You are on page 1of 2

Four approaches to defining AI systems

CS 4633/6633: Artificial Intelligence


Intelligent Agents (chapter 2)
Systems that think like humans Systems that think rationally.

Systems that act like humans.

Systems that act rationally.

Cognitive science
An interdisciplinary field of study that combines AI and psychology It builds computer systems that simulate human mental processes, as a way of understanding how the human mind works Based on experimental investigation of humans (and animals)

Rational agent approach to AI


Russell and Norvig define AI as the study and construction of rational (intelligent) agents. This is a broader definition of intelligence than human intelligence (we should not expect computer intelligence to be the same as human intelligence anymore than we expect airplanes to fly the same as birds)

Intelligent behavior is also a broader concept than intelligent thought (intelligent behavior may depend on reflex, not thinking)

Agent
An agent is anything that can be viewed as perceiving its environment through sensors and acting upon that environment through effectors.

Type of environments
Fully vs. partially observable
Chess vs. poker

Deterministic vs. stochastic


Poker vs. backgammon

Episodic vs. sequential


Package delivery vs. product assembly

Static vs. dynamic


Navigation in storage area vs. supermarket

Discrete vs. continuous


Computer diagnosis vs. medical diagnosis

Single agent vs. multiagent


Soitaire vs. chess

Characterize these environments


Environment Chess Poker Backgammon Taxi driving Medical diagnosis Fully observable? Deterministic? Episodic? Static? Discrete? Single agent?

Rational agent
A rational agent is something that attempts to achieve its goals, given its beliefs A rational agent is capable of autonomous behavior Examples: robots, softbots (software agents) How are agents different from traditional computer systems?

The real world is partially observable, stochastic, sequential, dynamic, continuous and multi-agent

Types of agent designs


(in order of increasing generality)

Some example agents


Agent type
Medical diagnonsis system Satellite image analysis system Part-picking robot Automated taxi driver

Reactive agents (pre-programmed behavior)


Simple reflex agents Agents with memory to keep track of the world

Percepts
Symptoms, findings, patients answers Pixels of various intensity, color Pixels of varying intensity Cameras, speedometer, GPS, sonar, microphone

Actions
Questions, tests, treatements Print a categorization of scene Pick up parts and sort into bins Steer, accelerate, brake, talk to passenger

(performance measure)

Goals

Environment
Patient, hospital

Healthy patient, minimize costs Correct categorization Place parts in correct bins

Deliberative agents (can think and plan)


Goal-based agents Utility-based agents

Images from orbiti satellite Conveyor belt with parts

Learning agents (can learn from experience)

Safe, fast, legal, Roads, other comfortable trip, traffic, maximize profits pedestrians, customers

Some agents we will study this semester


Problem-solving agent
a type of goal-based agent that uses search to find a sequence of actions that will achieve a goal state

Knowledge-based agent
performs reasoning based on knowledge that is explicitly represented in a logical language

Planning agent
similar to a problem-solving agent, but uses more sophisticated methods of representation and search

Learning agent
can adapt to a new or changing environment

The book discusses many kinds of agents

You might also like