10/1/24

BIM309 Artificial Intelligence

Week 2 – Introduction to AI

Outline
§ Overview of AI
§ Applications of AI
§ AI Foundations and History

1
 10/1/24

AI in the Movies

Definition of AI
“Intelligence: The ability to learn and solve problems”
Webster’s 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

2
 10/1/24

What is AI?

Why AI?

“If we can make computers more intelligent and understand the world and the
environment better, it can make life so much better for many of us. 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

3
 10/1/24

What is AI?
Measure success in Measure success against
terms of fidelity to an ideal concept of
human performance intelligence, rationality

Thinking Humanly 1 Thinking Rationally 3

concerned “The exciting new effort to make computers think … “The study of mental faculties through the use of
with machines with minds, in the full and literal sense.” computational models.” (Charniak and McDermott,
Thought
processes (Haugeland, 1985) 1985)
and
reasoning
“[The automation of] activities that we associate with “The study of the computations that make it possible
human thinking, activities such as decision-making, to perceive, reason, and act.” (Winston, 1992)
problem solving, learning ...” (Bellman, 1978)

Acting Humanly 2 Acting Rationally 4

“The art of creating machines that perform functions “Computational Intelligence is the study of the design
that require intelligence when performed by people.” of intelligent agents.” (Poole et al., 1998)
address
Behaviour
(Kurzweil, 1990)

“The study of how to make computers do things which, “AI … is concerned with intelligent behavior in
at the moment, people are better.” (Rich and Knight, artifacts.” (Nilsson, 1998)
1991)

Four schools of thoughts (Russel & Norvig)

7

What is AI?
1. Thinking Humanly: Cognitive Approach
machines with minds

§ Requires to determine how humans think!

§ 1960’s “cognitive revolution”
§ Requires scientific theories of internal activities of the brain
§ What level of abstraction?
§ Is it “Knowledge” or “circuits”?
§ How to validate the knowledge?

Today, Cognitive Science and Artificial Intelligence are distinct

disciplines.

4
 10/1/24

What is AI?
2. Acting Humanly:
machines do things just like humans

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

intelligence, if it can fool a human interrogator
Can machines think?
Imitation Game
Problem:
Turing test is not
reproducible or amenable
to mathematical analysis

§ Major components of AI: knowledge, reasoning, language,

understanding, learning Turing, Alan M. "Computing machinery and
intelligence." Mind 59.236 (1950): 433-460.

What is AI?
3. Thinking Rationally: Law of thoughts
involves using math and logic to do AI
§ 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
3. Logical systems tend to do the wrong thing in the presence of uncertainty

10

5
 10/1/24

What is AI?
4. Acting Rationally: Rational Agents
machines to act rationally rather than humanly
§ Rational behavior: doing the “right thing”
§ The right thing: which is expected to maximize goal achievement, given the available information about the environment
§ Doesn’t necessarily involve thinking, e.g., blinking reflex
§ Thinking should be in the service of rational action
§ Entirely dependent on goals!
§ Rational ≠ successful OR Irrational ≠ insane, irrationality is sub-optimal action
§ Our focus here: Rational Agents
§ A rational agent is one that acts so as to achieve the best outcome, or when there is uncertainty, the best expected outcome
§ Systems which make the best possible decisions given goals, evidence, and constraints
§ In the real world, usually lots of uncertainty … and lots of complexity
§ Usually, we’re just approximating rationality

§ Aristotle (Nicomachean Ethics):

“Every art and every inquiry, and similarly every action and pursuit, is thought to aim at some good.”

11

What is AI?
Thinking Humanly 1 Thinking Rationally 3
“The exciting new effort to make computers think … “The study of mental faculties through the use of
machines with minds, in the full and literal sense.” computational models.” (Charniak and McDermott,
(Haugeland, 1985) 1985)

“[The automation of] activities that we associate with “The study of the computations that make it possible
human thinking, activities such as decision-making, to perceive, reason, and act.” (Winston, 1992)
problem solving, learning ...” (Bellman, 1978)

Acting Humanly 2 Acting Rationally à OUR APPROACH 4

“The art of creating machines that perform functions “Computational Intelligence is the study of the design
that require intelligence when performed by people.” of intelligent agents.” (Poole et al., 1998)
(Kurzweil, 1990)

“The study of how to make computers do things which, “AI … is concerned with intelligent behavior in
at the moment, people are better.” (Rich and Knight, artifacts.” (Nilsson, 1998)
1991)

Four schools of thoughts (Russel & Norvig)

12

6
 10/1/24

What is AI?
Thinking Humanly 1 Thinking Rationally 3
“The exciting new effort to make computers think … “The study of mental faculties through the use of
machines with minds, in the full and literal sense.” computational models.” (Charniak and McDermott,
(Haugeland, 1985) machines with minds 1985) involves using math and logic to do AI
Cognitive Approach Law of thoughts
“[The automation of] activities that we associate with “The study of the computations that make it possible
human thinking, activities such as decision-making, to perceive, reason, and act.” (Winston, 1992)
problem solving, learning ...” (Bellman, 1978) codifying the right thinking with logic

Acting Humanly 2 Acting Rationally à OUR APPROACH 4

“The art of creating machines that perform functions “Computational Intelligence is the study of the design
that require intelligence when performed by people.” of intelligent agents.” (Poole et al., 1998)
(Kurzweil, 1990)
machines do things just like humans machines to act rationally rather than humanly
“The study of how to make computers do things which, “AI … is concerned with intelligent behavior in
at the moment, people are better.” (Rich and Knight, artifacts.” (Nilsson, 1998)
1991)

Four schools of thoughts (Russel & Norvig)

13

AI Today: In Vitro vs. Real-World

§ One of the biggest changes in AI is the transition from the lab to the
real-world
§ AI was limited to artificial enviroments and datasets
§ Spur the development of new methods
§ Now much more real-world deployment of AI in ways that have a
direct impact on people's lives

15

7
 10/1/24

Applications of AI

Speech Recognition & Question Answering

§ Virtual assistants: Siri (Apple), Alexa (Amazon),

Google Assistant, Cortana (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

16

Applications of AI
Handwriting recognition (check, zipcode)

17

8
 10/1/24

Applications of AI
Machine Translation
§ Historical motivation: Translate Russian to English
§ First systems using mechanical translation (one-to-one
correspondence) failed?
§ “Out of sight, out of mind” à “Invisible, imbecile”.
Oops!
§ 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

18

Applications of AI
Machine Translation

Voice & Text Translator

https://www.skype.com/en/features/skype-translator/

19

9
 10/1/24

Applications of AI
Robotics
Awesome robots today!

RoboCup - https://www.youtube.com/watch?v=_PC-V5GJP6Q

21

Applications
of AI

22

10
 10/1/24

Applications of AI
Sentiment Analysis

Input: movie review

Shows moments of promise
but ultimately succumbs to
clichés and pat storytelling.

Output: sentiment
POSITIVE or NEGATIVE

23

Applications
of AI

AI Powered Search

24

11
 10/1/24

Applications of AI
Email

25

Applications of AI
Face Detection

Viola-Jones method

26

12
 10/1/24

Applications of AI
Face Recognition

27

Can you tell the difference between a

real face and an AI-generated fake?
http://whichfaceisreal.com/

28

13
 10/1/24

NVIDIA AI DEMO!

https://www.nvidia.com/en-us/research/ai-playground/

NVIDIA's GauGAN AI has

learned to conversion the
drawings into a chart!

29

Applications of AI
Computer Vision

YOLO v8 (2023)

30

14
 10/1/24

Applications of AI
Text-to-Image

Open AI (2023) Dall-e-3

31

AI artwork?
§ Jason M. Allen created an award-
winning AI artwork using Midjourney,
an artificial intelligence technology that
turns lines of text into hyper-realistic
pictures.
§ Mr. Allen’s piece, “Théâtre D’opéra
Spatial,” won the fair’s new digital
artists category, making it one of the
first AI-generated works to do so.
§ This spurred intense debates on the
ethics of AI-generated art, as well as
criticism from others who claim these
apps are just a high-tech version of
plagiarism.
§ Allen declined to provide the exact
written command he gave Midjourney
to create “Théâtre D’opéra Spatial.”

“Théâtre D’opéra Spatial” by Jason M. Allen via Midjourney

32

15
 10/1/24

Applications of AI
Visual Assistive Technology

§ Seeing AI, the free Microsoft app for iOS for

those who are blind or who have low vision
§ Read printed text, currency, and describe
physical objects, product labels and colors,
among other things
§ Supports English, Dutch, French, German,
Japanese and Spanish

https://www.microsoft.com/en-us/ai/seeing-ai

33

Applications of AI
Detection of breast cancer in
Chest Radiology mammography images

Diabetic Retinopathy Drug Screening for COVID-19

Computer
AI in Healthcare Vision

34

16
 10/1/24

Applications of AI
Poverty Mapping

35

Applications of AI
Chess (1997): Kasparov vs. IBM Deep Blue

Powerful search algorithms!

36

17
 10/1/24

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

Natural Language Understanding and Information Extraction!

37

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

"I misjudged the capabilities of

AlphaGo and felt powerless.”,
quote after game 3.

Deep Learning, Reinforcement Learning and Search Algorithms!

38

18
 10/1/24

Applications of AI
Dota2 (2019): OpenAI wins OG world champions team of 5 pros

AI controlled 5 bots using different layers of same network

§ Trained with Reinforcement Learning and self-play
§ Equivalent to 45000 years (over 10 months)
39

Applications of AI
Autonomous Driving

§ DARPA Grand Challenge

§ 2005: 132 miles
§ 2007: Urban challenge
§ 2009: Google self-driving car
§ 2015: Uber started self-driving car

https://www.theverge.com/2023/8/10/23827790/waymo-cruise-
cpuc-vote-robotaxi-san-francisco

40

19
 10/1/24

REMBRANDT - New Rembrandt Painting with AI

https://www.youtube.com/watch?v=7z90lbbFz-c

41

State-of-the-art AI Applications
§ Speech recognition § Fraud detection
§ Autonomous planning and scheduling § Recommendation systems
§ Financial forecasting § Web search engines
§ Game playing, video games § Autonomous cars
§ Spam fighting § Energy optimization
§ Logistics planning § Question answering systems
§ Robotics (household, surgery, navigation) § Social network analysis
§ Machine translation § Medical diagnosis, imaging
§ Information extraction § Route finding
§ VLSI layout § Traveling salesperson
§ Automatic assembly § Protein design
§ Sentiment analysis § Document summarization
§ Computer animation § Transportation/scheduling
Many more!!!
42

20
 10/1/24

Characteristics of AI Tasks
What’s in common with all of these AI applications?
§ High societal impact (affect billions of people)
§ It’s clear that AI applications tend to be very high impact
§ Diverse (language, games, robotics)
§ They are incredibly diverse, operating in very different domains, and requiring integration
with many different modalities (natural language, vision, robotics)
§ Complex (really hard)
§ These applications are also mind-bogglingly complex to the point where we shouldn’t
expect to find solutions that solve these problems perfectly
§ Two sources of complexity in AI tasks Computation Information
§ Computational complexity: exponential explosion (time/memory) (data)
§ Information complexity: need to acquire knowledge,
cope with uncertainty

43

Foundations of AI

44

21
 10/1/24

AI Founders
§ Aristotle § Herbert Simon
§ Alan Turing § Allen Newell
§ John Mc Carthy § David Waltz
§ Warren McCulloh § Tom Mitchell
§ Walter Pitts § Stuart J. Russell
§ Claude Shannon § Peter Norvig
§ Marvin Minsky § etc.
§ Dean Edmonds

49

History of AI
§ 1940-1950: Gestation of AI
§ McCulloch & Pitts: Boolean circuit to model of brain
§ Turing’s “Computing Machinery and Intelligence”
http://www.turingarchive.org/browse.php/B/9

§ 1950-1970: Early enthusiasm, great expectations

§ Early AI programs, Samuel’s checkers program, chess, theorem providing
§ Birth of AI @ Dartmouth meeting 1956, John McCarthy coined “AI”
§ Check out the MIT video “The thinking Machine” on YouTube
https://www.youtube.com/watch?v=aygSMgK3BEM

§ 1970-1990: Knowledge-based approaches

§ Expert systems (XCON, MYCIN), AI becomes an industry
§ AI winter

50

22
 10/1/24

History of AI
§ 1990-2012: Scientific Approaches
§ Rise of Machine Learning in 1990s
§ Neural Networks: le retour
§ The emergence of intelligent agents
§ AI becomes “scientific”, use of probability to model uncertainty
§ AI Spring!
§ 2012-2022: Deep Learning – Excitement
§ Big Data, big compute, neural networks
v Latest rebirth à new machine learning techniques, tons of data, and tons of computation
§ The availability of very large datasets
v Data will drive future discoveries and alleviate the complexity in AI
§ 2022-present: More excitement
§ AI is being used by public.

51

Risks of AI
§ Energy Consumption

§ Privacy

§ Security

§ Fairnes

54

23
 10/1/24

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 environment
but we will still aim to build rational agents

55

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.”

56

24
 10/1/24

Next Week: Rational Intelligent Agents

§ This course is about designing intelligent agents
§ An agent perceives the environment and act upon that environment 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 performance measure
§ AI aims to design the best agents (programs) that achieve the best
performance given the computational limitations
Agent = Architecture + Program

57

25