Symbolic AI — the dominant paradigm from 1956 to roughly 1990 — held that intelligence is symbol manipulation according to rules. The Physical Symbol System Hypothesis (Newell and Simon, 1976) is the canonical statement. The paradigm produced real systems, then hit real limits.
- 01State the Physical Symbol System Hypothesis.
- 02Name three canonical symbolic AI systems and what they did.
- 03Explain why the paradigm plateaued.
The claim
Intelligence is the manipulation of symbols according to formal rules. Reasoning is search through a space of symbolic representations. Learning is the acquisition of new rules. All human cognition is, on this view, in principle reducible to this process.
The limits
Symbolic systems required exhaustive hand-coded knowledge. They were brittle at the edges of their explicit rules and useless in domains where the rules were tacit — vision, speech, common-sense reasoning. Statistical and neural methods eventually filled the gap.
- Symbolic AI is rule-based reasoning over explicit representations.
- It scales badly on domains humans handle tacitly.