Artificial Intelligence: Foundations and Intelligent Agents

Essential vocabulary flashcards covering AI definitions, Turing Test concepts, intelligent agents, the PEAS framework, environment types, and agent architectures.

Artificial Intelligence (AI)

The computational field concerned with reproducing human reasoning and intelligent behavior through computational methods, categorized historically into thinking humanly, acting humanly, thinking rationally, and acting rationally.

Thinking Humanly

An approach to AI that models cognitive processes and mental faculties by introspecting thoughts, conducting psychological experiments, or using brain imaging such as MRI to understand how human brains work.

Acting Humanly

An approach to AI focused on building machines that perform functions requiring human intelligence, measured primarily by whether a machine can pass the Turing Test.

Turing Test

A test designed by Alan Turing in 1950 where a human interrogator engages in written conversation with a human and a computer; the machine passes if the interrogator cannot tell whether the responses came from a machine or a person.

Total Turing Test

An extended version of the Turing Test requiring physical interaction with the environment, testing computer vision to recognize objects and gestures, as well as robotics to move and manipulate physical objects.

Thinking Rationally

An approach to AI based on the formal 'laws of thought' and logical reasoning (pioneered by Aristotle's syllogisms) to ensure correct deductions, though often limited by computational complexity and real-world uncertainty.

Acting Rationally

An approach to AI that focuses on creating rational agents that act to achieve the best expected outcome or maximize expected utility, providing a more general and mathematically principled foundation than thinking rationally.

Cognitive Science

An interdisciplinary field formed by the merger of AI computational models and experimental psychology to study how human minds perceive, process information, and think.

Dartmouth Meeting (1956)

A historic academic conference that marked the inception of artificial intelligence as an official field of study, introducing early programs such as the Logic Theorist.

AI Winter (1966–1973)

A period in AI history marked by funding cutbacks and declining enthusiasm due to the computational intractability of many problems and the temporary disappearance of neural networks.

Knowledge-Based Systems

AI systems developed prominently between 1969 and 1979 that rely on domain-specific human knowledge to facilitate powerful reasoning and expert decision-making.

Agent

Anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators.

Sensor

A device or organ that receives input from the environment (e.g., eyes and ears in human agents; cameras and infrared range finders in robotic agents).

Actuator

A mechanism or body part through which an agent performs actions to alter its environment (e.g., hands, legs, and mouth in humans; motors in robotic agents).

Percept

The agent's perceptual inputs at any specific given moment.

Percept Sequence

The complete history of everything the agent has ever perceived over time.

Agent Function

An abstract mathematical mapping from percept histories to actions, represented mathematically as $f: P^* \rightarrow A$.

Agent Program

The concrete computational algorithm that runs on an agent's physical architecture to implement the agent function.

Agent Architecture

The physical computing hardware and devices on which an agent program executes, summarized as $\text{agent} = \text{architecture} + \text{program}$.

Intelligent Agent vs. Robot

A robot is a tangible physical entity that manipulates the physical world (with or without intelligence), whereas an intelligent agent makes cognitive decisions and may be entirely software-based.

Performance Measure

An external objective criterion used to evaluate the success and effectiveness of an agent's behavior.

Rational Agent

An agent that selects an action expected to maximize its performance measure for each possible percept sequence, given the percept evidence and its built-in knowledge.

Rationality vs. Perfection

Rationality maximizes expected outcome based on available knowledge and percepts, whereas perfection maximizes actual outcome and requires complete foreknowledge.

Autonomy

The extent to which an agent's behavior is determined by its own experience rather than solely by the prior knowledge designed into it.

PEAS

A task environment specification framework that defines an intelligent agent by its Performance measure, Environment, Actuators, and Sensors.

Fully Observable Environment

An environment where an agent's sensors provide complete access to the full state of the environment at any given time, providing perfect information.

Partially Observable Environment

An environment where parts of the state are missing, noisy, or inaccessible via sensors, requiring the agent to make informed guesses.

Deterministic Environment

An environment where the next state is completely determined by the current state and the action chosen by the agent.

Stochastic Environment

An environment that involves aspects of randomness or uncertainty beyond the agent's control, meaning the next state cannot be predicted with certainty.

Episodic Environment

An environment divided into discrete, independent episodes where current action choices do not affect subsequent episodes, eliminating the need to plan ahead.

Sequential Environment

An environment where current actions affect future states and decisions, requiring the agent to plan ahead.

Static Environment

An environment that remains unchanged while the agent is deliberating and deciding which action to take.

Dynamic Environment

An environment that continuously changes while the agent is deliberating, forcing the agent to anticipate changes or make fast decisions.

Semidynamic Environment

An environment where the world itself does not change over time, but the agent's performance score does change with the passage of time.

Discrete Environment

An environment with a finite number of distinct, clearly defined states, percepts, and possible actions (e.g., chess moves).

Continuous Environment

An environment characterized by states, percepts, or actions that smoothly vary over a continuous range of numerical values (e.g., automated vehicle speed and steering angle).

Single-Agent Environment

An environment in which an agent operates independently without interacting with or being affected by other agents.

Multiagent Environment

An environment containing more than one agent, where agents interact either cooperatively or competitively.

Simple Reflex Agent

An agent architecture that chooses actions based solely on the current percept using condition-action rules, without maintaining percept history or internal memory.

Model-Based Reflex Agent

An agent architecture that maintains an internal state based on percept history to track unseen parts of the world, relying on models of how the world evolves and how its own actions affect the world.

Markovian State Space

A state space representation where each internal state includes all relevant information for making decisions, satisfying $f: P^*, A^* \rightarrow S \rightarrow A$, such that the future depends only on the current state.

Perfect Recall

The property of an agent's memory in game theory where each state explicitly contains the complete history of all percepts and actions leading to it.

Goal-Based Agent

An agent architecture that combines world state knowledge with explicit goal descriptions to evaluate alternative actions and plan sequences that achieve desired destinations.

Utility-Based Agent

An agent architecture that uses a utility function to score states on a real-valued scale, evaluating trade-offs between competing goals and measuring the degree of success or happiness.

Utility Function

A mathematical function that maps a world state onto a real number to quantify the degree of happiness, goodness, or success associated with that state.

Learning Agent

An agent architecture composed of a critic, a learning element, a performance element, and a problem generator, enabling it to learn from experience and adapt to initially unknown environments.

Critic (Learning Agent)

The component of a learning agent that evaluates the agent's behavior by comparing sensory percepts against a fixed external performance standard to provide feedback.

Learning Element

The component of a learning agent that uses feedback from the critic to make improvements and modify the performance element for future operations.

Performance Element

The operational component of a learning agent responsible for collecting percepts and selecting external actions, corresponding to the entire agent in non-learning architectures.

Problem Generator

The component of a learning agent that suggests exploratory actions that lead to new experiences and problems, preventing the agent from settling into suboptimal routines.

Exploration vs. Exploitation

The strategic tradeoff in learning agents between trying new actions to gain information (which may incur short-term costs) and taking familiar actions to maximize short-term rewards.