Artificial Intelligence

Introduction
AI in the movies
Definition of AI
Intelligence: The ability to learn and solve problems
Websters Dictionary.
Definition of AI
Intelligence: The ability to learn and solve problems
Websters Dictionary.

Artificial intelligence (AI) is the intelligence exhibited

by machines or software
Wikipedia.
Definition of AI
Intelligence: The ability to learn and solve problems
Websters Dictionary.

Artificial intelligence (AI) is the intelligence exhibited

by machines or software
Wikipedia.

The science and engineering of making intelligent

machines
McCarthy.
Definition of AI
Intelligence: The ability to learn and solve problems
Websters Dictionary.

Artificial intelligence (AI) is the intelligence exhibited

by machines or software
Wikipedia.

The science and engineering of making intelligent

machines
McCarthy.

The study and design of intelligent agents, where an

intelligent agent is a system that perceives its environment
and takes actions that maximize its chances of success.
Russel and Norvig AI book.
Why AI?

Just as the Industrial Revolution freed up a lot of humanity

from physical drudgery, I think AI has the potential to free
up humanity from a lot of the mental drudgery.
Andrew Ng.
What is AI?
Four schools of thoughts (Russel & Norvig)

Thinking humanly Thinking rationally

The exciting new eort to make The study of mental faculties

computers think... machines with through the use of computational
minds, in the full and literal sense. models. (Charniak and McDermott,
(Haugeland, 1985) 1985

Acting humanly Acting rationally

The study of how to make com- Computational Intelligence is the

puters do things which, at the mo- study of the design of intelligent
ment, people are better. (Rich and agents. (Poole et al., 1998)
Knight, 1991)
What is AI?
Thinking humanly: cognitive approach

Requires to determine how humans think!

1960s cognitive revolution.
Requires scientific theories of internal activities of the brain
What level of abstraction? Knowledge or circuits?
How to validate?

Today, Cognitive Science and Artificial Intelligence are

distinct disciplines.
What is AI?
Acting humanly:

Turing test (Alan Turing 1950): A computer passes the

test of intelligence, if it can fool a human interrogator.

Credit: From Russel and Norvig slides.

Major components of AI: knowledge, reasoning, language,

understanding, learning.
What is AI?
Acting humanly:
What is AI?
Thinking rationally: Laws of thoughts.

Codify right thinking with logic.

Several Greek schools developed various forms of logic:notation

and rules of derivation for thoughts.

Problems:

1. Not all knowledge can be expressed with logical notations.

2. Computational blow up.
What is AI?
Acting rationally:

The right thing: that which is expected to maximize goal

achievement, given the available information.

A rational agent is one that acts so as to achieve the best

outcome, or when there is uncertainty, the best expected out-
come.

Aristotle (Nicomachean Ethics):

Every art and every inquiry, and similarly every action and
pursuit, is thought to aim at some good.
What is AI?
Four schools of thoughts (Russel & Norvig)

Thinking humanly Thinking rationally

The exciting new eort to make The study of mental faculties

computers think... machines with through the use of computational
minds, in the full and literal sense. models. (Charniak and McDermott,
(Haugeland, 1985) 1985

Acting humanly Acting rationally: Our approach

The study of how to make com- Computational Intelligence is the

puters do things which, at the mo- study of the design of intelligent
ment, people are better. (Rich and agents. (Poole et al., 1998)
Knight, 1991)
Applications of AI
Applications of AI
Speech recognition

Virtual assistants: Siri (Apple),

Echo (Amazon), Google Now, Cor-
tana (Microsoft).

They helps get things done: send

an email, make an appointment,
find a restaurant, tell you the
weather and more.

Leverage deep neural networks to

handle speech recognition and
natural language understanding.
Applications of AI
Handwriting recognition (check, zipcode)
Applications of AI
Machine translation

Historical motivation: translate Russian to English.

First systems using mechanical translation (one-to-one cor-

respondence) failed!

Out of sight, out of mind ) Invisible, imbecile.

Applications of AI
Machine translation

Historical motivation: translate Russian to English.

First systems using mechanical translation (one-to-one cor-

respondence) failed!

Out of sight, out of mind ) Invisible, imbecile.

Oops!
Applications of AI
Machine translation

MT has gone through ups and downs.

Today, Statistical Machine Translation leverages the vast

amounts of available translated corpuses.

While there is room for improvement, machine translation has

made significant progress.
Applications of AI
Machine translation

100+ languages
Applications of AI
Machine translation
Applications of AI
Robotics: Awesome robots today! NAO, ASIMO, and more!

Credit: By Momotarou2012, via Wikimedia Commons.

Applications of AI
Recommendation systems (collaborative filtering)
Applications of AI
Search engines
Applications of AI
Email
Applications of AI
Face detection

Viola-Jones method.
Applications of AI
Face detection

Viola-Jones method.
Applications of AI
Face detection

Viola-Jones method.
Applications of AI
Face recognition
Applications of AI
Detection of breast cancer in mammography images
Applications of AI
Chess (1997): Kasparov vs. IBM Deep Blue

(Left) Copyright 2007, S.M.S.I., Inc. - Owen Williams, The Kasparov Agency, via Wikimedia

Commons (Right) By James the photographer, via Wikimedia Commons

Powerful search algorithms!

Applications of AI
Jeopardy! (2011): Humans vs. IBM Watson

By Rosemaryetoufee (Own work), via Wikimedia Commons

Natural Language Understanding and information extraction!

Applications of AI
Go (2016): Lee Sedol versus Google AlphaGo

(Left) By LG Electronics, via Wikimedia Commons (Right) By Google DeepMind, via

Wikimedia Commons

Deep Learning, reinforcement learning, and search algorithms!

Applications of AI
Autonomous driving

By User Spaceape on en.wikipedia, via Wikimedia Commons

DARPA Grand Challenge

2005: 132 miles
2007: Urban challenge
2009: Google self-driving car
State-of-the-art applications
Speech recognition Fraud detection
Autonomous planning and Recommendation systems
scheduling Web search engines
Financial forecasting Autonomous cars
Game playing, video games Energy optimization
Spam fighting Question answering systems
Logistics planning Social network analysis
Robotics (household, surgery, Medical diagnosis, imaging
navigation) Route finding
Machine translation Traveling salesperson
Information extraction Protein design
VLSI layout Document summarization
Automatic assembly Transportation/scheduling
Sentiment analysis Computer animation
State-of-the-art applications
Speech recognition Fraud detection
Autonomous planning and Recommendation systems
scheduling Web search engines
Financial forecasting Autonomous cars
Game playing, video games Energy optimization
Spam fighting Question answering systems
Logistics planning Social network analysis
Robotics (household, surgery, Medical diagnosis, imaging
navigation) Route finding
Machine translation Traveling salesperson
Information extraction Protein design
VLSI layout Document summarization
Automatic assembly Transportation/scheduling
Sentiment analysis Computer animation
Many more!
Foundation of AI
Foundation of AI
Philosophy
Logic, methods of reasoning.
Mind as physical system that operates as a set of rules.
Foundations of learning, language, rationality.

Mathematics
Logic: Formal representation and proof.
Computation, algorithms.
Probability.

Economics
Formal theory of rational decisions.
Combined decision theory and probability theory for decision
making under uncertainty.
Game theory.
Markov decision processes.
Foundation of AI
Neuroscience
Study of brain functioning.
How brains and computers are (dis)similar.

Psychology
How do we think and act?
Cognitive psychology perceives the brain as an information
processing machine.
Led to the development of the field cognitive science: how
could computer models be used to study language, memory,
and thinking from a psychological perspective.

Computer engineering
Cares about how to build powerful machines to make AI
possible.
E.g., Self-driving cars are possible today thanks to advances
in computer engineering.
Foundation of AI
Control theory and cybernetics
Design simple optimal agents receiving feedback from the
environment.
Modern control theory design systems that maximize an
objective function over time.

Linguistics
How are language and thinking related.
Modern linguistics + AI = Computational linguistics (Nat-
ural language processing).
AI founders
Aristotle
Alan Turing
John Mc Carthy
Warren McCulloh
Walter Pitts
Claude Shannon
Marvin Minsky
Dean Edmonds
Herbert Simon
Allen Newell
David Waltz
Tom Mitchell
Stuart J. Russell
Peter Norvig
etc.
AI Resources
Major journals/conferences: JAIR, TPAMI, JMLR, IJCAI,
AAAI, IAAI, CVPR, ECAI, ICML, NIPS, etc.

Video lectures:

http://videolectures.net/Top/Computer_Science/Artificial_Intelligence/
History of AI
1940-1950: Gestation of AI
McCulloch & Pitts: Boolean circuit to model of brain
Turings Computing Machinery and Intelligence
http://www.turingarchive.org/browse.php/B/9

1950-1970: Early enthusiasm, great expectations

Early AI programs, Samuels checkers program
Birth of AI @ Dartmouth meeting 1956.
Check out the MIT video The thinking Machine on
youtube
https://www.youtube.com/watch?v=aygSMgK3BEM

1970-1990: Knowledge-based AI
Expert systems, AI becomes an industry
AI winter
History of AI
1990-present: Scientific approaches

Neural Networks: le retour

The emergence of intelligent agents

AI becomes scientific, use of probability to model uncer-

tainty

AI Spring!

The availability of very large datasets.

Data will drive future discoveries and alleviate the com-
plexity in AI.
Course logistics
Course level: Masters challenging!

Prerequisites: You are required to have some knowledge of

programming and an understanding of probability. Python is
the programming language in this course.

Assignments: There will be two kinds of assignments:

Quizzes (conceptual): Test your understanding of the

lectures. Please read the questions very carefully.

Projects (programming): The course oers an excellent

opportunity for students to dive into Python while solving
AI problems and learning its applications.
Course logistics
Suggested readings:

We recommend this book, which is the main reference in

the field:
Artificial Intelligence, A Modern Approach. Stuart Russell
and Peter Norvig. Third Edition. Pearson Education.

http://aima.cs.berkeley.edu/

Check out the list of readings, useful links we suggest for

this course.
What you will learn
Introduction to artificial intelligence, history of Artificial In-
telligence.

Building intelligent agents (search, games, logic, constraint

satisfaction problems).

Machine Learning algorithms.

Applications of AI (Natural Language Processing, Robotics,

and Vision).

Solving real AI problems through programming Python.

Course roadmap
1. Rational intelligent agents

2. Search agents (uninformed search, informed search)

3. Adversarial search/games

4. Machine Learning (ML)

5. Constraint satisfaction problems (CSPs)

6. Logic (propositional logic, first order logic)

7. Markov Decision Processes (MDPs) and Reinforcement Learn-

ing (RL)

8. Application to Natural language Processing (NLP)

9. Application to vision and robotics

Rational intelligent agents
This course is about designing intelligent agents.

An agent perceives the environment and act upon that envi-

ronment to achieve some task.

An agent is function from percepts to actions.

We care specifically about rational agents.

Rationality is relative to how to act to maximize a perfor-

mance measure.

AI aims to design the best agents (programs) that achieve the

best performance given the computational limitations.

Agent = Architecture + Program

Search agents
Agents that work towards a goal.

Agents consider the impact of actions on future states.

Agents job is to identify the action or series of actions that

lead to the goal.

Paths come with dierent costs and depths.

Two kinds of search:

Uninformed Search (use no domain knowledge): BFS,
DFS, UCS, etc.

Informed Search (use heuristic to reach the goal faster):

Greedy search, A, etc.
Search agents

The 8-queen problem: on a chess board, place 8 queens so that no

queen is attacking any other horizontally, vertically or diagonally.
Search agents
Start: Las Vegas
Goal: Calgary

Explore + Execute
Adversarial Search: games
Solved games!
Checkers:
Chinook ended 40-year-reign of human world champion Marion
Tinsley in 1994.

Used an endgame database defining perfect play for all po-

sitions involving 8 or fewer pieces on the board, a total of
443,748,401,247 positions.
Adversarial search: games
Adversarial search problems game

Adversarial There is an opponent we cant control!

Game vs. search: optimal solution is not a sequence of actions

but a strategy (policy). If opponent does a, agent does b, else
if opponent does c, agent does d, etc.

Tedious and fragile if hard-coded (i.e., implemented with

rules).

Concepts/methods: Minimax algorithm, pruning,

stochastic games.
Machine learning

How do we create computer programs that improve

with experience?
Tom Mitchell
Machine learning
Binary classification (categorization)
Input: Examples with labels.
(x1, y1), . . . , (xn, yn) / xi 2 X Rn, yi 2 Y = { 1, +1}

Output: h : X ! Y
Example: Approve credit yes/no, spam/ham.

!"#$%&"'('

!"#$%&"')'
*"+,-,./'0.%/1#&2'

Concepts/methods: Supervised learning, classification, K near-

est neighbors, perceptrons, neural networks, linear regression, etc.
Machine learning
Data segmentation or Clustering
Input: Examples without labels.
x1, . . . , xn, xi 2 X Rn

Output: f : X ! {C1, . . . Ck } (set of clusters).

Feature'2'

Feature'1'

Concepts/methods: Unsupervised learning, clustering, k-means,

association rules, etc.
Constraint satisfaction
A search problem too!

We dont care about the path but about the goal itself.

All paths are of same depth.

Problem is formulated using variables, domains and con-

straints.

Solving the CSP: finding the assignment(s) that satisfy all

constraints.

Concepts/methods: problem formalization, backtracking

search, arc consistency, etc.
Constraint satisfaction

Variables: Xl,c for 1 l 9 and 1 c 9.

Constraints: All 3x3 grid, row, column, must contain digits 1..9
and all of them!
Solution: Find the assignments to the variables that satisfy the
constraints.
Constraint satisfaction

Variables: Xl,c for 1 l 9 and 1 c 9.

Constraints: All 3x3 grid, row, column, must contain digits 1..9
and all of them!
Solution: Find the assignments to the variables that satisfy the
constraints.
Logical Agents
Logic can be used by an agent to model the world.

Sentences in PL and FOL have a fixed syntax.

With symbols and connectives we can form logical sentences:

Example: hot ^ sunny ) beach _ pool

Syntax and Semantic represent two important and distinct

aspects in logic.

Inference: Given a Knowledge Base (KB) (set of sentences in

logic), given a query , output whether KB entails , noted:
KB |=

Concepts/methods: Modus Ponens, sound and complete in-

ference, horn clauses, etc.
Reinforcement learning
Agent evolves in a stochastic and uncertain environment.

Agent learns from reinforcement or delayed reward.

Learning approaches for decision making in situations where

outcomes are stochastic.

Agent continues to plan and learn to aect its environment.

Reinforcement learning agents are driven by maximizing their

rewards on the long run.
Applications
Natural Language Processing (NLP): concerned with the
interactions between computers and human languages.

Vision/perception: concerned with image processing and

building computer vision agents. Goals: information extrac-
tion for tasks such as manipulation, navigation, and object
recognition.

Robotics: concerned with intelligent agents that manipulate

the physical world. Dierent aspects: planning of robot mo-
tion, vision and object recognition.
Historical moment today

In memory of Alan Turing (1912-1954)

Famous British mathematician.

Code breaker during World War II.
Proposed an operational test for intelligent behavior: The Im-
itation Game.
In Computing machinery and intelligence (1950), he laid
down AI major components:
(language, reasoning, knowledge, learning, understanding).
http://www.turingarchive.org/browse.php/B/9
Summary
AI is a hard (computational complexity, language, vision, etc),
and a broad field with high impact on humanity and society.

What can AI do for us is already amazing!

AI systems do not have to model human/nature but can act

like or be inspired by human/nature.

How human think is beyond the scope of this course.

Rational (do the right thing) agents are central to our approach
of AI.

Note that rationality is not always possible in complicated en-

vironment but we will still aim to build rational agents.
Summary
AI may be perceived as a scary area! Is AI a threat to our
humankind?

Professor Stephen Hawking, eminent scientist told BBC:

The development of full artificial intelligence could spell

the end of the human race.

AI is a flourishing and exciting field: everyone can contribute.

Looking forward for an exciting journey together!

Credit
Artificial Intelligence, A Modern Approach. Stuart Russell and
Peter Norvig. Third Edition. Pearson Education.

http://aima.cs.berkeley.edu/