Symbolic AI

Symbolic AI is the tradition in artificial intelligence that represents knowledge as explicit symbols and manipulates those symbols through formal rules of inference. In contrast to deep learning and other statistical approaches, symbolic systems operate on discrete structures — propositions, predicates, frames, and ontologies — that are compositional and transparently interpretable. The approach dominated AI from the 1950s through the 1980s, producing expert systems, theorem provers, and knowledge bases, before encountering scaling limits that the connectionist and statistical revolutions would later address. Symbolic AI is now experiencing renewed interest as the interpretable half of the neural-symbolic integration agenda.

The persistent limitation of symbolic systems is not their logic but their brittleness: they require human experts to specify what they know, and they fail gracefully only when grace has been explicitly programmed in.