AI and Machine Learning
What a language model does with your prompt, from tokens to the next word, and how models learn at all.
Tokenization
Models do not read words. A sentence is split into characters, then the most common pairs are merged into tokens.
Not started.Beginner8 minEmbeddings and Semantic Search
Search by meaning: text becomes an arrow, and the nearest arrow wins even when no words match.
Not started.Beginner9 minAttention
How a model knows what "it" means: every token looks at every other one, which is also why doubling the prompt quadruples the work.
Not started.Intermediate11 minHow an LLM Generates Text
One prompt in: tokens, vectors, layers, scores, one token out. Then the whole thing runs again for the next token.
Not started.Beginner10 minTemperature and Sampling
The same prompt gives a different answer. Temperature reshapes the odds before the next token is drawn.
Not started.Beginner8 minContext Windows and Memory
Models do not remember your chat. Each message is sent with the whole conversation so far, and the app is what remembers.
Not started.Beginner7 minRAG or Fine-Tuning
Does the model lack facts, or lack a skill? Facts change, so they are looked up. Behaviour is trained.
Not started.Intermediate9 minInside One Neuron
Multiply, add, squash. Change one weight and the same inputs give the opposite answer.
Not started.Beginner7 minGradient Descent
A model learns by rolling downhill. Three learning rates: one crawls, one bounces out of the valley, one settles.
Not started.Intermediate9 minEight AI Words, Explained
Token, context window, embedding, temperature, top-p, fine-tuning, RAG and hallucination, each shown happening.
Not started.Beginner8 min