facing students. Switch tabs to see how time shifts.

This isn't about job loss, it's about tasks shifting. With a global teacher shortage, demand isn't falling; as repetitive work is automated, what remains for the teacher is their core job: human connection. Another angle: the same tool gives every student with a device and internet access a low-cost personal tutor; the student without access is left further behind. AI can both close and deepen inequality.

So what should students learn?

The surprising answer: largely the same core skills. Writing well, thinking critically, solving problems, a solid knowledge base, and both using digital tools and understanding how they work. One more skill is added on top: managing AI.

Five questions before you use it.

A framework distilled from workshop participants' own questions. Run through these five questions quickly before any AI use.

Fill in the blanks.