What is AI? - Homepages | The University of Aberdeen

Course Overview
What is AI?

Done
What are the Major Challenges?
 What are the Main Techniques?
 Where are we failing, and why?
 Step back and look at the Science
 Step back and look at the History of AI
 What are the Major Schools of Thought?
 What of the Future?
Done
Course Overview
 What is AI?
 What are the Major Challenges?
What are the Main Techniques? (How do we do it?)
 Where are we failing, and why?
 Step back and look at the Science
 Step back and look at the History of AI
 What are the Major Schools of Thought?
 What of the Future?
Course Overview
 What is AI?
 What are the Major Challenges?
What are the Main Techniques? (How do we do it?)
 Where are weSearch
failing, and why?
 Logics (knowledge representation and reasoning)
 Planning
and
acting
 Step back and
look at the
Science
 Bayesian belief networks
 Neural
 Step back and
look atnetworks
the History of AI
 Evolutionary computation
 Reinforcement
 What are the
Major Schools oflearning
Thought?
 Language parsing and speech techniques
Statistical methods (language, learning)
 What of the Future?
Course Overview
 What is AI?
 What are the Major Challenges?
What are the Main Techniques? (How do we do it?)
 Where are weSearch
failing, and why?
 Logics (knowledge representation and reasoning)
 Planning
and
acting
 Step back and
look at the
Science
 Bayesian belief networks
 Neural
 Step back and
look atnetworks
the History of AI
 Evolutionary computation
 Reinforcement
 What are the
Major Schools oflearning
Thought?
 Language parsing and speech techniques
Statistical methods (language, learning)
 What of the Future?
How to represent knowledge for the
computer?
 Natural language is ambiguous
–
–
–
–
We gave the monkeys the bananas because they were hungry.
We gave the monkeys the bananas because they were over-ripe.
Mary threw a rock at the window and broke it.
purgé la totalité de sa peine
 Need something precise for the computer
 Want to represent
– For communication
•
Meaning of natural language sentences
– For robot to reason about what to do
For Monkey-banana type problems
•
•
State of the world,
Effects of actions
– For Vision
•
Things that are seen, and known objects
What are out Requirements?
(for representing knowledge)
 Expressive
– John believes nobody likes brussels sprouts
 Reasoning ability
– Automatically deduce things
• John believes Mary doesn’t like brussels sprouts
– Don’t force us to represent loads of silly facts
• If we know Gordon Brown is the Prime Minister
Then we should know Joe Bloggs is not
• We shouldn’t need to say
–
–
–
–
Joe Bloggs is not the Prime Minister
His brother Jimmy Bloggs is not the Prime Minister
Their dog Spot is not the Prime Minister
…
– Same for properties of a student
• Tell the computer:
– All students take courses and learn new subjects
• If Mary is a student, we shouldn’t have to tell the computer
– Mary takes courses and learns new subjects
 Natural representation
Subject, Predicate, Object
 Subject
– The first spirit amazed Scrooge.
– Scrooge saw images of his past.
– The images caused a change in Scrooge's attitude.
 Predicate
– The first spirit amazed Scrooge.
– Scrooge saw images of his past.
– The images caused a change in Scrooge's attitude.
 Object
– The first spirit amazed Scrooge.
– Scrooge saw images of his past.
– The images caused a change in Scrooge's attitude.
 Objects and subjects are just things (nouns)
– Scrooge
– Spirit
– Images
Monkey
Chair
Brick
 Predicate describes relation among things (subject, object)
– amazed(Spirit,Scrooge)
– saw(Scrooge,images)
– caused(images,change)
Predicate Logic
 Objects and subjects are just things
– Monkey
– Chair
– Brick
 Predicate is a relationship or property
– Relationship like
• amazes(spirit,Scrooge)
• father(Vader,Luke)
• bigger(Glasgow,Aberdeen)
– Property like
• red(car)
• city(Aberdeen)
 Logic deals with:
– objects and relationships or properties
AND, OR, NOT, IF-THEN
 Some examples…
–
–
–
–
tall(Mary) AND beautiful(Mary)
sunny(day) OR rainy(day)
IF sunny(day) THEN go_for(Mary,walk)
IF rainy(day) THEN NOT (go_for(Mary,walk))
– IF rainy(day) THEN (go_for(Mary,study) OR go_for(Mary,cinema))
 Note “IF-THEN” does not necessarily mean “cause-effect”
– IF depressed(Mary) THEN rainy(day)
– Means that whenever she is depressed, it must be a rainy day
– She can never be depressed on a fine day
 Note “OR” is not exclusive
– IF (sweet(sauce) OR salty(sauce)) Then likes(Mary,sauce)
– “sweet(sauce) OR salty(sauce)”
• true if the sauce is both salty and sweet
– Logic is not natural language
• it has a precise (sometimes unnatural) meaning
Variables
Do not specify the exact value
– Just like x + x = 2x
Some examples…
– IF beautiful(Mary) THEN likes(John,Mary)
– IF beautiful(X) THEN likes(John,X)
– IF beautiful(X) THEN likes(Y,X)
– IF sunny(day) THEN go_for(Mary,Z)
– IF rainy(day) THEN NOT (go_for(Mary,Z))
– IF (rainy(day) AND student(S))
THEN (go_for(S,study) OR go_for(S,cinema))
FOR-ALL, THERE-EXISTS
 Some examples…
– All students study
• FOR-ALL S IF student(S) THEN study(S)
– Nobody likes brussels sprouts
• FOR-ALL X NOT likes(X, brussels_sprouts)
– Anybody who likes brussels sprouts is weird
• FOR-ALL X IF likes(X, brussels_sprouts) THEN weird(X)
– Somebody likes brussels sprouts
• THERE-EXISTS X likes(X, brussels_sprouts)
– Everybody has a mother
• FOR-ALL X THERE-EXISTS Y mother(y,x)
– Everybody is loved by somebody
• FOR-ALL X THERE-EXISTS Y loves(y,x)
– Everybody loves somebody
• FOR-ALL X THERE-EXISTS Y loves(x,y)
Reasoning
 Suppose we have
– FOR-ALL X IF p(X) THEN q(X)
 and we have
– p(X)
 then we could deduce
– q(X)
Try an example…
 Suppose
– FOR-ALL X IF tall(X) THEN strong(X)
 and suppose somebody tells us
– tall(john)
 What could we deduce?
 and what if somebody else tells us
– FOR-ALL X IF X strong(X) THEN loves(mary, X)
Ontology
 Can use logic to represent a hierarchy of concepts
–
–
–
–
–
–
–
isa(Tweety, canary)
isa(canary, bird)
isa(bird, animal)
isa(animal, living_thing)
isa(living_thing, physical_thing)
isa(physical_thing, tangible_thing)
isa(tangible_thing, thing)
 Lower ontologies / domain ontologies
– Computer hardware
– Medicine
 Upper ontologies
– http://www.cyc.com/cycdoc/img/UpperOntology.gif
– Several incompatible upper ontologies exist
Example Ontology
Ontology
 Can use logic to represent a hierarchy of concepts
–
–
–
–
–
–
–
isa(Tweety, canary)
isa(canary, bird)
isa(bird, animal)
isa(animal, living_thing)
isa(living_thing, physical_thing)
isa(physical_thing, tangible_thing)
isa(tangible_thing, thing)
 Lower ontologies / domain ontologies
– Computer hardware
– Medicine
 Upper ontologies
– http://www.cyc.com/cycdoc/img/UpperOntology.gif
– Several incompatible upper ontologies exist
Ontology
Represent a broad selection of objects and
relations
– Getting some common sense into the computer
 Domain ontologies – for a limited domain
 Cyc type projects – attempt to do all knowledge
 Also try to apply it to the Internet
– Make Web pages in Logic
– So that computer can understand and deduce things
• Example: for a travel itinerary
– “Semantic Web”
– Retreat from original goal of making computers
understand Natural Language
Logic Success/Failure?
 As with other areas
– Success for limited domains
• Medicine
• Engineering
– Still little progress towards general knowledge
 Reasoning
– Works well in a constrained domain
– Takes too long if large domain
– Hard to represent things like
•
•
•
•
•
Default/defeasible (i.e. usually/unless proved otherwise)
Time/events/changes
Beliefs, must be, could be, ought to be
Uncertainty – “IF X THEN probably Y”
These can be done… but reasoning becomes more difficult / slower
 Ultimate problem…
–
–
–
–
Too much to write down
Too many consequences to deduce/infer
What is relevant and not relevant?
common sense would tell you, but how to get commonsense in the first place?!
Course Overview
 What is AI?
 What are the Major Challenges?
What are the Main Techniques? (How do we do it?)
 Where are weSearch
failing, and why?
 Logics (knowledge representation and reasoning)
 Planning
and
acting
 Step back and
look at the
Science
 Bayesian belief networks
 Neural
 Step back and
look atnetworks
the History of AI
 Evolutionary computation
 Reinforcement
 What are the
Major Schools oflearning
Thought?
 Language parsing and speech techniques
Statistical methods (language, learning)
 What of the Future?