Artificial Intelligence Fundamentals and Concepts
Core concepts
Artificial intelligence is the engineering of systems that perform tasks associated with human cognition. Candidates should distinguish narrow (weak) AI, which is task-bound, from general AI, which is hypothetical, and avoid treating today's large models as general intelligence. Know the classic categories: rule-based/symbolic systems (explicit logic, brittle outside their rules) versus data-driven/statistical systems (patterns learned from examples). Also understand search (state spaces, heuristics), knowledge representation (ontologies, graphs), and the agent model (perception, reasoning, action, environment).
Vocabulary that gets tested
Be precise about inference (drawing conclusions), heuristics (rules of thumb that trade optimality for speed), deterministic vs stochastic behavior, and supervised vs unsupervised vs reinforcement learning at a conceptual level. Know that an algorithm is a procedure and a model is the artifact produced by training.
Common traps
Questions often blur AI, machine learning, and deep learning as synonyms; they are nested subsets (DL ⊂ ML ⊂ AI). Another trap is confusing automation with intelligence: a scripted workflow is automation, not necessarily AI. Watch for claims that a system "understands" or "knows" in a human sense; exam-correct answers describe behavior and capability, not consciousness. Finally, do not assume more data always helps-relevance and quality matter.
Rule of thumb
Use the capability test: ask what the system can do that it was not explicitly programmed to do. If the answer is "nothing-every step was hand-coded," it is automation, not learning.
Worked example
A spam filter that follows a fixed list of banned words is rule-based automation; one that learns word patterns from labeled emails is supervised machine learning. The label (spam/not spam) is what makes it supervised.