Perception · representation · reasoning · learning · action

Artificial Intelligence

Artificial intelligence studies computational systems that perceive or receive information, represent a task or environment, reason or search, choose actions or outputs, and sometimes improve from experience. Machine learning is one important method family inside that broader problem.

Agent loop

Intelligence problems couple information to choice.

Not every AI system needs every stage, and the stages can be tightly fused. The loop is a useful way to ask what information the system has, what structure it maintains, how it chooses, and what can change with experience.

Observe

Receive a state description, sensor stream, language input, database, or other evidence about the environment and task.

Represent

Encode relevant state, goals, uncertainty, relationships, features, beliefs, memory, or learned representations in a computational form.

Reason / search

Infer consequences, compare alternatives, solve constraints, search a state space, or estimate likely outcomes under the model.

Choose & act

Select an answer, plan, control action, prediction, generated output, or other behavior according to the system's objective and constraints.

Learn / revise

Use feedback, data, reward, errors, or new evidence to update models, parameters, representations, strategies, or beliefs.

Primary branches · navigation

Different AI problems emphasize different parts of the loop.

choose a path

Search & Planning

How can an agent efficiently explore possible states and action sequences toward a goal?

planned

represent and infer

Knowledge & Reasoning

How can structured knowledge, rules, probabilities, or beliefs support computational reasoning?

planned

improve from data

Machine Learning

How can model parameters or representations be fitted from examples and evaluated on new cases?

open

learn from consequences

Reinforcement Learning

How can an agent learn a policy when actions influence future observations and delayed rewards?

planned

interpret rich signals

Language & Perception

How can systems extract, represent, or generate meaning from language, images, audio, and other observations?

planned

Method families

AI is not one algorithmic paradigm.

Symbolic & search methods

Represent states, rules, constraints, goals, or logical relationships explicitly, then manipulate those structures with search, inference, planning, or optimization.

Probabilistic methods

Represent uncertainty directly and update beliefs or decisions using probability models, graphical structures, estimation, and expected outcomes.

Learning methods

Fit functions, representations, policies, or generative models from data or interaction instead of specifying every behavior rule manually.

Hybrid systems

Combine learned perception or prediction with symbolic constraints, search, planning, tools, memory, optimization, simulation, or human oversight.

Algorithms & DataSearch, optimization, graph traversal, and representation are core tools throughout AI.Formal LogicSymbolic representation and inference connect AI to formal semantics and proof.StatisticsProbability, inference, experimental design, and uncertainty support learning and evaluation.