Traditional AI
Traditional AI is an informal label for established AI methods contrasted with a newer approach. Depending on the speaker, it can mean rule-based reasoning and search, or classical machine learning contrasted with deep or generative models.
A support system might route requests using explicit rules: payment failures go to billing, login failures to technical support. Another might learn the same labels with a support vector machine. A speaker could call either “traditional AI,” but one applies written rules and the other learns a statistical decision function.
The field includes search, logical reasoning, planning and statistical learning. “Traditional” does not identify a single architecture, training method or date cutoff. Symbolic AI is a more precise name for approaches that manipulate explicit symbols and rules; it is not interchangeable with every use of traditional AI.
This matters when choosing components. A rule can enforce a known routing condition; a learned classifier can handle variation that would require many hand-written rules. Name the method, its inputs and its limits before comparing accuracy, maintainability or auditability. An older method is not automatically unsuitable for a task.
Sources
- UC Berkeley CS188: Introduction to Artificial Intelligence — Covers search, logical reasoning, planning and learning as separate AI approaches; it does not establish a formal “traditional AI” taxonomy.
Go deeper
- Poole and Mackworth: Propositions and Inference course
Explore a concrete reasoning method built on propositions and inference rules.