AI Concept Map
The Building Blocks of Modern AI
Click any concept to expand its explanation
Modern AI Systems
LAYER 01Primitives
Prompt
Prompt
Temperature
Context Window
inference parameters
LLM
Large Language Model
Transformer
MoE
model architectures
Embedding
Embedding
Vector Space
Cosine Similarity
semantic math foundations
LAYER 02Compositions
Function Calling
MCP
Vector Database
RAG
Guardrails
Multimodal
Prompt Caching
LAYER 03Model Adaptation
SFT
LoRA
RLHF / DPO
Knowledge Distillation
These adapt a pre-trained base model, not training from scratch. Pre-training costs tens of millions of dollars; only a handful of labs do it.
LAYER 04Deployment
Agent
Fine-tuning
Framework
Evals
Red Teaming
Small Model
LAYER 05Emerging
Multi-Agent Systems
Synthetic Data
Interpretability
Thinking Model
Test-Time Compute