Artificial intelligence: the model is one part of the system

Artificial intelligence (AI) studies computational systems that perceive, reason, learn or act toward goals. Machine learning is one approach within that wider field; search, planning and explicit reasoning also belong to it.

A system can combine these approaches. A chess engine can use a learned position evaluator inside a search procedure. A language-model agent can combine a learned model with retrieval and tool calls.

Capability comes from the combination

The model may propose an action, a planner may compare alternatives, and ordinary software may execute the chosen tool. Better search or a better interface can improve the whole system without adding a single model parameter.

This makes “does it use AI?” a less useful technical question than “which operation is learned, which is searched, and which is specified in code?”

Evaluate the resulting behaviour on the intended task. A fluent interface, a neural component and an autonomous loop describe different properties; none alone establishes broad competence.