How AI Really Works
An intuitive model of LLMs, no code.
What you will learn
- Describe an LLM as autocomplete at a huge scale
- Explain what a token is
- Describe sampling as a weighted pick
- Put pretraining, fine-tuning and human feedback in order
- Explain an embedding in one sentence
- Say why an AI invents facts
- Describe a tool call in plain words
- Walk through the six steps of an answer
Lessons
Inside the Chat
- Your phone already predicts
- Autocomplete, scaled up
- Watch a reply get built
- Run the loop by hand
- One step, many options
- Fluent is not the same as checked
IdeaSample card Autocomplete, scaled up
Autocomplete guesses your next word from a short phrase. An LLM guesses the next piece from pages of context.
It learned from huge amounts of writing. Your question starts a text it must continue.
Enrol free to open every lesson. Your progress is saved.
Lesson 2: Tokens: How AI Reads Text
Lesson 3: Context Window: Short-Term Memory
Lesson 4: Temperature: Same Prompt, Different Answers
How It Learns
Lesson 5: Training in Plain Words
Lesson 6: Embeddings: Meaning as a Map
Lesson 7: Why It Makes Things Up
Using It Wisely
Lesson 8: Big vs Small, Open vs Closed
Lesson 9: Tools and Search: How Chat Apps Do More
Lesson 10: Explain It Back
What you need
- A phone with a browser. A computer works too.
- A free Google account, to save your progress.
- No tech background. Every new word is explained.
AI apps you can use
Every task works with the free plan of any of these apps.
- Gemini
- ChatGPT
- Claude
Start this module
Course: AI Beginner's Course