2 Lectures 2 4 AI Introduction
-
Upload
aainashuha-yusli -
Category
Documents
-
view
22 -
download
0
description
Transcript of 2 Lectures 2 4 AI Introduction
1Dr. Tarek Helmy, ICS-KFUPM
ICS-381Principles of Artificial Intelligence
Lectures 2- 4
Introducing Artificial Intelligence
Dr.Tarek Helmy El-Basuny
Dr. Tarek Helmy, ICS-KFUPM 2
Introduction to Artificial Intelligence
Brain StormingWhy do we study AI?What is Artificial Intelligence?
Characteristics of IntelligenceAI is a Multi-Disciplinary FieldCommonly Accepted Definitions of Artificial IntelligenceWhy Does Industry and the Government Care about AI?What might be involved in building a “smart” computer?Typical AI ProgramsFeatures of Using Artificial IntelligenceHow to Achieve AIAI Technologies and ApplicationsAI Brief HistoryCan a machine be truly “intelligent”?: Turing Test
Dr. Tarek Helmy, ICS-KFUPM 3
Why do we study AI?
Search enginesScience
Medicine/Diagnosis
LaborWhat else?Appliances
Dr. Tarek Helmy, ICS-KFUPM 4
Honda Humanoid Robot
Walk Turn
Stairshttp://world.honda.com/ASIMO/
Dr. Tarek Helmy, ICS-KFUPM 6
Brain Storming: What is Artificial Intelligence?
Good Question, but exactly, what is intelligence? Can we say, he is intelligent means:He knows a lotHe thinks fastHe talks muchHe learns quicklyHe memorizes well
Is it learned?Are you born with it?Can we use tests to measure it?
IQ Test!Intelligence = Knowledge + ability to perceive, feel, understand, process, communicate, judge, and learn.What is Artificial Intelligence?There is no official agreed upon definition of Artificial Intelligence.Why?
In practice, it is an “umbrella term”It is multidisciplinary subjectTechnologies enter and exit the AI “umbrella” regularly.
Dr. Tarek Helmy, ICS-KFUPM 7
Characteristics of Intelligence
Ability to Communicate
Creativity
Internal Knowledge
Ability to Learn
Perceive World Knowledge
Goal-Directed Behavior
Self Awareness
Dr. Tarek Helmy, ICS-KFUPM 8
An Intelligent Entity
Has understanding/intentionality
Exhibits behavior
SeeHearTouchTasteSmell
INPUTS INTERNAL PROCESSES
OUTPUTS
Senses environment
Can Reason
Has knowledge
Dr. Tarek Helmy, ICS-KFUPM 9
Commonly Accepted Definitions of Artificial Intelligence
Winston: “AI is the study of ideas which enable computers to do things which make people seem intelligent.”Steven Tanimoto, “Computational techniques for performing tasks that apparently require intelligence when performed by humans.”David Parnas, “Artificial intelligence is to artificial flowers as natural intelligence is to natural flowers.”Luger: The branch of computer science that is concerned with automation of intelligent behavior.Rich: “AI is the study of how to make computers do things which, at themoment, people do better.”
Fahlman: AI is the study of intelligence using the ideas and methods of computation.”
Found on the Web: AI is the reproduction of the methods or results of human reasoning or intuition.We can define it too: AI is a field of computer science that simulates human performance to make a computer reasons in a manner similar to humans.
Dr. Tarek Helmy, ICS-KFUPM 10
Why Do Industry and Government Care about AI?
AI shows promise for handling many complex problems, saving lives and resources:
• Solving information overload problems.
• Intelligent search engines
• Operating in risky environments.
• Robots
• Distributing scarce commercial knowledge.
• Data-mining software sort through massive databases, looking for patterns that would take a human years to spot.
• Enhancing training through use of simulation.
• ES
• Adaptive Computer Based Tests
Dr. Tarek Helmy, ICS-KFUPM 11
Main Areas of AI
Knowledge representation(including formal logic)Search, especially heuristic search (puzzles, games)PlanningReasoning under uncertainty, including probabilistic reasoningLearningAgent architecturesRobotics and perceptionNatural language processing
Search
Knowledgerep.Planning
Reasoning
Learning
Agent
RoboticsPerception
Naturallanguage Expert
Systems
Constraintsatisfaction
...
Dr. Tarek Helmy, ICS-KFUPM 12
A Hierarchical Model of Intelligence
Wisdom
Knowledge
Information
Data Context+
Vision+Experience+
Dr. Tarek Helmy, ICS-KFUPM 13
AI is a Multi-Disciplinary Field
ControlTheory
LinguisticsMathematics
Psychology
Artificial IntelligenceComputerScience
Philosophy
ComputerEngineering
Dr. Tarek Helmy, ICS-KFUPM 14
AI is a Multi-disciplinary
Many disciplines contribute to the goal of creating/modeling intelligent
entities:
Computer Engineering (Building fast computers)
Psychology (Perceive, process information, represent knowledge.)
Philosophy (Logic, methods of reasoning, mind as physical system,
foundations of learning, etc)
Linguistics (Structure and meaning of language)
Human Biology (How brain works)
Mathematics (Formal representation and proof, algorithms, etc.)
Control theory (Design systems that maximize an objective function
over time)
Dr. Tarek Helmy, ICS-KFUPM 15
Intelligent System Should do:
How can we make computer based systems more intelligent?
In practical terms, intelligent systems:
Should have the ability to automatically perform tasks that normally
require a human expert.
Should have more autonomy; less requirement for human intervention
or monitoring.
Should have Flexibility in dealing with variability in the environment
in an appropriate manner.
Are easier to use: able to understand what the user wants from limited
instructions.
Can improve their performance by learning from experience.
Dr. Tarek Helmy, ICS-KFUPM 16
Typical AI Programs
Intelligent entities (or “agents”) need to be able to do both “ordinary” and “expert” tasks:
Ordinary tasks - consider going shopping:Planning a route, and sequence of shops to visit!Recognizing (through vision) buses, people.Communicating (through natural language).Navigating round obstacles on the street, and manipulating objects for purchase.
Expert tasks are things like:Medical diagnosis.Equipment repair.
Dr. Tarek Helmy, ICS-KFUPM 17
How to Achieve AI?
How is AI research done?
AI research has both theoretical and experimental sides. The experimental side has both basic and applied aspects.
There are two main lines of research:One is biological, based on the idea that since humans are intelligent, AI should study humans and imitate their psychology or physiology. The other is phenomenal, based on studying and formalizing common sense facts about the world and the problems that the world presents to the achievement of goals.
The two approaches interact to some extent, and both should eventually succeed. It is a race, but both racers seem to be walking. [John McCarthy]
Dr. Tarek Helmy, ICS-KFUPM 18
What might be involved in building a “smart” computer?
What are the “components” that might be useful?Fast hardware?Foolproof [never makes error] software?Speech interaction?
Speech separation [segmentation/synthesis]Speech recognitionSpeech understanding
Image recognition and understanding?Learning?Planning and decision-making?
Dr. Tarek Helmy, ICS-KFUPM 19
Can we build hardware as complex as the brain?
How complicated is our brain?A neuron, or nerve cell, is the basic information processing unitEstimated to be on the order of 1012 neurons in a human brainMany more synapses (1014) connecting these neuronsCycle time: 10-3 seconds (1 millisecond)
How complex can we make computers?106 or more transistors per CPU Supercomputer: hundreds of CPUs, 109 bits of RAM Cycle times: order of 10- 8 seconds
ConclusionYES: we can have computers with as many basic processing elements as our brain, but with
Far fewer interconnections (wires or synapses) than the brainMuch faster updates than the brain
But building hardware is very different from making a computer behave like a brain!
Dr. Tarek Helmy, ICS-KFUPM 20
Must an Intelligent System be Foolproof?
A “foolproof” system is one that never makes an error:Types of possible computer errors
Hardware errors, e.g., memory errorsSoftware errors, e.g., coding bugs“Human-like” errors
Clearly, hardware and software errors are possible in practiceWhat about “human-like” errors?
An intelligent system can make errors and still be intelligentHumans are not right all of the timeWe learn and adapt from making mistakes
Conclusion:NO: intelligent systems will not (and need not) be foolproof
Dr. Tarek Helmy, ICS-KFUPM 21
Can Computers Talk with Understanding?
This is known as “speech synthesis”Translate text to phonetic form
e.g., “fictitious” -> fik-tish-esUse pronunciation rules to map phonemes to actual sound
e.g., “tish” -> sequence of basic audio soundsDifficulties
Sounds made by this “lookup” approach sound unnaturalSounds are not independent
e.g., “act” and “action”A harder problem is emphasis, emotion, etc
Humans understand what they are sayingMachines don’t: so they sound unnatural
Conclusion: NO, for complete understanding, but YES for pronouncing and translating.
Dr. Tarek Helmy, ICS-KFUPM 22
The End!!
Thank you
Any Questions?
Dr. Tarek Helmy, ICS-KFUPM 23
Can Computers Recognize Speech?
Speech Recognition:Mapping sounds from a microphone into a list of words.Hard problem: noise, more than one person talking, speech variability,.. Even if we recognize each word, we may not understand its meaning.
Recognizing single words from a small vocabularySystems can do this with high accuracy (order of 99%)e.g., directory inquiries for phone companies
• Limited vocabulary (area codes, city names)• Computer tries to recognize you first, if unsuccessful hands you over to a human
operator• Saves millions of dollars a year for the phone companies
Recognizing normal speech is much more difficultSpeech is continuous: where are the boundaries between words?Large vocabularies
Can be many thousands of possible wordsWe can use context to help figure out what someone said
Background noise, other speakers, accents, colds, etcOn normal speech, modern systems are only about 60% accurate
Conclusion: NO/with little accuracy, normal speech is too complex to accurately recognize, but YES for restricted problems
Dr. Tarek Helmy, ICS-KFUPM 24
Can Computers Understand speech?
Understanding is different to recognition:“Time flies like an arrow”
Assume the computer can recognize all the wordsBut how could it understand it?How could a computer figure this out?• Clearly humans use a lot of implicit common sense
knowledge in communication
Conclusion: NO with full semantic, much of what we say is beyond the capabilities of a computer to understand at present.
Dr. Tarek Helmy, ICS-KFUPM 25
Can Computers Learn and Adapt ?
Learning and AdaptationConsider a computer learning to drive on the freewayWe could code lots of rules about what to doWe could let it drive and steer it back of course when it heads for the embankment
Systems like this are under development.Machine learning allows computers to learn to do things without explicit programming.
Conclusion: YES, computers can learn and adapt, when presented with information in the appropriate way.
Dr. Tarek Helmy, ICS-KFUPM 26
Can Computers “see”?
Recognition v. Understanding (like Speech)Recognition and Understanding of Objects in a scene
Look around this roomYou can effortlessly recognize objectsHuman brain can map 2d visual image to 3d “map”
Why is visual recognition a hard problem?
Conclusion: Computers can partially “see” certain types of objects under limited circumstances: but YES/fully for certain constrained problems (e.g., face recognition).
Dr. Tarek Helmy, ICS-KFUPM 27
Can Computers Plan and Make Decisions?
IntelligenceInvolves solving problems and making decisions and planse.g., you want to visit your cousin in Mekah
You need to decide on dates, flightsYou need to get to the airport, etcInvolves a sequence of decisions, plans, and actions
What makes planning hard?The world is not predictable:
Your flight might be canceled or there will be a backup.There are a potentially huge number of details
Do you consider all flights? all dates?• No: common sense constrains your solutions
AI systems are only successful in constrained planning problems
Conclusion: NO, real-world planning and decision-making is still beyond the capabilities of modern computers. But YES for very well-defined, constrained problems.
Dr. Tarek Helmy, ICS-KFUPM 28
Intelligent Systems in Your Everyday Life
Post OfficeAutomatic address recognition and sorting of mail
BanksAutomatic check readers, signature verification systemsAutomated loan application classification
Telephone CompaniesAutomatic voice recognition for directory inquiriesAutomatic fraud detection,
Credit Card CompaniesAutomated fraud detection
Computer CompaniesAutomated diagnosis for help-desk applications
Dr. Tarek Helmy, ICS-KFUPM 29
AI Applications: Consumer Marketing
Have you ever used any kind of credit/ATM/store card while shopping?If so, you have very likely been “input” to an AI algorithm
All of this information is recorded digitally
Companies gather this information weekly and search for patternsGeneral changes in consumer behaviorTracking responses to new productsIdentifying customer segments: targeted marketing, e.g., they find out that consumers with sports cars who buy textbooks respond well to offers of new credit cards.Currently a very hot area in marketing
How do they do this?Algorithms (“data mining”) search data for patternsBased on mathematical theories of learningCompletely impractical to do manually
Dr. Tarek Helmy, ICS-KFUPM 30
AI Applications: Identification Technologies
ID cards e.g., ATM cardsCan be a security risk:
Cards can be lost, stolen, passwords forgotten, etc
Biometric IdentificationWalk up to a locked door
CameraFingerprint deviceMicrophone
Computer uses your biometric signature for identificationFace, eyes, fingerprints, voice pattern
Dr. Tarek Helmy, ICS-KFUPM 31
AI Applications: Predicting the Stock Market
The Prediction ProblemGiven the past, predict the futureVery difficult problem!We can use learning algorithms to learn a predictive model from historical data
Prob (increase at day t+1 | values at day t, t-1,t-2....,t-k)Such models are routinely used by banks and financial traders to manage portfolios worth millions of dollars
Time in days
?
?
Value ofthe Stock
Dr. Tarek Helmy, ICS-KFUPM 32
AI Brief History
1950: Alan Turing: Turing test
1950: Claude Shannon publishes a paper on chess playing. Shows that a game of chess involved about 10120 moves shows the need for heuristics
1943-56: McCulloch/Pitts: Research on the structure of the brain gives a model of neurons of the brain artificial neural networks
1951: von Neumann helps Minsky and Edmonds to build the first neural network computer.
1956: The first AI workshop sponsored by IBM Birth of AI
1958: McCarthy presents a paper “Program with common sense”.
1962: Rosenblatt proves the perception convergence theorem (learning algorithm)
Dr. Tarek Helmy, ICS-KFUPM 33
AI Brief History
1965: Lotfy Zadeh introduces Fuzzy sets
Early 70s: shift from a general purpose, knowledge-sparse, weak methods to domain-
specific, knowledge-intensive techniques (ES)
Mycin: rule-based expert system for diagnosis of infectious blood diseases
Mid 80s: use of neural networks for machine learning.
Generalization of single-layer network: Hopfield network, back-propagation.
Knowledge engineering: use of Fuzzy logic improves computational power,
improves cognitive modeling, allows to represent multiple experts
In 1995 The emergence of intelligent agents
Dr. Tarek Helmy, ICS-KFUPM 34
Can a machine be truly “intelligent”? : Turing’s Test
Alan Turing's 1950 article in Computing Machinery and Intelligence discussed conditions for considering a machine to be intelligentCan someone tell which is the machine, when communicating to human and to a machine in another room? If not, can we call the machine intelligent?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?
A B
Dr. Tarek Helmy, ICS-KFUPM 35
What would a computer need to pass the Turing test?
Passing Turing test requires the computer to have the following capabilities:1. NLP to communicate in English with the examiner2. Knowledge Representation to store information provided during the
test3. Automated reasoning to use stored information to answer questions
and draw conclusions. 4. Machine learning to adapt to new circumstances and to detect and
extrapolate patterns.5. Computer vision to recognize the examiner’s actions and various
objects presented by the examiner.5. Robotics to act upon objects as requested.
Dr. Tarek Helmy, ICS-KFUPM 36
AI Technologies
Previous, Today, Future
Cognitive-Based AI
(has to percept, learn and reason)Expert SystemsDecision Support SystemsNatural Language ProcessingIntelligent AgentsCollaborative Intelligent Agent NetworksFuzzy Logic
Biologically-Based AINeural NetsGenetic AlgorithmsSpeech RecognitionComputer VisionEvolutionary (changeable) ProgrammingMachine LearningRobotics
Dr. Tarek Helmy, ICS-KFUPM 37
The End!!
Thank you
Any Questions?