Concept · Chapter 1: What Is Artificial Intelligence?
Symbolic AI
Symbolic AI represents knowledge as explicit symbols and rules written by people, and produces intelligent behaviour by manipulating them — through logic and search.
The problem
How do you make a machine reason, plan, prove theorems or diagnose disease?
The solution
Write down what the machine needs to know as symbols (facts, rules, goals) and have it manipulate them with general procedures: logical inference and search through possible steps.
The consequence
It produced theorem provers, planners, chess engines and expert systems — and hit a wall wherever knowledge was too vast, fuzzy or tacit to write down.
You should understand first
- The Turing Test
- Symbolic AI
The idea
The founders of AI noticed that much of what we call intelligence looks like symbol manipulation: proving a theorem, solving a puzzle, planning a route, following a chain of reasoning. So they built programs that manipulate symbols.
A symbolic system has two parts:
- Knowledge, written explicitly: facts like
parent(alice, bob), rules like "if X is a parent of Y and Y is a parent of Z, then X is a grandparent of Z". - General procedures that operate on it: logical inference (derive new facts from rules) and search (explore sequences of possible steps until one reaches a goal).
Newell and Simon's physical symbol system hypothesis (1976) stated the bold version: symbol manipulation is both necessary and sufficient for general intelligence Interpretation. Few researchers would defend the "necessary" half today.
What it achieved
Early successes were real and impressive: the Logic Theorist (1955) proved mathematical theorems; SRI's Shakey robot (1966–72) planned its own actions; A* search (1968) found optimal paths; expert systems in the 1970s–80s diagnosed infections and configured computers; and in 1997 Deep Blue beat the world chess champion — with search, not learning.
Where it struggled
Symbolic AI works best where the world is well-defined: clear rules, discrete states, checkable goals. It struggled with perception (what are the "symbols" in a photo?), with uncertainty and exceptions, and above all with the sheer amount of everyday knowledge a system needs — the knowledge-acquisition bottleneck.
Why should I care?
As a researcher
Current debates about reasoning, planning and 'neurosymbolic' methods revisit symbolic AI's goals with learned components. Knowing what it achieved — and why it stalled — keeps you from reinventing either.
As an engineer
Rules, search and knowledge graphs are still everywhere: route planners, compilers, SQL optimizers, business-rule engines, and the tool-calling scaffolding around LLMs.
Modern systems that depend on it
- Search and planning algorithms
- Knowledge graphs
- Rule engines and expert systems
- Agent planning loops
- Game-playing search (with learned evaluation)
Historical context
Before
Cybernetics and early neural models in the 1940s–50s; Turing's framing of machine intelligence.
After
Statistical machine learning from the late 1980s, then deep learning — which learn their knowledge from data instead of having it written in.
Used today
In search and planning (maps, logistics, games), in compilers and databases, in knowledge graphs, and increasingly as structured tools and plans wrapped around learned models.
What to remember
- Knowledge = explicit symbols and rules, written by people.
- Reasoning = manipulating symbols: logical inference and search.
- Strengths: precise, inspectable, good at well-defined problems.
- Weakness: brittle and expensive where knowledge is vast, uncertain or tacit.
- It dominated AI from the 1950s to the 1980s.
Key papers
A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, August 31, 1955
John McCarthy, Marvin L. Minsky et al. · 1955 · Reprinted in AI Magazine 27(4), 2006
The document that named the field 'artificial intelligence' and set out its founding ambition.
How to read it: Short. Notice how many of its topics — language, neuron nets, abstraction, self-improvement — are still open problems.
Computer science as empirical inquiry
Allen Newell, Herbert A. Simon · 1976 · Communications of the ACM
The classic statement of the symbolic-AI worldview: the physical symbol system hypothesis and heuristic search as the heart of intelligence.
How to read it: Their 1975 Turing Award lecture. Read it as the best case for the approach this chapter shows running into limits.