Introduction to Artificial Intelligence — Lecture 1 Study Materials Flashcards

Front

Turing Test

Back

A behavioral test where a machine is evaluated by whether an interrogator can distinguish its responses from a human's. Passing the test suggests the system acts humanly in conversation, though it does not prove understanding.

Front

Chinese Room

Back

A philosophical thought experiment by Searle arguing that syntactic symbol manipulation does not imply semantic understanding. It challenges claims that running the right program suffices for genuine mental states.

Front

PEAS

Back

A framework for specifying AI tasks: Performance measure, Environment, Actuators, Sensors. It helps designers clearly define what an agent should do and how it interacts with its world.

Front

Rational Agent

Back

An agent that acts to maximize expected performance according to a predefined performance measure. Rationality depends on the agent's percepts, prior knowledge, and computational limits.

Front

Simple Reflex Agent

Back

An agent that selects actions based solely on current percepts using condition-action rules. It is fast but can be short-sighted and fail in partially observable environments.

Front

Reflex Agent with State

Back

A reflex agent augmented with internal state that summarizes past percepts. This state allows the agent to handle partially observable environments by remembering relevant history.

Front

Goal-Based Agent

Back

An agent that chooses actions by planning toward states that satisfy explicit goals. It supports flexible behavior but requires search and can be sensitive to changes during deliberation.

Front

Utility-Based Agent

Back

An agent that uses a utility function to evaluate and compare states, enabling trade-offs among competing goals. Utility provides a smooth preference measure beyond binary goal satisfaction.

Front

Learning Agent

Back

An agent that improves its performance over time by updating its knowledge, components, or policies based on experience and feedback. Learning enables adaptation to new or changing environments.

Front

Fully Observable

Back

An environment property where the agent's sensors give access to the complete state of the environment at each time. Fully observable environments simplify decision making and planning.

Front

Partially Observable

Back

An environment where sensors provide incomplete or noisy information about the true state. Agents need memory, state estimation, or probabilistic reasoning to act effectively.

Front

Deterministic Environment

Back

An environment where the next state is completely determined by the current state and the agent's action. Deterministic settings allow predictable planning without modeling stochastic transitions.

Front

Episodic Task

Back

A task where each agent-environment interaction is divided into separate episodes, with each episode independent of previous ones. Episodic tasks reduce the need for long-term memory and planning.

Front

Dynamic Environment

Back

An environment that can change while the agent is deliberating or acting, often requiring real-time responses and continuous re-planning. Dynamic settings favor reactive or incremental planning approaches.

Front

Multiagent System

Back

A setting with multiple agents whose actions may affect each other, leading to strategic or cooperative behavior. Multiagent problems often require game-theoretic or coordination techniques.

Front

Reinforcement Learning

Back

A learning paradigm where agents learn policies by trial-and-error interactions, receiving rewards that guide behavior. It is effective for sequential decision problems with feedback signals.

Front

Bayesian Network

Back

A probabilistic graphical model representing dependencies among variables using directed acyclic graphs. It supports compact representation and inference under uncertainty.

Front

A* Search

Back

A best-first search algorithm that uses a heuristic to guide path finding toward a goal while guaranteeing optimality if the heuristic is admissible. A* balances exploration and exploitation via estimated cost.

Front

Dartmouth Conference

Back

The 1956 workshop widely considered the founding event of AI as an academic field. Key attendees included McCarthy, Minsky, Newell, and Simon, who helped establish early goals and approaches.

Front

Loebner Prize

Back

An annual competition implementing a restricted Turing Test to evaluate conversational systems. It awards prizes to programs deemed most human-like by human judges.