The whole series
Every week of Build Real AI
12 weekly sessions, each its own step. Built for ages 10 to 20. Every week builds on the last, so the learning is a real climb, not a set of one-off games.
What AI really is
A computer that finds a pattern from examples, instead of you writing the rules.
Teach a machine from examples
Show it labeled examples and it draws a boundary; more examples move the boundary.
Features: how a machine sees
It sees only the numbers you chose to measure. Representation is a choice you make.
Inside the neuron
Multiply each input by a weight, add them up, then a squish decides. That is a neuron.
Learning is adjusting
The machine reads which way lowers its wrongness and steps that way, again and again.
A network of neurons
Wire neurons in layers and the machine can draw curves one neuron never could.
The training loop
Guess, check the loss, adjust the weights, repeat. That loop is all of training.
Overfitting and honesty
It can ace the practice and still be bad. The honest score lives on held-out examples.
Predict the next word
A language model just predicts the next token, over and over. You trained one last week.
Bias and safety
A model inherits the skew of its data. Split the score by group and find the gap.
Build your own tiny AI (capstone)
Run the whole pipeline yourself: data, train, break it, audit it, and write the honest card.
Family Showcase Day
Each learner shows the AI they built and explains, honestly, how it learned.