AI Foundations and Applications
What this domain tests
AI Foundations and Applications examines whether you can define AI precisely, separate it from ordinary automation and from machine learning, and match AI capabilities to real products. Questions often ask you to classify a system and justify the classification.
Core rules and concepts
- AI is the design of systems that perform tasks associated with human intelligence: perception, language, reasoning, learning, planning and action.
- Narrow AI is built for a bounded task. General AI would transfer across arbitrary tasks; it is not a deployed commercial product.
- The families nest: AI contains machine learning, and machine learning contains deep learning. ML learns patterns from data rather than relying only on hand-written rules.
- A learning system needs data, a model, an objective or loss, and an evaluation method.
- Rule-based systems can still count as AI when they encode reasoning, but they do not improve from data unless a learning component is added.
- Common applications include recommendation, speech assistants, spam filtering, medical triage support and navigation.
Worked example
A school wants to flag abusive comments. This is narrow AI because the task is bounded. It is machine learning if labelled examples train a classifier; it is rule-based if it only blocks a banned-word list. Perception is the text input, reasoning is classification, and action is flagging or hiding the comment. Performance must be checked on unseen comments, not on the training examples.
Traps that cost marks
Do not call every robot AI or every AI a robot. Do not treat ML and AI as synonyms. Do not describe general AI as currently available. Do not claim a rule-based chatbot learns when it has no training mechanism.