From Sand to Agents — K-12 Artificial Intelligence

Some machines just do one thing the same way every time, like a light switch. Others notice the world with sensors and choose what to do — those are the smart machines, and they're all around us.

A pattern is something that repeats, and once you spot one you can guess what comes next. That everyday guessing is the simplest version of how a smart machine makes a prediction.

Machines learn from the examples we show them, a bit like teaching a friend a new game. Show good, fair examples and it guesses well; show messy or unfair ones and it guesses wrong.

Smart machines should be fair, keep private things private, and be honest. And people — not the machines — are the ones who decide what's right and set those rules.

Everything a computer does is built from millions of tiny switches that are either on or off. Stack enough of those simple signals together and you can build anything a computer knows how to do.

An algorithm is a set of clear, in-order steps — like a recipe — that a computer follows exactly. Loops repeat steps, conditionals choose between them, and debugging is how you fix the steps that go wrong.

Computers sort and search through data to find patterns, and those patterns let them make smart guesses about what you might want or what comes next.

A computer's guesses are only as good as the data behind them. If the data is one-sided or unfair, the results will be too — so where data comes from really matters.

Computer chips are built from sand-based silicon shaped into billions of microscopic switches called transistors. Wired into logic gates, those switches are the hardware that does every calculation.

Machine learning means a program learns patterns from many examples instead of being handed every rule. It builds a model from training data, then uses it to handle things it has never seen before.

A neural network is a stack of tiny math machines called neurons: each multiplies its inputs by weights, adds them up, and fires if the total crosses a threshold. Wire many together in layers and the simple pieces add up to something that learns.

A model can look confident and still be wrong. This unit is about measuring whether a model is actually good — and whether it's fair — instead of just trusting it.

Powerful systems run on data, so protecting that data and using these tools responsibly is part of the skill. Good digital citizens guard privacy and think about impact before they act.

How AI changes industries and jobs — and the real routes into them.

Computers turn words, images, and sounds into numbers so math can work on meaning. These number-representations, called embeddings, are the foundation modern AI is built on.

Deep networks learn by making a guess, measuring how wrong it is, and nudging themselves to be a little less wrong — millions of times over. The real trick is learning the pattern without just memorizing the examples.

Transformers are the design behind today's most powerful AI. Their key idea — attention — lets a model weigh which parts of the input matter most for what it's trying to figure out.

Large language models learn from enormous amounts of text and generate language one piece — one token — at a time. From all that scale, surprising new abilities emerge.

How you ask shapes what you get back. Prompting and context are the skills of giving a model clear instructions and the right background so its answers are reliable.

Running AI at scale costs real money and time. Engineers use techniques like caching to make systems fast and affordable enough for millions of people to use at once.

Agents are models that don't just answer — they act. They use tools, make plans, and work in loops to get real tasks done. These are the 'Agents' the whole curriculum builds toward.

Models can sound sure while being wrong. Retrieval connects them to real, current information so their answers are grounded in facts they can point to.

When one system serves millions of people, small biases cause large harm. This unit weighs fairness, accountability, and who is responsible when AI gets it wrong.

When AI helps make something, who owns it and who gets the credit? This unit looks at intellectual property, attribution, and honest use of AI-assisted work.

The final unit steps back to the big picture: keeping AI safe and aligned with human goals, its effect on society, and the new kinds of jobs and careers it is creating.