Meridian's ramp-safety team built a rule-based system in the 1990s that flags foreign-object debris only when it exactly matches one of 200 pre-coded object shapes, and it has never been updated to reason about partial matches or uncertainty. What kind of system is this, and what is its key limitation?

CPMAI Task 1 asks candidates to recognize symbolic systems, expert systems, and fuzzy logic as related but distinct AI concepts. A system driven by 200 hand-coded exact-match rules is a symbolic/expert system: it encodes human expert knowledge as explicit rules and executes them deterministically, which is exactly why it is reliable on known cases but brittle on anything outside its coded rule set — it cannot reason about partial matches or degrees of confidence, which is what fuzzy logic (or a modern learned model) would add. The reinforcement-learning distractor is wrong because nothing in the scenario describes an agent taking actions, receiving rewards, and exploring an environment — it's a static rule matcher, not a learning agent. The generative-AI/hallucination distractor misapplies a large-language-model failure mode (fabricating plausible-sounding but false output) to a rigid rule-matcher, which by design cannot 'invent' a match — it either matches a coded shape or does not. The unsupervised-clustering distractor is wrong on the mechanism (there is no clustering of unlabeled data happening) and also on the stated limitation (explainability is generally a strength of rule-based systems, not a weakness, since every decision traces to an explicit rule). The correct diagnosis — brittle symbolic system — points the PM toward augmenting or replacing it with pattern-recognition AI rather than adding more hand-coded rules.