Foundations of Large Language Models: From N-grams to LLMs
A beginner-friendly video course on how language models work, from counting words to transformers and AI assistants. No machine learning background needed.
How does a model like ChatGPT or Claude actually work? This course answers that question from the ground up. You start with the simplest idea, predicting the next word by counting, and build step by step to neural networks, attention, and the transformer, then see how a raw model becomes a helpful assistant.
What you will learn
- How text is split into tokens and turned into numbers
- How n-gram models predict the next word, and why they fall short
- How a model learns weights from examples, and how embeddings give words meaning
- How neural networks learn with loss and gradient descent
- How attention and the transformer power modern LLMs
- How instruction tuning and human feedback create assistants, and where retrieval, agents and interpretability fit
Who it is for
College students and working professionals who want a clear mental model of LLMs. No machine learning background is needed, and basic high-school math is enough.
Lessons
10 video lessons · 31 minutes · the first 3 are free
Instruction Tuning and Human Feedback
Enroll to unlock this lesson and the rest of the course.
9. Instruction Tuning and Human Feedback
How a base model becomes an assistant, and the limits to remember.
- 1IntroductionFree2:01
- 2How This Course Is OrganizedFree1:18
- 3Why Text Must Be Turned Into TokensFree4:13
- 4How Machines Predict the Next Word🔒5:42
- 5From Counting Words to Learning Weights🔒3:17
- 6Words as Vectors: Embeddings🔒2:53
- 7How Neural Networks Learn🔒2:48
- 8Attention and the Transformer🔒3:03
- 9Instruction Tuning and Human Feedback🔒2:59
- 10Your Roadmap: Retrieval, Agents, and Interpretability🔒3:01