Symbolic AI represents knowledge explicitly and solves problems by manipulating hand-crafted rules. It was the dominant AI approach from the 1950s through the late 1980s, before machine learning displaced it for many tasks. section 1.2
Hard-coded chess programs
An early chess program could use rules written by programmers rather than learn from data. This worked for a well-defined logical problem, but writing explicit rules became impractical for fuzzy tasks such as image classification, speech recognition, and language translation. section 1.2
Symbolic AI is associated with:
- Explicit databases that store knowledge for rule-based manipulation. section 1.2
- Expert systems, a later symbolic-AI form that attracted major corporate investment during the 1980s. section 1.11
- High performance on well-defined logical problems, alongside difficulty scaling to complex, fuzzy problems. section 1.2 section 1.11
Its limitations contributed to a shift toward machine learning. Expert systems also proved expensive to maintain, difficult to scale, and limited in scope, helping trigger the second AI winter in the early 1990s. section 1.11 section 1.2
