What is Symbolic AI?
Four approaches: symbolic vs statistical vs subsymbolic vs embodied
Symbolic AI is the branch of artificial intelligence that uses symbols —
discrete, manipulable representations of concepts — to reason about the world.
There are four main approaches to AI today. Each has different assumptions
about how intelligence arises:
1. Symbolic AI — intelligence comes from manipulating symbols (this course!)
2. Statistical AI — intelligence comes from learning patterns in data
3. Subsymbolic AI (neural / connectionism) — intelligence emerges from
many small numerical units firing together
4. Embodied AI — intelligence cannot exist without a body in an environment
Symbolic AI shines when:
- The problem has clear rules and structure (math, logic, programs)
- You need to explain a decision (provenance is preserved)
- You want correctness guarantees (proofs, types, formal verification)
Statistical / subsymbolic AI shines when:
- Rules are hard to write but data is plentiful
- Slight errors are tolerable
- You have lots of compute and training time
Most modern AI systems are hybrid: a symbolic shell around a statistical core,
or a statistical core wrapped in symbolic constraints.