How AI Works
16 lessons that take you from ‘what is AI?’ to how reasoning models think. About 2 hours in total.
Beginner16 lessons in 4 modules
0 of 16 lessons done
We'll save your progress in this browser.
Module A · Big picture
3 lessons- 1What is AI, really?AI vs ML vs deep learning vs generative AI, with everyday Indian examples6 min readUp next6 min readUp next
- 2From rules to reasoning: a short historyTuring to ChatGPT to agents10 min readNot started10 min readNot started
- 3Meet the LLMWhat a large language model is and what ‘large’ means6 min readNot started6 min readNot started
Module B · How an LLM talks
5 lessons- 4Predicting the next wordGeneration one token at a time, probabilities, temperature6 min readNot started6 min readNot started
- 5Tokens: the LLM's alphabetTokenization, token IDs, context windows, cost6 min readNot started6 min readNot started
- 6Search engine vs LLMKnowledge cutoffs, hallucinations, why confident isn't correct6 min readNot started6 min readNot started
- 7Embeddings: meaning as numbersVectors, similarity, semantic search6 min readNot started6 min readNot started
- 8RAG: giving the model an open bookRetrieval-augmented generation in plain words7 min readNot started7 min readNot started
Module C · Inside the model
4 lessons- 9Neural networks without the scary mathsNeurons, weights and layers, explained with pictures instead of equations6 min readNot started6 min readNot started
- 10The Transformer, step by stepAttention and the building blocks inside every modern LLM7 min readNot started7 min readNot started
- 11How models learnTraining data, loss and gradient descent in plain words7 min readNot started7 min readNot started
- 12From base model to assistantFine-tuning and human feedback: how a text predictor becomes a helpful chatbot6 min readNot started6 min readNot started
Module D · Today and next
4 lessons- 13Can AI reason?What reasoning models do differently, and where they still slip6 min readNot started6 min readNot started
- 14Beyond text: multimodal AI and agentsModels that see and hear, and models that take actions with tools7 min readNot started7 min readNot started
- 15Using AI responsiblyBias, privacy, copyright, and checking the output before you trust it6 min readNot started6 min readNot started
- 16Capstone: the journey of one promptFollow one prompt from your keyboard to the reply, using everything so far6 min readNot started6 min readNot started