AI isn't magic. It reads words by chopping them into pieces, it "sees" pictures as grids of numbers, and it learns in three ways — with a teacher (supervised), all by itself (unsupervised), or by trial & reward (reinforcement). Take each idea apart in its own little playground — then see how a real LLM learns and answers.
Chop a sentence into pieces and give each a number.
Play → 2🧹Stop-words, stemming & lemmatization — words to their roots.
Play → 3🔢Turn words into numbers so similar words sit close together.
Play → 4🔮The probability game every language model plays.
Play → ★🤖The big one. Watch the model learn from text, then write one token at a time.
Open the big lab →Learn from labelled examples, draw a boundary, then sort new things. (Needs a teacher.)
Play → 7🪄No labels at all — watch the computer find the groups by itself.
Play → 8🕹️A robot learns a path to the 🏆 by trial & reward — no labels, no teacher. Train it and watch it get smarter. (The "RL" behind game-playing AIs and chatbots.)
Play →Unzip and double-click index.html — runs with no internet.