Rules And Search
Symbolic AI And Search
Early AI treated intelligence as searching for solutions in spaces governed by explicit rules, laying foundations for planning, games, and pathfinding.
View Demo 01: SearchOverview
AI Evolution Timeline
This timeline is not a paper list. It places each era's core problem, technical shift, and matching demo on one learning path.
Knowledge Engineering
Expert Systems
Expert knowledge was written as if-then rules. The systems were explainable, but acquiring knowledge and maintaining exceptions became bottlenecks.
View Demo 02: Expert SystemsProbability And Statistics
Bayes And Statistical Learning
AI moved from deterministic rules toward uncertainty modeling, updating beliefs with evidence and gradually adopting data-driven learning.
View Demo 03: BayesDeep Learning
CNNs And Deep Vision
Convolutional networks combine visual features layer by layer, from edges and textures to shapes, using local receptive fields and shared parameters.
View Demo 05: CNNBefore Foundation Models
Transformer
Attention creates direct connections between tokens and became a key architecture for large language and multimodal models.
View Demo 06: AttentionModern AI Systems
LLM Systems
Large language models have powerful generation capabilities, but modern applications usually organize them with context, tools, memory, and evaluation.
View Chapter 07: LLM System MapModern AI Systems
RAG
RAG retrieves external knowledge into the context, improving factuality, freshness, and citation value.
View Demo 08: RAGAgentic AI
Agent
An LLM enters a loop of planning, tool calls, observation, and revision, allowing it to execute multi-step tasks instead of only answering once.
View Demo 09: Agent