Foundations
Symbolic AI
Also known as: GOFAI, rule-based AI
The older approach to AI based on explicit rules, logic and symbols, in contrast to today's data-driven, learned approach.
Symbolic AI is the "old school" way of building AI: people write explicit rules and logic, like "IF the patient has a fever AND a rash, THEN consider measles". It works with clear symbols and reasoning steps that humans can read. It powered early AI (expert systems) and is excellent when rules are known and precise, but it struggles with messy, real-world data, exactly where today's learning-based AI shines.
Symbolic AI (sometimes "GOFAI", good old-fashioned AI) represents knowledge as explicit symbols and rules, and reasons using logic and search. It dominated AI from the 1950s into the 1980s (expert systems, logic programming, knowledge bases).
Strengths and weaknesses
- Strengths: transparent and explainable, precise, good at formal reasoning, no training data needed.
- Weaknesses: brittle, hand-crafting rules does not scale, and it struggles with perception and ambiguity (images, natural language).
Symbolic vs learned
Modern AI is dominated by machine learning and deep learning, which learn from data rather than following written rules. There is growing interest in neuro-symbolic approaches that combine the two: learning's flexibility plus symbolic reasoning's rigour and transparency.
Symbolic AI encompasses paradigms that manipulate explicit symbolic representations, logic, rules, ontologies, search, planning, under the physical-symbol-system hypothesis. It delivered early successes (expert systems, theorem proving, constraint solving) and offers transparency, verifiability and strong formal reasoning without data, but suffers brittleness, the knowledge-acquisition bottleneck, and poor handling of perception and ambiguity, limitations that drove the shift to statistical/connectionist methods.
The contrast with data-driven deep learning is complementary rather than absolute: symbolic methods excel at compositional, verifiable reasoning; learned methods excel at perception and generalisation from data. Neuro-symbolic research seeks integration, e.g. LLMs invoking solvers/tools, program synthesis, or structured reasoning, to combine flexibility with reliability and explainability. Symbolic techniques also persist throughout modern AI stacks (search, planning, knowledge graphs, formal verification) even where neural models handle perception and generation.