AI Dictionary
LEARNING PATH

Learn AI from zero — in the right order.

The terms are laid out in a reading order — each section builds on the previous. Start anywhere; reading time is shown next to each section.

39 terms~74 min total
Step 1

Foundations

3 terms · ~6 min total

The big picture first: what is AI, how does ML fit, why were neural networks a breakthrough?

  1. 1Artificial IntelligenceAI · ~2 min
  2. 2Machine LearningML · ~2 min
  3. 3Neural NetworkLayered learning model · ~2 min
Step 2

Language Models

4 terms · ~8 min total

The core of modern AI: LLMs, their architecture (Transformer), and the atomic unit (token).

  1. 1LLMLarge Language Model · ~2 min
  2. 2TransformerThe architecture behind LLMs · ~2 min
  3. 3TokenThe atomic unit of text for an LLM · ~2 min
  4. 4Context WindowHow much an LLM can see at once · ~2 min
Step 3

Talking to LLMs

5 terms · ~10 min total

The craft of steering a model: prompting, system messages, temperature, and learning by example.

  1. 1PromptThe instruction you give an LLM · ~2 min
  2. 2System PromptThe persistent instruction · ~2 min
  3. 3TemperatureThe randomness dial · ~2 min
  4. 4Few-shot LearningTeaching by example · ~2 min
  5. 5Chain-of-ThoughtCoT — Step-by-Step Reasoning · ~2 min
Step 4

Limits & Risks

3 terms · ~6 min total

The model's failure modes, security gaps, and how much text it can handle — production limits you must know.

  1. 1HallucinationConfident wrong answers · ~2 min
  2. 2JailbreakBypassing safety guardrails · ~2 min
  3. 3Prompt InjectionHidden commands in user data · ~2 min
Step 5

Vectors & Meaning

3 terms · ~3 min total

The infrastructure of RAG: meaning as numbers, vector databases.

  1. 1VectorNumeric representation · ~1 min
  2. 2EmbeddingMeaning, encoded as numbers · ~1 min
  3. 3Vector DatabaseSimilarity search store · ~1 min
Step 6

Custom Knowledge with RAG

5 terms · ~9 min total

Giving a model external knowledge without retraining — RAG's full stack.

  1. 1RAGRetrieval-Augmented Generation · ~2 min
  2. 2ChunkingSplitting documents for retrieval · ~2 min
  3. 3Hybrid SearchKeyword + semantic, combined · ~1 min
  4. 4BM25Best Matching 25 — classic relevance score · ~2 min
  5. 5RerankerThe second-pass ranker · ~2 min
Step 7

Agents & Tools

3 terms · ~6 min total

Going from chat to action: tool use, agents, the MCP standard.

  1. 1Function CallingTool Use · ~2 min
  2. 2AI AgentAn LLM that takes actions · ~2 min
  3. 3MCPModel Context Protocol · ~2 min
Step 8

Training & Optimization

6 terms · ~12 min total

How models are trained, specialized, and made faster.

  1. 1Fine-tuningSpecializing a pretrained model · ~2 min
  2. 2RLHFReinforcement Learning from Human Feedback · ~2 min
  3. 3AlignmentTuning models to human values · ~2 min
  4. 4LoRALow-Rank Adaptation · ~2 min
  5. 5QuantizationModel compression · ~2 min
  6. 6Knowledge DistillationTeacher → Student transfer · ~2 min
Step 9

Advanced Topics

7 terms · ~14 min total

Frontier terms — reasoning models, MoE, multimodal, generative architectures.

  1. 1Reasoning ModelModels that think first · ~2 min
  2. 2Mixture of ExpertsMoE · ~2 min
  3. 3MultimodalMany modes, one model · ~2 min
  4. 4Diffusion ModelGeneration by gradual denoising · ~2 min
  5. 5Image GenerationText-to-image · ~2 min
  6. 6TTSText-to-Speech · ~2 min
  7. 7ASRAutomatic Speech Recognition · ~2 min