artificial intelligence a brief introduction

aslab 12004/03/24Sanz / Artificial Intelligence: An Introduction

aslab.org

May 20, 2004

Artificial IntelligenceA Brief Introduction

Ricardo Sanz

autonomous systems laboratory

2004/03/24Sanz / Artificial Intelligence: An Introduction 2aslab

ContentsContents

Basic Ideas History Technology Robots Agents


Core Ideas

What is AI ?


What is AI?What is AI?

Acting humanly: The Turing test (1950) What do we need to pass the test

Thinking humanly: Cognitive modeling “Think-aloud” to learn from human and recreate in computer

programs (GPS)

Thinking rationally: Syllogisms, Logic Acting rationally: A rational agent


Foundations of AIFoundations of AI

Philosophy (428 B.C. - Present) – reasoning and learning Can formal rules be used to draw valid conclusions? How does the mental I arise from a physical brain? Where does knowledge come from? How does knowledge lead to action?



Mathematics (c. 800 - Present) - logic, probability, decision making, computation What are the formal rules to draw conclusions? What can be computed? How do we reason with uncertain information?

Economics (1776-present) How should we make decisions so as to maximize payoff? How should we do this when others may not go along? How should we do this when the payoff may be far in the

future?



Neuroscience (1861-present) How do brains process information

Psychology (1879 - Present) - investigating human mind How do humans and animals think and act?

Computer engineering (1940 - Present) - ever improving tools How can we build an efficient computer?



Control theory and Cybernetics (1948-present) How can artifacts operate under their own control?

Linguistics (1957 - Present) - the structure and meaning of language How does language relate to thought?


What is Intelligence?What is Intelligence?

Intelligence, taken as a whole, consists of the following skills:

1. the ability to reason

2. the ability to acquire and apply knowledge

3. the ability to manipulate and communicate ideas


Has understanding/intentionality

Exhibits behaviour

SeeHearTouchTasteSmell

INPUTSINTERNAL PROCESSES

OUTPUTS

Senses environment

Can Reason

Has knowledge

An Intelligent EntityAn Intelligent Entity


The Age of Intelligent MachinesThe Age of Intelligent Machines

1st Industrial Revolution: the Age of Automation: Machines extend & multiply man's physical capabilities

2nd Industrial Revolution: the Age of Info Tech: Machines extend & multiply man's mental capabilities

Knowledge Revolution?: the Age of Knowledge Technology "..working smarter, not harder."

How do we make our systems smarter? - by building in intelligence?


More Definitions of AIMore Definitions of AI

AI is the science of making machines do things that would require intelligence if done by humansMarvin Minsky

AI is the part of computer science concerned with designing intelligent computer systemsEd Feigenbaum

Systems that can demonstrate human-like reasoning capability to enhance the quality of life and improve business competitivenessJapan-S’pore AI Centre


Behaviourist’s View on Intelligent MachinesBehaviourist’s View on Intelligent Machines

Many scientists believe that only things that can be directly observed are “scientific”

Therefore if a machine behaves “as if it were intelligent” it is meaningless to argue that this is an illusion.

Turing was of this opinion and proposed the “Turing Test”

This view can be summarized as:“If it walks like a duck, quacks like a duck and looks like a duck - it is a duck”


In 1950 Alan Turing published his now famous paper "Computing Machinery and Intelligence." In that paper he describes a method for humans to test AI programs.

In its most basic form, a human judge sits at a computer terminal and interacts with the subject by written communication only. The judge must then decide if the subject on the other end of the computer link is a human or an AI program imitating a human.

http://www.turing.org.uk/turing/

Turing’s TestTuring’s Test


Which one’s the man?

BA

Turing’s Test - Part 1Turing’s Test - Part 1


If the computer succeeds in fooling the judge then it has managed to exhibit a human level of intelligence in the task of pretending to be a woman, the definition of intelligence the machine has shown itself to be intelligent.

Which one’s the computer?

AB

Turing’s Test - Part 2Turing’s Test - Part 2


Some History

From hype to work


Brief History of AIBrief History of AI

Gestation of AI (1943 -1955) McCulloch and Pitts’s model of artificial neurons Minsky’s 40-neuron network

Birth of AI (1956) A 2-month Dartmouth workshop of 10 attendees – the name

of AI Newell and Simon’ Logic Theorist

Early enthusiasm, great expectations (1952 - 1969) GPS by Newell and Simon, Lisp by McCarthy, Blockworld by

Minsky



AI facing reality (1966 - 1973) Many predictions of AI coming successes

A computer would be a chess champion in 10 years (1957) Machine translation – Syntax is not enough Intractability of the problems attempted by AI

Knowledge-based systems (1969 - 1979) Knowledge is power, acquiring knowledge from experts Expert systems (MYCIN)

AI - an industry (1980 - present) Many AI systems help companies to save money and

increase productivity



The return of neural networks (1986 – present) PDP books by Rumelhart and McClelland Connectionist models vs. symbolic models

AI – a science (1987 – present) Build on existing theories vs. propose brand new ones Rigorous empirical experiments Learn from data – data mining

AI – intelligent agents (1995 – present) Working agents embedded in real environments with

continuous sensory inputs

AI - conscious machines (Now !!) Making machines that feel and and have a self


Degree ofMotivation

DartmouthConference

AIWinter

Support Technology

Time1948 1970s - 80s mid-1980s

Japan 5thGeneration Computer

mid-1990s

History of AIHistory of AI


Examples of AI systemsExamples of AI systems

Robots Chess-playing program Voice recognition system Speech recognition system Grammar checker Pattern recognition Medial diagnosis System malfunction rectifier

Game Playing Machine Translation Resource Scheduling Expert systems (diagnosis,

advisory, planning, etc) Machine learning Intelligent interfaces


AI Case Study - RoboCupAI Case Study - RoboCup

The Robocup Competition pits robots (real and virtual) against each other in a simulated soccer tournament.

The aim of the RoboCup competition is to foster an interdisciplinary approach to robotics and agent-based AI by presenting a domain that requires large-scale co-operation and coordination in a dynamic, noisy, complex environment.

Common AI methods used are variants of neural networks and genetic algorithms.


Intelligent Technologies

Resources for Sophisticated Information Processing


User

General Knowledge-baseInference engine

User interfacemay employ:

Question-and-Answer,

Menu-driven,

Naturallanguage,

GraphicsInterfaceStyles

Etc.

Knowledge-baseeditor

Explanationsubsystem

Case-specific data

Knowledge-Based Systems (KBS)Knowledge-Based Systems (KBS)


Synapse Neuron

Connectionbetween neurons

Outputs

Inputs

Plant

Sensors

InputOutput

Artificial Neural NetworksArtificial Neural Networks

What are Artificial Neural Networks (ANNs)? ANN or connecionist systems are systems that were developed based on

the learning characteristics of biological creatures. ANN solve problems though a process of learning and adaptation.

How are ANNs represented?


Genetic AlgorithmsGenetic Algorithms

We will use the processes loosely based on natural selection, crossover, and mutation to find solutions to certain problems.

GAs are adaptive (search, learning) methods based on the genetic processes of biological organisms.

A

1st generation of possible solutions

2nd generation of possible solutions


Fuzzy LogicFuzzy Logic

For systems with little complexity, hence little uncertainty, closed-form mathematical expressions provide precise description of the system.

For systems that are a little more complex, but for which significant data exists, model free methods such as artificial ANNs, provide a powerful and robust means to reduce uncertainty through learning.

For most complex systems where few numerical data exists and where only ambiguous or imprecise information may be available, fuzzy reasoning provides a way to understand system behavior.

Pre

cisi

on

in t

he

mo

del

Complexity (uncertainty) of the system

Mathematicalequations Model-free

Methods(e.g., ANNs)

Fuzzy Systems


Towards intelligent machines

Are we ready to build the next generation of intelligent robots?


Some problems remain…Some problems remain…

Vision Audition / speech processing Natural language processing Touch, smell, balance and other senses Motor control


Computer PerceptionComputer Perception

Perception: provides an agent information about its environment. Generates feedback. Usually proceeds in the following steps.

Sensors: hardware that provides raw measurements of properties of the environment Ultrasonic Sensor/Sonar: provides distance data Light detectors: provide data about intensity of light Camera: generates a picture of the environment

Signal processing: to process the raw sensor data in order to extract certain features, e.g., color, shape, distance, velocity, etc.

Object recognition: Combines features to form a model of an object

And so on to higher abstraction levels


Perception for what?Perception for what?

Interaction with the environment, e.g., manipulation, navigation Process control, e.g., temperature control Quality control, e.g., electronics inspection, mechanical parts Diagnosis, e.g., diabetes Restoration, of e.g., buildings Modeling, of e.g., parts, buildings, etc. Surveillance, banks, parking lots, etc. … And much, much more


Sample perception: Computer visionSample perception: Computer vision

1. Grab an image of the object (digitize analog signal)

2. Process the image (looking for certain features)1. Edge detection2. Region segmentation3. Color analysis4. Etc.

3. Measure properties of features or collection of features (e.g., length, angle, area, etc.)

4. Use some model for detection, classification etc.


State of the artState of the art

Can recognize faces? – yes

Can find salient targets? – sure

Can recognize people? – no problem

Can track people and analyze their activity? – yep

Can understand complex scenes? – not quite but in progress


Face recognition case studyFace recognition case study


Pedestrian recognitionPedestrian recognition


How about other senses?How about other senses?

Speech recognition -- can achieve user-undependent recognition for small vocabularies and isolated words

Other senses -- overall excellent performance (e.g., using gyroscopes for sense of balance, or MEMS sensors for touch) except for olfaction and taste, which are very poorly understood in biological systems also.


How about actuationHow about actuation

Robots have been used for a long time in restricted settings (e.g., factories) and, mechanically speaking, work very well.

For operation in unconstrained environments, Biorobotics has proven a particularly active line of research:

Motivation: since animals are so good at navigating through their natural environment, let’s try to build robots that share some structural similarity with biological systems.


Robot examples: constrained environmentsRobot examples: constrained environments


Towards unconstrained environmentsTowards unconstrained environments


They’re here …They’re here …

Robot lawn mowers and vacuum-cleaners are here already…


The time is nowThe time is now

It is a particularly exciting time for AI because… CPU power is getting not a problem anymore Many physically-capable robots are available Some vision and other senses are partially available

Many AI algorithms for constrained environment are available

So for the first time we have all the components required to build smart robots that interact with the real world.


Agents

Recent IA software focus


What is an Agent?What is an Agent?

in general, an entity that interacts with its environment perception through sensors actions through effectors or actuators


Examples of AgentsExamples of Agents

human agent eyes, ears, skin, taste buds, etc. for sensors hands, fingers, legs, mouth, etc. for effectors

powered by muscles

robot camera, infrared, bumper, etc. for sensors grippers, wheels, lights, speakers, etc. for effectors

often powered by motors

software agent functions as sensors

information provided as input to functions in the form of encoded bit strings or symbols

functions as effectors results deliver the output


Agents and Their ActionsAgents and Their Actions

a rational agent does “the right thing” the action that leads to the best outcome

problems: what is “ the right thing” how do you measure the “best outcome”


Performance of AgentsPerformance of Agents

criteria for measuring the outcome and the expenses of the agent often subjective, but should be objective task dependent time may be important


Performance Evaluation ExamplesPerformance Evaluation Examples

vacuum agent A number of tiles cleaned during a certain period

based on the agent’s report, or validated by an objective authority

doesn’t consider expenses of the agent, side effects energy, noise, loss of useful objects, damaged furniture,

scratched floor might lead to unwanted activities

agent re-cleans clean tiles, covers only part of the room, drops dirt on tiles to have more tiles to clean, etc.


Rational Agent considerationsRational Agent considerations

performance measure for the successful completion of a task

complete perceptual history (percept sequence) background knowledge

especially about the environment dimensions, structure, basic “laws”

task, user, other agents

feasible actions capabilities of the agent


OmniscienceOmniscience

a rational agent is not omniscient it doesn’t know the actual outcome of its actions it may not know certain aspects of its environment

rationality takes into account the limitations of the agent percept sequence, background knowledge, feasible actions it deals with the expected outcome of actions


Ideal Rational AgentIdeal Rational Agent

selects the action that is expected to maximize its performance based on a performance measure depends on the percept sequence, background knowledge,

and feasible actions


From Percepts to ActionsFrom Percepts to Actions

if an agent only reacts to its percepts, a table can describe the mapping from percept sequences to actions instead of a table, a simple function may also be used can be conveniently used to describe simple agents that

solve well-defined problems in a well-defined environment e.g. calculation of mathematical functions


Agent or ProgramAgent or Program

our criteria so far seem to apply equally well to software agents and to regular programs

autonomy agents solve tasks largely independently programs depend on users or other programs for “guidance” autonomous systems base their actions on their own

experience and knowledge requires initial knowledge together with the ability to learn provides flexibility for more complex tasks


Structure of Intelligent AgentsStructure of Intelligent Agents

Agent = Architecture + Program architecture

operating platform of the agent computer system, specific hardware, possibly OS functions

program function that implements the mapping from percepts to

actions


Software AgentsSoftware Agents

also referred to as “softbots” live in artificial environments where computers and

networks provide the infrastructure may be very complex with strong requirements on the

agent World Wide Web, real-time constraints,

natural and artificial environments may be merged user interaction sensors and effectors in the real world

camera, temperature, arms, wheels, etc.


Agent Program TypesAgent Program Types

different ways of achieving the mapping from percepts to actions

different levels of complexity

simple reflex agents agents that keep track of the world goal-based agents utility-based agents


Simple Reflex AgentsSimple Reflex Agents

instead of specifying individual mappings in an explicit table, common input-output associations are recorded requires processing of percepts to achieve some abstraction frequent method of specification is through condition-action

rules if percept then action

similar to innate reflexes or learned responses in humans efficient implementation, but limited power


Sensors

Effectors

What the world is like now

What should I do now

Condition-action rules

Agent

En

viro

nm

ent

Reflex Agent DiagramReflex Agent Diagram


Reflex Agent with Internal StateReflex Agent with Internal State

an internal state maintains important information from previous percepts sensors only provide a partial picture of the environment


Sensors

Effectors


State

How the world evolves

What my actions do

Agent

Environment

Condition-action rules


Agent with State DiagramAgent with State Diagram


Goal-Based AgentGoal-Based Agent

the agent tries to reach a desirable state may be provided from the outside (user, designer), or

inherent to the agent itself

results of possible actions are considered with respect to the goal may require search or planning

very flexible, but not very efficient


Sensors

Effectors


What happens if I do an action


State


What my actions do

Goals

Agent

Goal-Based Agent DiagramGoal-Based Agent Diagram


Utility-Based AgentUtility-Based Agent

more sophisticated distinction between different world states states are associated with a real number

may be interpreted as “degree of happiness” allows the resolution of conflicts between goals permits multiple goals


Sensors

Effectors


What happens if I do an action

How happy will I be then


State


What my actions do

Utility

Agent

Utility-Based Agent DiagramUtility-Based Agent Diagram


EnvironmentsEnvironments

determine to a large degree the interaction between the “outside world” and the agent the “outside world” is not necessarily the “real world” as we

perceive it

in many cases, environments are implemented within computers they may or may not have a close correspondence to the

“real world”


Environment PropertiesEnvironment Properties

accessible vs. inaccessible sensors provide all relevant information

deterministic vs. non-deterministic changes in the environment are predictable

episodic vs. non-episodic independent perceiving-acting episodes

static vs. dynamic no changes while the agent is “thinking”

discrete vs. continuous limited number of distinct percepts/actions


Agents SummaryAgents Summary

agents perceive and act in an environment ideal agents maximize their performance measure autonomous agents act independently

basic agent types simple reflex reflex with state goal-based utility-base

some environments may make life harder for agents inaccessible, non-deterministic, non-episodic, dynamic,

continuous


References

Basic literature


Recommended BooksRecommended Books

Artificial Intelligence : A Modern Approach by Stuart J. Russell, Peter Norvig

Logical Foundations of Artificial Intelligence by Michael R. Genesereth, Nils J. Nislsson, Nils J. Nilsson

Artificial Intelligence by Patrick Henry Winston Artificial Intelligence by Elaine Rich, Kevin Knight (good for

logic, knowledge representation, and search only)


General ReferenceGeneral Reference

Whatis.com (Computer Science Dictionary)http://whatis.com/search/whatisquery.html

Technology Encyclopediahttp://www.techweb.com/encyclopedia/

Computing Dictionaryhttp://wombat.doc.ic.ac.uk/

Webster Dictionaryhttp://work.ucsd.edu:5141/cgi-bin/http_webster

artificial intelligence a brief introduction

Documents