PowerPoint Presentation
Wrap-Up
Lesson Preview
Major topics revisited
Principles of KBAI revisited
Modern KBAI research
Final thoughts
Cognitive System
Reaction
Deliberation
Metacognition
Reasoning
Learning
Memory
Input
Output
Cognitive System
The World
Input
Output
Cognitive System
The World
Cognitive System
Cognitive System
Planning
Design & Creativity
Analogical Reasoning
Common Sense Reasoning
Metacognition
Learning
Fundamentals
Visuospatial Reasoning
Knowledge-Based Artificial Intelligence
Semantic Networks
Generate and Test
Problem Reduction
Means-Ends Analysis
Fundamentals
Production Systems
Planning
Logic
Planning
Common Sense Reasoning
Frames
Understanding
Scripts
Common Sense Reasoning
Learning by Recording Cases
Incremental Concept Learning
Version Spaces
Classification
Learning
Learning by Recording Cases
Case-Based Reasoning
Analogical Reasoning
Explanation-Based Learning
Analogical Reasoning
Visuospatial Reasoning
Constraint Propagation
Visuospatial Reasoning
Design & Creativity
Configuration
Diagnosis
Creativity
Design
Metacognition
Learning by Correcting Mistakes
Meta-Reasoning
Ethics in Artificial Intelligence
⛵
⛵
⛵
⛵
⛵
⛵
⛵
Principle #1: KBAI agents represent and organize knowledge into knowledge structures to guide and support reasoning.
kill. verb. /kil/
to cause the death of
to put an end to
kill1
agent :
victim :
kill2
agent :
object :
Sentence 1: A serious earthquake killed 25 people in Lower Slabovia.
Sentence 2: The President of Lower Slabovia killed 25 proposals for earthquake prediction.
Principle #1: KBAI agents represent and organize knowledge into knowledge structures to guide and support reasoning.
Action Frame
prim. : move-body-part
agent : Ashok
part : body
destination : sitting
Action Frame
prim. : see
agent : Andrew
object : Ashok
Action Frame
prim. : move-object
agent : Andrew
object : Andrew
destination : Ashok
Action Frame
prim. : move-object
agent : Andrew
object : menu
destination : Ashok
S = Ashok; P = Olive Garden; W = Andrew
Principle #1: KBAI agents represent and organize knowledge into knowledge structures to guide and support reasoning.
Object
Cup
is
Object
Bottom
has
Bottom
Flat
is
Object
Concavity
has
Object
Light
is
Object
Handle
has
Object
Porcelain
made-of
Object
Decoration
has
Object
Liquids
carries
Object
Liftable
is
Object
Drinking
enables
Object
Stable
is
Principle #1: KBAI agents represent and organize knowledge into knowledge structures to guide and support reasoning.
Given new problem a
Retrieve most similar prior problem, b, from memory
Apply b’s solution to problem a
b
a
a
b
Principle #2: Learning in KBAI agents is often incremental.
Route DestinationX DestinationY
A 10E 8N
B 1E 8N
C 10E 4N
D 2E 1N
X 8E 2N
Y 8E 2N
Z 1E 9N
Is the origin
North of 5N?
East of 5E?
East of 5E?
B
C
D
Yes
No
Yes
Yes
No
No
East of 3E?
Yes
No
Y
East of 2E?
X
A
Yes
No
Principle #2: Learning in KBAI agents is often incremental.
Current Concept
An Arch
Brick
Brick
Brick
Brick
supports
supports
left-of
supports
Brick
Brick
Brick
supports
supports
left-of
New Concept
Brick
Brick
Brick
supports
supports
left-of
Principle #2: Learning in KBAI agents is often incremental.
Sam’s
breakfast
Friday
cheap
[any]
[any]
[any]
[any]
Visit3
restaurant : Sam’s
meal : lunch
day : Saturday
cost : cheap
Specific
General
[any]
[any]
[any]
cheap
Sam’s
[any]
[any]
cheap
[any]
Breakfast
[any]
[any]
Sam’s
[any]
[any]
[any]
Principle #2: Learning in KBAI agents is often incremental.
Object
Cup
is
Object
Concavity
has
Object
Light
is
Object
Handle
has
Object
Liquids
carries
Object
Liftable
is
Object
Drinking
enables
Handle
Fixed
is
Object
Manipulable
is
Object
Orientable
is
Principle #2: Learning in KBAI agents is often incremental.
Angela ate lasagna with her dad last night at Olive Garden.
Ate
subject : Angela
object : lasagna
location : Olive Garden
time : night
utensils :
object-alive : false
object-is : in-subject
subject-mood : happy
Principle #3: Reasoning in KBAI agents is top-down as well as bottom-up.
kill. verb. /kil/
to cause the death of
to put an end to
kill1
agent :
victim :
kill2
agent :
object :
Sentence 1: A serious earthquake killed 25 people in Lower Slabovia.
Sentence 2: The President of Lower Slabovia killed 25 proposals for earthquake prediction.
Principle #3: Reasoning in KBAI agents is top-down as well as bottom-up.
Action Frame
prim. : move-body-part
agent : Ashok
part : body
destination : sitting
Action Frame
prim. : see
agent : Andrew
object : Ashok
Action Frame
prim. : move-object
agent : Andrew
object : Andrew
destination : Ashok
Action Frame
prim. : move-object
agent : Andrew
object : menu
destination : Ashok
S = Ashok; P = Olive Garden; W = Andrew
Principle #3: Reasoning in KBAI agents is top-down as well as bottom-up.
Y-Constraint
W-Constraint
L-Constraint
T-Constraint
blade
fold
fold
fold
blade
blade
fold
blade
blade
blade
fold
fold
fold
blade
blade
blade
fold
blade
blade
blade
Principle #3: Reasoning in KBAI agents is top-down as well as bottom-up.
Principle #3: Reasoning in KBAI agents is top-down as well as bottom-up.
Image credit:
https://commons.wikimedia.org/wiki/File:Chair_4a.jpg
28
Principle #4: KBAI agents match methods to tasks.
Methods
Generate & Test
Means-Ends Analysis
Problem Reduction
Production Systems
Case-Based Reasoning
Planning
Analogical Reasoning
Tasks
Configuration
Diagnosis
Design
Meta-Reasoning
Creativity
Classification
Systems Thinking
Brick
Brick
Brick
supports
supports
left-of
Current Concept
Brick
Brick
Brick
left-of
Not an Arch
Brick
Brick
Brick
must-
support
left-of
New Concept
must-
support
“require-link” heuristic
Principle #5: KBAI agents use heuristics to find solutions that are good enough, though not necessarily optimal.
A
C
B
A
C
B
Initial State
Goal State
A on Table
B on Table
C on A
Δ = 3
A on B
B on C
C on Table
Δ = 0
A
C
B
Δ = 3
A
C
B
Δ = 3
A
C
B
Δ = 2
Principle #5: KBAI agents use heuristics to find solutions that are good enough, though not necessarily optimal.
Principle #5: KBAI agents use heuristics to find solutions that are good enough, though not necessarily optimal.
?
What color is this block?
Block World
Blue
Orange
Purple
Black
Green
Red
Principle #6: KBAI agents make use of recurring patterns in the problems they solve.
Principle #6: KBAI agents make use of recurring patterns in the problems they solve.
Principle #6: KBAI agents make use of recurring patterns in the problems they solve.
Principle #6: KBAI agents make use of recurring patterns in the problems they solve.
Image credit:
https://commons.wikimedia.org/wiki/File:Chair_4a.jpg
36
Action Frame
prim. : move-body-part
agent : Ashok
part : body
destination : sitting
Action Frame
prim. : see
agent : Andrew
object : Ashok
Action Frame
prim. : move-object
agent : Andrew
object : Andrew
destination : Ashok
Action Frame
prim. : move-object
agent : Andrew
object : menu
destination : Ashok
S = Ashok; P = Olive Garden; W = Andrew
Principle #6: KBAI agents make use of recurring patterns in the problems they solve.
inning : 5th
portion : bottom
game : 131
weather : windy
runners : 1st, 3rd
outs : 1
batter : Pierzynski
average : .283
bats : left-handed
score : 1-4
goal : pitch
pitch : throw-fast-ball
result : homerun
(r8) If two operators selected and one has an episode with result homerun
then prefer other operator
“chunking”
Principle #7: The architecture of KBAI agents enables reasoning, learning, and memory to support and constrain each other.
Flies(animal)
∧
¬Bird(animal)
If Flies(animal) ∧ ¬Bird(animal):
Then Bat(Animal)
If an animal flies and is not a bird, it is a bat.
Principle #7: The architecture of KBAI agents enables reasoning, learning, and memory to support and constrain each other.
Object
Cup
is
Object
Bottom
has
Bottom
Flat
is
Object
Concavity
has
Object
Light
is
Object
Handle
has
Object
Porcelain
made-of
Object
Decoration
has
Object
Liquids
carries
Object
Liftable
is
Object
Drinking
enables
Object
Stable
is
Principle #7: The architecture of KBAI agents enables reasoning, learning, and memory to support and constrain each other.
Principle #7: The architecture of KBAI agents enables reasoning, learning, and memory to support and constrain each other.
Object
Cup
is
Object
Concavity
has
Object
Light
is
Object
Handle
has
Object
Liquids
carries
Object
Liftable
is
Object
Drinking
enables
Handle
Fixed
is
Object
Manipulable
is
Object
Orientable
is
CALO:
“Cognitive Assistant that Learns and Organizes”
Cyc and OMCS: Knowledgebases of everyday common sense knowledge
Wolfram Alpha:
A computational knowledge and answer engine.
VITA:
A computational model of visual thinking in autism, based on RPM.
Dramatis:
A computational model of suspense and drama in stories.
DANE:
Support for design based on analogies to natural systems.
What are some things Watson must be able to do to participate in Jeopardy? Write each on its own line below.
Read the clue
Search its knowledge base
Decide on an answer
Properly phrase that answer
Fundamental Conundrums of Artificial Intelligence
Intelligent agents have limited resources.
Computation is local, but problems have global constraints.
Logic is deductive, but many problems are not.
The world is dynamic, but knowledge is limited.
Problem solving, reasoning, and learning are complex, but explanation and justification are even more complex.
Characteristics of AI Problems
Knowledge often arrives incrementally.
Problems exhibit recurring patterns.
Problems have multiple levels of granularity.
Many problems are computationally intractable.
The world is dynamic, but knowledge of the world is static.
The world is open-ended, but knowledge is limited.
Characteristics of AI Agents
Agents have limited computing power.
Agents have limited sensors.
Agents have limited attention.
Computational logic is fundamentally deductive.
AI agents’ knowledge is incomplete relative to the world.
Deliberation
Reasoning
Learning
Memory
thinking
acting
optimally
like humans
Agents that think rationally
Agents that think like humans
Agents that act like humans
Agents that act rationally
machine learning
airplane autopilot
semantic web
improvisational robots
thinking
acting
optimally
like humans
thinking
acting
optimally
like humans
1
3
2
4
1. Roomba (automated vacuum)
2. C-3PO (Star Wars)
3. Siri (Apple virtual assistant)
4. Google Maps (route navigation)
What are cognitive systems?
Cognitive: dealing with human-like intelligence.
Systems: multiple interacting components such as learning, reasoning, and memory.
Cognitive Systems: Systems that exhibit human-like intelligence through processes like learning, reasoning, and memory.
Cognitive System
The World
Input
Output
Cognitive System
The World
Cognitive System
Cognitive System
Cognitive System
Reaction
Deliberation
Metacognition
Reasoning
Learning
Memory
Input
Output
Planning
Design & Creativity
Analogical Reasoning
Common Sense Reasoning
Metacognition
Learning
Fundamentals
Visuospatial Reasoning
Knowledge-Based Artificial Intelligence
Semantic Networks
Generate and Test
Problem Reduction
Means-Ends Analysis
Fundamentals
Production Systems
Planning
Logic
Planning
Common Sense Reasoning
Frames
Understanding
Scripts
Common Sense Reasoning
Learning by Recording Cases
Incremental Concept Learning
Version Spaces
Classification
Learning
Learning by Recording Cases
Case-Based Reasoning
Analogical Reasoning
Explanation-Based Learning
Analogical Reasoning
Visuospatial Reasoning
Constraint Propagation
Visuospatial Reasoning
Design & Creativity
Configuration
Diagnosis
Creativity
Design
Metacognition
Learning by Correcting Mistakes
Meta-Reasoning
Ethics in Artificial Intelligence
Knowledge-Based Artificial Intelligence
Logic
Planning
Frames
Understanding
Scripts
Common Sense Reasoning
Constraint Propagation
Visuospatial Reasoning
Configuration
Diagnosis
Computational Creativity
Design
Learning by Correcting Mistakes
Meta-Reasoning
Ethics in Artificial Intelligence
Semantic Networks
Generate and Test
Problem Reduction
Means-End Analysis
Production Systems
Learning by Recording Cases
Incremental Concept Learning
Scripts
Classification
Learning by Recording Cases
Case-Based Reasoning
Analogical Reasoning
Explanation-Based Reasoning
To recap…
Conundrums and characteristics
Four schools of AI
What is KBAI?
Cognitive Systems
Topics in AI
Introduction to CS7637
Lesson Preview
Class goals, outcomes, and strategies
Class projects and assessments
Computational psychometrics
Raven’s Progressive Matrices
Principles of CS7637
You will learn…
Core methods of knowledge-based AI
Tasks addressed by knowledge-based AI
How knowledge-based AI agents use these methods to address these tasks
The relationship between AI and human cognition
You will be able to…
Design and implement a knowledge-based AI agent
Use these strategies and agents to address complex, practical problems
Use the design and results of these agents to reflect on human cognition
You will learn by…
Learning by Example
Learning by Doing
Project-Based Learning
Personalization
Learning by Reflection
You will complete…
Projects
Short assignments
Tests
Exercises
Discussions
Raven’s Progressive Matrices
Test written in the 1930s to examine general intelligence.
Consists of 60 multiple-choice visual analogy problems.
Unique in that problems are strictly visual.
Widespread usage as a valid test for intelligence.
Problems
2×1 matrix problems
2×2 matrix problems
3×3 matrix problems
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
#
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
?
A
B
C
B
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
D
F
E
G
#
H
?
A
C
B
1
2
3
4
5
6
ο
ο
ο
ο
ο
ο
D
F
E
G
#
H
5. KBAI agents use heuristics to find solutions that are good enough, though not necessarily optimal.
6. KBAI agents make use of recurring patterns in the problems they solve.
7. The architecture of KBAI agents enables reasoning, learning, and memory to support and constrain each other.
1. KBAI agents represent and organize knowledge into knowledge structures to guide and support reasoning.
2. Learning in KBAI agents is often incremental.
3. Reasoning in KBAI agents is top-down as well as bottom-up.
4. KBAI agents match methods to tasks.
Principles of CS7637
Frequently-used readings
Artificial Intelligence by Patrick Winston
Knowledge Systems by Mark Stefik
Artificial Intelligence by Elaine Rich and Kevin Knight
Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig
To recap…
Class goals, outcomes, and strategies
The Project: Raven’s Progressive Matrices
Principles of CS7637
/docProps/thumbnail.jpeg