Summary, Glossary and What's Next

What you now know

• AI is the field; machine learning learns from data; deep learning uses large neural networks; generative AI creates content.
• Models learn by adjusting parameters to reduce error during training; you use them at inference.
• An LLM predicts the next token, over and over, using the Transformer and its attention mechanism.
• Assistants come from pretraining plus instruction tuning and human feedback.
• Their main limits: hallucinations, the knowledge cutoff, the context window, bias and privacy.

Glossary

Term	Meaning
Parameter (weight)	A number inside a model, set during training
Token	A piece of text a model reads or writes
Context window	The most tokens a model can consider at once
Inference	Using a trained model to produce output
Temperature	How much randomness goes into choosing each token
Hallucination	A plausible-sounding but false output
Knowledge cutoff	The date the model's training data ends
RAG	Giving the model relevant source text to answer from

Check yourself

Each chapter ends with a short quiz. Take them in order — Easy, Medium and Hard — and read the explanation for every answer you miss.

What's next

In the Level 0 path, the next topic is Python for AI: the language you will use to call these models, prepare data and build your first AI features.

Further reading

• Attention Is All You Need — the 2017 Transformer paper
• OpenAI: What are tokens and how to count them?
• OpenAI: Why language models hallucinate