1.1 What Is AI?
AI can be viewed through human-like thought or behavior, and through rational thought or behavior. The rational-agent view asks what action best advances the system’s objectives given its information.
Artificial Intelligence · Chapter 1
Introduction
AI studies agents that perceive their surroundings and choose actions. The central question is how to build systems that act effectively, reason well, and remain useful to people.
Chapter guide
1.2 The Foundations of Artificial Intelligence
AI draws on philosophy, mathematics, economics, neuroscience, psychology, computer engineering, control theory, and linguistics. These fields contribute ideas about knowledge, learning, decisions, brains, machines, and communication.
1.3 The History of Artificial Intelligence
The field grew from formal logic and early computing, through symbolic reasoning and expert systems, to statistical learning and data-driven methods. Progress has alternated between ambitious advances and periods of reduced expectations.
1.4 The State of the Art
Modern AI excels at bounded tasks such as perception, language, games, and prediction. Competence in one domain does not imply general human-level understanding or reliable performance in every setting.
1.5 Risks and Benefits of AI
AI can improve health, access, productivity, and scientific discovery. Safety, fairness, privacy, labor disruption, misuse, and alignment with human goals require deliberate evaluation and governance.
Key terms and a useful distinction
Agent: a system that perceives and acts. Rationality: choosing an action expected to do well against a stated performance measure, using the information available—not knowing the future. Model: the agent’s representation of how the world works. A narrow system can perform impressively on one task without having general intelligence.
Example: A delivery robot can choose a shorter route when the sidewalk is clear, but wait for a better observation when sensor readings are unreliable. The sensible choice depends on both its goal (timely delivery) and what it knows about the environment.
Interactive lab
An agent chooses an action
Change the goal pressure and uncertainty. Watch a simple rational agent trade off progress against risk—one small example of the agent perspective.
Current stateChoosing
WAITACT
Selected action: Gather more information
Moderate uncertainty makes information valuable.
A teaching model only: real agents need explicit objectives, calibrated uncertainty, and safeguards.
How to read this demo
The policy is deliberately transparent: if uncertainty is at least 65%, gather information; otherwise act when goal importance is at least 55%; otherwise wait. For example, goal = 70 and uncertainty = 30 selects the goal-directed action. These thresholds are a toy rule, not a utility-maximizing proof or a calibrated model of risk.
Complexity: Inputs are goal importance g and uncertainty r (each 0–100). For each input change, the policy reads these two scalars and updates a fixed number of labels/styles: O(1) time and O(1) auxiliary space. The displayed agent and controls are fixed-size UI; no search tree or history is stored.