AI lab

AI isn't magic: it's a program that finds patterns in examples. In the four experiments here, you'll train small but real models yourself. The lessons explain why they work the way they do.

Open the first experiment: Draw, teach, predict

Experiments: 0/7
Lessons: 0/9

Four experiments, one cycle

Every AI follows the same path: data is collected, the model learns from it, it makes predictions about new things, and people use it. Each experiment has you do one step of that path by hand.

1 · Data10 minDraw, teach, predictTrain your own model by drawing shapes. Then make a new drawing and see what the model thinks it looks like.DataLabelsSimilarityLearning from mistakes
2 · Learning12 minTune the neuronSet the weights of an artificial neuron that sorts fruit into ripe and unripe yourself, then watch it learn on its own.WeightThresholdTraining loopLayer
3 · Prediction8 minThe next wordBuild sentences with a tiny language model trained on short stories. Chatbots pick their words the same way.Language modelProbabilityTemperatureContext
4 · Learning12 minTeach with sentencesTrain a model with example sentences: it learns whether a sentence is happy or sad, and runs the lamp and fan in a smart room.Text classificationSentiment analysisTest dataBias
5 · Learning12 minDecision treeTrain a decision tree on days that show how a friend gets to school. Let the tree find the rule, then turn it into Python code.Decision treeFeatureIf / elseToo simple
6 · Prediction12 minThrow and predictCollect the powers and distances of the robot’s throws and train a line. Then let the model predict the power it takes to land in the basket.RegressionErrorLine of best fitOutside the data
7 · Use8 minChatbot detectiveLook closely at a chatbot’s answers: which can you trust, which need checking, and which should you never answer?HallucinationPrivacyChecking sources

Lessons

Each lesson starts with a question, opens up with an example and ends with a short quiz. At the end you can jump to the related experiment.