Back to Study Material Artificial Intelligence

Quick revision

This section pulls together the key definitions, distinctions, and formulas from across the entire Artificial Intelligence topic area, giving you a structu...

Quick revision

Core AI Concepts at a Glance

This section pulls together the key definitions, distinctions, and formulas from across the entire Artificial Intelligence topic area, giving you a structured last-pass review before an examination.

An intelligent agent perceives its environment through sensors and acts through actuators. PEAS (Performance, Environment, Actuators, Sensors) is the framework for specifying agents. Task environments are classified as fully or partially observable, deterministic or stochastic, episodic or sequential, static or dynamic, single or multi-agent, discrete or continuous.

Turing Test: a machine passes if a human judge cannot reliably distinguish it from a human in text conversation. Strong AI is hypothetical general intelligence; weak AI is all current systems — specialized, task-specific.

Knowledge representation: predicate logic and first order logic (FOL) use predicates, constants, variables, and quantifiers to express facts. Frames store stereotypical object knowledge in slot-filler structures. Scripts represent event sequences with roles and scenes. Non-monotonic and default reasoning allow retraction of conclusions when new information defeats assumptions.

Production systems use IF-THEN rules and a recognize-act cycle. Forward chaining is data-driven (facts to conclusions). Backward chaining is goal-driven (goal to supporting facts). Both are implemented in expert systems and Prolog.

Planning: STRIPS uses preconditions, add lists, and delete lists. The frame assumption holds unchanged facts constant. Means-end analysis identifies differences and applies operators to reduce them. Planning graphs provide heuristics for planners like GraphPlan.

Search: hill climbing gets stuck at local maxima, plateaus, and ridges. Simulated annealing escapes by accepting worse solutions with probability exp(delta E / T). Genetic algorithms use selection, crossover, and mutation on a population.

Swarm intelligence: ACO uses pheromone trails; PSO uses particle velocities influenced by personal and swarm best positions. Knowledge graphs store entity-relationship triples. AI safety addresses alignment, specification gaming, and robustness. Human-in-the-loop AI keeps people active in critical decisions.

Exam tip: Make flashcards for every bolded term above. In AI examinations, precise definitions are worth full marks and vague answers lose them.