In artificial intelligence, a symbolic approach was a dominant approach in the past. The idea is that there are well-defined discrete symbols and mechanisms to manipulate symbols. They were used to prove theorems in mathematics but were too abstract to generalise to the real-world.
Lisp was the main language used in the 20th century. Mechanisms included support vector machines.
Basics
There are several classes of symbolic AI problems:
- In search problems, we use problem-specific state representations and heuristics. We’re generally concerned about determining paths from the current state to goal states. And we view states as black boxes with no internal structures. We also generally use a generic algorithm.
- In constraint satisfaction problems (CSPs), we care less about paths and more about final configurations. We take advantage of a general state representation. This uniform state representation allows design of more efficient algorithms.
- Knowledge representation and reasoning (KR&R) allows us to encode the formal specification of the problems using logical statements.