AI Foundations — Index
AI Foundations
| # | Lesson | Question answered |
|---|---|---|
| 1 | AI, ML, deep learning, and generative AI | What belongs inside “AI”? |
| 2 | Data, training, validation, and inference | How does a model learn and later produce output? |
| 3 | Neural-network basics | What do weights, layers, loss, and gradients do? |
| 4 | Transformers and attention | Why did transformers change language modeling? |
| 5 | Tokens, embeddings, and context | What does the model actually receive? |
| 6 | Generation, decoding, and hallucinations | Why can plausible text be wrong? |
| 7 | Model families and multimodality | How do text, vision, audio, and specialist models differ? |
| 8 | AI as a glass box | Can I reconstruct and test the observable path through an AI system? |
| 9 | AI as a white box | Can I trace, train, and inspect a tiny language model? |
Exit check: reconstruct the observable path of one AI run, trace one token from text to next-token probability, explain how a training error changes parameters, and distinguish a model from an application and a harness without using the word “AI” as the definition.