Department of Computer Science

Introduction to Artificial Intelligence

COSC3112
 Course Detail
⚫ Course code: COSC3112
⚫ Credit hours: 3 ECTS: 5
⚫ Contact hrs: 2
⚫ Lab hrs: 3
⚫ Tutorial hrs: 2
⚫ Prerequisite:CoSc2092-Data Structures and Algorithms,
STAT2015 - Probability and Statistics
⚫ Course category: compulsory
⚫ Year: IIII Semester: I

2
 Course Description
⚫ T h e purpose of this course is to give students an
understanding of Artificial Intelligence methodologies,
techniques, tools and results.
⚫ Students will use p y t ho n p ro gra mmin g language to
demonstrate laboratory exercises.

⚫ Students will learn the theoretical and conceptual components

of this discipline and firm up their understanding b y using AI
and Expert System tools in laboratory sessions, projects and
h o m e assignments.

3
 Course objectives
At the end of this course the students will be able to:
⚫ Understand reasoning, knowledge representation and learning
techniques of artificial intelligence
⚫ Evaluate the strengths and weaknesses of these techniques and
their applicability to different tasks
⚫ Assess the role of AI in gaining insight into intelligence and
perception
⚫ Kn o w classical examples of artificial intelligence
⚫ Kn o w characteristics of programs that can be considered
"intelligent"
⚫ Understand the use of heuristics in search problems and
games

4
 Course objectives
⚫ Kn o w a variety of ways to represent and retrieve knowledge
and information
⚫ Kn o w the fundamentals of artificial intelligence programming
techniques in a mo de rn p ro gra mming language

⚫ Consider ideas and issues associated with social technical, and

ethical uses of machines that involve artificial intelligence
⚫ Introduce students for powerful learning algorithms and their
applications.

⚫ Letting students to develop simple AI powered applications

either in robotics, N L P or games.

5
 Course Outline
Topic Contents
Introduction to Objectives/Goals of AI, Types of AI(General and Specific AI), Approaches to AI –
AI(3 hrs) making computer: Think like a human ( Thinking humanly),Act like a human
(Acting humanly), Think rationally (Thinking rationally) & Act rationally (Acting
rationally). The Foundations of AI, Bits of History and the State of the Art,
Proposing and evaluating Application of AI

Intelligent Agents Foundation of Agents, Agents and Environments, Acting of Intelligent Agents
(4 hrs) (Rationality), Structure of Intelligent Agents: Agent Types, Simple reflex agent,
Model-based reflex agent, Goal-based agent, Utility-based agent. Multi agent
systems, Learning agent

Searching and Solving Problems by Searching and planning, Constraint Satisfaction Problem,
Planning (6 hrs) Problem Solving Agents, Problem spaces and search, Knowledge and rationality,
Heuristic search strategies, Search and optimization (gradient descent),
Adversarial search, Planning and scheduling, Avoiding Repeated States,
Dynamic game theory
Knowledge Logic and Inference, Logical Agents, Propositional Logic, Predicate (First-Order)
Representation Logic, Inference in First-Order Logic, Knowledge Representation, Knowledge
and Reasoning (8 Reasoning, Bayesian reasoning, Probabilistic reasoning, Temporal reasoning,
hrs) Knowledge-based Systems, Case study: Medical diagnosis
6
 Course Outline…
Machine Learning Knowledge in Learning, Learning Probabilistic Models, Supervised learning: Linear
Basics (3 hrs) classification models, Probabilistic models., Unsupervised learning: Clustering
models, Reinforcement learning, Deep Learning: Neural networks and back-
propagation, Convolution neural networks, Recurrent neural networks and LSTMs

Natural Language Intro to Natural Language Processing, Machine learning Application in NLP,
Processing (NLP) Natural language interaction, Computer vision and Image processing, Case study:
Basics (3 hrs) Sentiment Analysis, speech recognition, Chatbot

Robotic Sensing Introduction to robotics: Sensing, Manipulation, Human-robot interaction.

and Manipulation Navigation and path planning: Autonomous robotic systems
(3 hrs)

Ethical and Legal Privacy, Bias, AI and the future of work, Appropriate uses of AI
Considerations in
AI (2 hrs)

7
 Assessment methods
⚫ This course is assessed b y written exams, lab
assignments, reports and presentations of
assignments.
⚫ Assessment Breakdown:
⚫ Assignment/quizzes 10 %
⚫ Mi d semester examination 20%
⚫ Project ` 20%
⚫ Final examination 50%

8
 Reading Material
⚫ Text Books:
⚫ Russell, S. and P. Norvig (1995) Artificial Intelligence: A
Mo de rn Approach Prentice-Hall
⚫ References:
⚫ Luger, G. (2002) Artificial Intelligence, 4th ed. Addison-
Wesley.
⚫ Bratko, Ivan (1990) P R O L O G Programming for Artificial
Intelligence, 2nd ed. Addison-Wesley, 1990
⚫ Winston, P.H. (1992) Artificial Intelligence Addison-Wesley.
Ginsberg, M.L. (1993) Essentials of Artificial Intelligence.
Morgan Kaufman.

9
Topic 1:Introduction to AI
 Introduction to Artificial
Intelligence
 O ve rvie w of AI
 Objectives/Goals of AI
 T y p e s of AI(General and Specific AI)
 Approaches to AI – making computer:
 Thin k like a hu man ( Thinking humanly)
 Act like a hu man (Acting humanly)
 Think rationally (Thinking rationally)
 Act rationally (Acting rationally)
 T h e Foundations of AI
 Bits of History and the State of the Art
 Proposing and evaluating Application of AI

11
 Overview of AI
⚫ W e call ourselves H o m o sapiens—m a n the w i s e — because our
intelligence is so important to us.
⚫ For thousands of years, w e have tried to understand how we
think; that is, ho w a m e r e handful of matter can perceive,
understand, predict, and manipulate a world far larger and
m o r e complicated than itself.
⚫ T h e field of artificial intelligence, or AI, goes further still: it
attempts not just to understand but also to build intelligent
entities.
⚫ AI is one of the newest fields in science and engineering.
W o r k started in earnest soon after Wo rl d Wa r II, and the
n a me itself was coined in 1956.
12
 Overview of AI
⚫ AI currently encompasses a huge variety of subfields, ranging
f r o m the general (learning and perception) to the specific,
such as playing chess, proving mathematical theorems,
writing poetry, driving a car o n a crowded street, and
diagnosing diseases.
⚫ AI is relevant to any intellectual task; it is truly a universal
field.

13
 Intelligence
⚫ Intelligence is the computational part of the ability to achieve
goals in the world.
⚫ Varying kinds and degrees of intelligence occur in people,
m a n y animals and s o me machines.
⚫ Intelligence is the capability of observing, learning,
remembering, reasoning and acting.
⚫ AI attempts to develop intelligent agents.
Characteristics of Intelligent system
⚫ Use vast amount of knowledge
⚫ Learn f r o m experience and adopt to changing environment
⚫ Interact with hu man using language and speech
⚫ Respond in real time
⚫ Tolerate error and ambiguity in communication

14
 Artificial Intelligence
⚫ It is the science and engineering of making intelligent
machines, especially intelligent computer programs.
⚫ It is related to the similar task of using computers to
understand h u m a n intelligence, but AI does not have to
confine itself to methods that are biologically observable.
⚫ Isn't there a solid definition of intelligence that doesn't
depend on relating it to human intelligence?
⚫ Not yet. T h e problem is that we cannot yet characterize in
general what kinds of computational procedures we want to call
intelligent. W e understand some of the mechanisms of
intelligence and not others.

15
 Artificial Intelligence
⚫ T h e concern of AI is to develop computer based system that
behave like h u m a n and emulate the reasoning p ow e r of
humans in order to do tasks that require h u m a n intelligence.
Which task requires intelligence?
⚫ C o m p l e x arithmetic operations
⚫ For instance, Solving 220 * 350?
⚫ Mundane tasks: all tasks routinely done b y all of us in our day-
to-day activities:
⚫ Example, Natural language understanding; face recognition;
planning; robotics, …
⚫ Expert tasks: which require specialists knowledge
⚫ Example, Medical diagnosis; computer maintenance; financial
planning

16
 Artificial Intelligence
⚫ Artificial intelligence, or AI, is the field that studies the
synthesis and analysis of computational agents that act
intelligently. Consider each part of this definition.
⚫ An agent is something that acts in an environment; it does
something.
⚫ Agents include worms, dogs, thermostats, airplanes, robots,
humans, companies, and countries.
⚫ An agent is judged solely b y ho w it acts. Agents that have the
same effect in the world are equally good.

17
 Artificial Intelligence
⚫ Intelligence is a matter of degree. T h e aspects that go into an
agent acting intelligently include:
⚫ what it does is appropriate for its circumstances, its goals,
and its perceptual and computational limitations
⚫ it takes into account the short-term and long-term
consequences of its actions, including the effects on
society and the environment
⚫ it learns f r o m experience
⚫ it is flexible to changing environments and changing goals.

18
 Computational Agent
⚫ A computational agent is an agent whose decisions about its
actions can be explained in terms of computation.
⚫ That is, the decision can be broken do w n into primitive
operations that can be implemented in a physical device.
⚫ This computation can take m a n y forms. In humans, this
computation is carried out in “wetware”; in computers it is
carried out in “hardware.”
⚫ Although there are s o me agents that are arguably not
computational, such as the wind and rain eroding a landscape,
it is an o p e n question whether all intelligent agents are
computational.

19
 Computational Agent
⚫ All agents are limited. N o agent is omniscient (all knowing) or
omnipotent (can d o anything).
⚫ Agents can only observe everything in ve ry specialized and
constrained domains.
⚫ Agents have finite m e m o r y .
⚫ Agents in the real world d o not have unlimited time to act.

20
 Objectives/Goals of AI
⚫ T h e central scientific goal of AI is to understand the principles
that make intelligent behavior possible in natural or artificial
systems. This is done b y
⚫ the analysis o f natural and artificial agents
⚫ formulating and testing hypotheses about what it takes to construct
intelligent agents
⚫ designing, building, and experimenting with computational systems that
p e r f o r m tasks c o m m o n l y viewed as requiring intelligence.

⚫ As part o f science, researchers build empirical systems to test

hypotheses or to explore the space o f possible designs.
⚫ These are distinct f r o m applications that are built to be useful for
an application domain.
⚫ T h e definition is not for intelligent thought. T h e role o f thought
is to affect action and lead to m o r e intelligent behavior.
21
 Objectives/Goals of AI
⚫ T h e central engineering goal o f AI is the design and synthesis of
agents that act intelligently, which leads to useful artifacts.
⚫ Building general intelligence isn’t the o nl y goal o f AI researchers.
⚫ T h e aim o f intelligence augmentation is to augment human
intelligence and creativity.
⚫ A diagnostic agent helps medical practitioners m a k e better decisions,
a
search engine augments h u m a n m e m o r y , and natural language
translation systems help people communicate.
⚫ AI systems are often in h u m a n -in-the-loop mo d e , where humans
and agents work together to solve problems.
⚫ Sometimes the actions o f artificial agents are to give advice to a human.
⚫ Sometimes humans give advice or feedback to artificial agents, particularly
for cases where decisions are m a d e quickly or repeatedly.

22
 Types of AI(General and
Specific AI)
⚫ AI can generally be categorized into two types: narrow (Special-
purpose or weak) AI and general (or strong) AI.
⚫ General-purpose AI like the robots o f science fiction is incredibly
hard
⚫ H u ma n brain appears to have lots of special and general functions,
integrated in s om e amazing w a y that w e really d o not understand at
all (yet)
⚫ T h e strong AI view holds that it is possible to create intelligent
machines that can really reason and solve problems.
⚫ Such machines are considered to be conscious and self-aware, can
independently think about problems and work out optimal solutions
to problems, have their ow n system of values and world views, and
have all the same instincts as living things, such as survival and
security needs.
23
⚫ It can be regarded as a new civilization in a certain sense.
 Types of AI(General and
Specific AI)
⚫ Special-purpose AI is m o r e doable (nontrivial)
⚫ E.g., chess/poker playing programs, logistics planning,
automated translation, voice recognition, web search, data
mining, medical diagnosis, keeping a car o n the road, … …
……
⚫ T h e weak AI view holds that intelligent machines cannot
really reason and solve problems.
⚫ These machines only look intelligent, but do not have real
intelligence or self-awareness.

24
 Strong AI vs Weak AI
⚫ Strong AI argues that it is possible that o ne day a computer will be
invented which can be called a m i n d in its fullest sense.
⚫ Strong AI aims to create an agent that can replicate humans
intelligence completely; i.e., it can think, reason, imagine, etc., & d o all
the things that are currently associated with the h u m a n brain.
⚫ W e a k AI, o n the other hand, argue that computers can only
appear to think & are not actually conscious in the same w a y as
h u m a n brains are.
⚫ T h e weak AI position holds that AI should try to develop systems
which have facets of intelligence, but the objective is not to build a
completely sentient/conscious entity.
⚫ W ea k AI researchers see their contribution as systems like expert
systems used for medical diagnosis, which use "intelligent" models,
but they d o not help create a conscious agent
25
 Strong AI vs Weak AI
⚫ A ke y counter o f AI is to achieve a superhuman level in
challenging fields through self-learning without any prior
knowledge.
⚫ T h e followers of strong AI believe that b y giving a computer program sufficient
processing power, and b y providing it with enough intelligence, one can create a
computer that can literally think and is conscious in the same way that a hu man is
conscious.
⚫ Strong AI can compete with humans in all aspects.
⚫ Therefore, it aims to enable robots to implement human-like
capabilities in all aspects rather than a specific field.
⚫ Strong AI can think, make plans, solve problems, perf orm abstract thinking, understand
compl ex concepts, quickly learn, and learn f r o m experience.

⚫ Currently, it is believed that if we can simulate the hu man brain and

c o p y all its neurons and synapses o n the same scale, strong AI will

26
naturally occur(??).
 Strong AI vs Weak AI
⚫ N o w w e are in the weak AI phase.
⚫ Weak AI, in contrast, is simply the view that intelligent behavior can be modeled and used b y computers
to solve complex problems. This point of view argues that just because a computer behaves intelligently
does not prove that it is actually intelligent in the way that a human is.

⚫ T h e emergence of weak AI alleviates hu man intellectual labor, similar

to advanced bionics.
⚫ Both AlphaGo and robots that can write press releases and novels fall
in the weak AI phase because they are better than humans only in
some ways.
⚫ T h e roles of data and computing power are self-evident in the era of
weak AI, and promote the commercialization of AI.
⚫ In the era of strong AI, these two factors are still critical.
⚫ At the same time, the research o n quantum computing b y technology
giants like Google and IBM also provides powerful support for

27
humans to enter the era of strong AI.
 Strong AI vs Weak AI
Weak AI Strong AI
Limited to perform specific tasks Perform intelligent human-level
activities
Programmed for fixed function Have the ability to learn, think, and
perform new activities like humans

It doesn’t have any consciousness It poses creativity, common sense,

or awareness of its own. and logic like humans.

They have a goal to complete a They have a goal to solve problems

task with creative and accurate at a faster pace.
solutions.
Examples of weak AI include Alexa, There are no real examples of
Siri, and Google Assistant. strong AI because it is a
hypothetical theory. Some fictional
examples are Wall-E and Big Hero
28 6.
 Approaches to AI – making
computer:
⚫ AI is found o n the premise that:
⚫ T h e workings of hu man min d can be explained in terms of
computation, and
⚫ computers can do the right thing given correct premises and
reasoning rules.
⚫ Intelligence is:
⚫ “the capacity to learn and solve problems” (Websters dictionary)
⚫ in particular,
⚫ the ability to solve novel problems
⚫ the ability to act rationally
⚫ the ability to act like humans
⚫ Views of AI fall into four categories:

Thinking humanly Thinking rationally

Acting humanly Acting rationally

29
 Approaches to AI – making computer:

30
 Thinking humanly: The
Cognitive Modeling
⚫ Reasons like humans do
⚫ Programs that behave like humans

⚫ Requires understanding of the internal activities of the

brain
⚫ see h o w humans behave in certain situations and see
if y o u could m a ke computers behave in that same
way.
Example. write a program that plays chess.
⚫ Instead of making the best possible chess-playing
program, y o u would m a ke o n e that play chess like
people do.

31
 Acting humanly: The Turing
Test
Ca n machines act like h u m a n do? Ca n machines behave
intelligently?
⚫ Turing Test: Operational test for intelligent behavior
⚫ do experiments o n the ability to achieve human-level
performance,
⚫ Acting like humans requires AI programs to interact with
people
⚫ Suggested major components of AI include: knowledge,
reasoning, language understanding, learning

32
 Thinking Rationally: The Laws
of Thought
⚫ A system is rational if it thinks/does the right thing through
correct reasoning.
⚫ Aristotle: provided the correct arguments/thought structures
that always gave correct conclusions given correct premises.
⚫ Belay is a man; all m e n are mortal; therefore Belay is
mortal
⚫ These Laws of thought governed the operation of the
min d and initiated the field of Logic.

33
 Acting rationally: The rational
agent
⚫ Doing the right thing so as to achieve one’s goal, given one’s
beliefs.
⚫ AI is the study and construction of rational agents (an agent
that perceives and acts)
⚫ Rational action requires the ability to represent knowledge
and reason with it so as to reach good decision.
⚫ Learning for better understanding of how the world works

34
 The Foundations of AI
⚫ AI is Interdisciplinary
⚫ AI is generally associated with C o m p u t e r Science, but it
has m a n y important links with other fields such as:
⚫ Mathematics,
⚫ Biology,
⚫ Economics,
⚫ Cognition,
⚫ Psychology,
⚫ Philosophy and
⚫ Neuroscience.

35
 Philosophy
⚫ Can formal rules be used to draw valid conclusions?
⚫ How does the mind arise from a physical brain?
⚫ Where does knowledge come from?
⚫ How does knowledge lead to action?
⚫ Descartes was a proponent o f dualism. H e held that there is a part
o f the h u m a n m i n d (or soul or spirit) that is outside o f nature,
e xe mpt f r o m physical laws. Animals, o n the other hand, did not
possess this dual quality; they could be treated as machines.
⚫ An alternative to dualism is materialism, which holds that the
brain’s operation according to the laws o f physics constitutes the
mind. Free will is simply the w a y that the perception o f available
choices appears to the choosing entity.

36
 Philosophy
⚫ T h e empiricism is characterized b y a dictum o f John Locke
(1632–1704): “Nothing is in the understanding, which was not first
in the senses.”
⚫ T h e logical positivism holds that all knowledge can be
characterized b y logical theories connected, ultimately, to
observation sentences that correspond to sensory inputs; thus
logical positivism combines rationalism and empiricism.
⚫ T h e confirmation theory attempted to analyze the acquisition of
knowledge f r o m experience b y quantifying the degree o f belief
that should be assigned to logical sentences based o n their
connection to observations that co nfi rm or disconfirm them.

37
 Philosophy
⚫ Utilitarianism: that rational decision making based on
maximizing utility should apply to all spheres o f h u m a n activity,
including public policy decisions m a d e o n behalf o f m a n y
individuals. Utilitarianism is a specific kind o f consequentialism:
the idea that what is right and wrong is determined b y the
expected outcomes o f an action.
⚫ In contrast, a theory o f rule-based or deontological ethics, in
which “doing the right thing” is determined not b y outcomes but
b y universal social laws that govern allowable actions, such as
“don’t lie” or “don’t kill.” M a n y m o d e r n AI systems adopt exactly
this approach.

38
 Mathematics
⚫ What are the formal rules to draw valid conclusions?
⚫ What can be computed?
⚫ How do we reason with uncertain information?
⚫ T h e theory of probability can be seen as generalizing logic to
situations with uncertain information —a consideration of great
importance for AI. T h e formalization of probability, combin ed with
the availability of data, led to the emergen ce of statistics as a field.
⚫ T h e history of computation is as old as the history of numbers, but
the first nontrivial algorithm is thought to be Euclid’s algorithm for
computing greatest c o m m o n divisors. T h e incompleteness theorem
showed that in any formal theory as strong as Peano arithmetic (the
elementary theory of natural numbers), there are necessarily true
statements that have n o proof within the theory.

39
 Mathematics
⚫ Alan Turing tried to characterize exactly which functions are
computable — capable of being compu ted b y an effective procedure.
For example, n o machine can tell in general whether a given program
will return an answer o n a given input or run forever.

⚫ Although computability is important to an understanding of

computation, the notion of tractability has had an even greater
impact o n AI. A problem is called intractable if the time required to
solve instances o f the problem grows exponentially with the size of
the instances.
⚫ T h e theory of NP-completeness provides a basis for analyzing the
tractability of problems: any problem class to which the class of NP-
complete problems can be reduced is likely to be intractable.

40
 Economics
⚫ How should we make decisions in accordance with our preferences?
⚫ How should we do this when others may not go along?
⚫ How should we do this when the payoff may be far in the future?
⚫ Decision theory, which combines probability theory with utility
theory, provides a formal and complete framework for individual
decisions (economic or otherwise) made under uncertainty—that is, in
cases where probabilistic descriptions appropriately capture the
decision maker’s environment.
⚫ Models based o n satisficing—making decisions that are “good
enough,” rather than laboriously calculating an optimal decision—
gave a better description of actual hu man behavior.

41
 Neuroscience
⚫ How do brains process information?
⚫ Neuroscience is the study of the nervous system, particularly the
brain. A collection of simple cells can lead to thought, action, and
consciousness.
⚫ Even with a computer of virtually unlimited capacity, we still require
further conceptual breakthroughs in our understanding of
intelligence.

42
 Psychology
⚫ How do humans and animals think and act?
⚫ Wu n dt insisted o n carefully controlled experiments in which his
workers would perform a perceptual or associative task while
introspecting o n their thought processes. T h e behaviorism
m o v e m e n t rejected any theory involving mental processes o n the
grounds that introspection could not provide reliable evidence.
⚫ Cognitive psychology views the brain as an information-processing
device. T h r e e key steps of a knowledge-based agent: (1) the stimulus
must be translated into an internal representation, (2) the
representation is manipulated b y cognitive processes to derive new
internal representations, and (3) these are in turn retranslated back
into action.
⚫ Intelligence augmentation states that computers should augment
hu man abilities rather than automate away hu man tasks.
43
 Computer engineering
⚫ How can we build an efficient computer?

⚫ Moore’s law states that performance doubled every 18 months.

⚫ Quantum computing holds out the promise of far greater
accelerations for some important subclasses of AI algorithms.

44
 Control theory and cybernetics
⚫ How can artifacts operate under their own control?

⚫ Modern control theory, especially the branch known as stochastic

optimal control, has as its goal the design of systems that maximize a
cost function over time.

45
 Linguistics
⚫ How does language relate to thought?

⚫ Modern linguistics and AI, then, were “born” at about the same time,
and grew u p together, intersecting in a hybrid field called
computational linguistics or natural language processing.

46
 THE HISTORY OF ARTIFICIAL
INTELLIGENCE
The inception of artificial intelligence (1943-1956)
⚫ T h e first work that is n o w generally recognized as AI was don e by
Warren McCulloch and Walter Pitts (1943).
⚫ T h e y proposed a mo de l o f artificial neurons. Year 1949: Donald Hebb
demonstrated an updating rule for modifyi ng the connection strength
between neurons. His rule is n o w called Hebbian learning.
Early enthusiasm, great expectations (1952-1969)
⚫ T h e intellectual establishment of the 1950s, b y and large, preferred o
t
believe that “a machine can never do X”. John McCarth y referred to
this period as the “Look, Ma, n o hands!” era.
A does of reality (1966-1973)
⚫ In almost all cases these early systems failed o n m o r e difficult
problems. T h e illusion of unlimited computational power was not
47 confined to problem-solving programs.
 THE HISTORY OF ARTIFICIAL
INTELLIGENCE
Expert systems (1969-1986)
⚫ T h e picture of problem solving that had arisen during the first decade
of AI research was of a general-purpose search mechanism trying to
string together elementary reasoning steps to find complete solutions.
Such approaches have been called weak methods. T h e alternative to
weak methods is to use m o r e powerful, domain-specific knowledge
that allows larger reasoning steps and can m o r e easily handle typically
occurring cases in narrow areas of expertise.
The return of neural networks (1986-present)
⚫ In the mid-1980s at least four different groups reinvented the back-
propagation learning algorithm first developed in the early 1960s.

48
 THE HISTORY OF ARTIFICIAL
INTELLIGENCE
Probabilistic reasoning and machine learning (1987-present)
⚫ In the 1980s, approaches using hidden Markov models (HMMs) came
to dominate the area. Pearl’s development of Bayesian networks
yielded a rigorous and efficient formalism for representing uncertain
knowledge as well as practical algorithms for probabilistic reasoning.
Big data (2001-present)
⚫ Remarkable advances in computing power and the creation of the
World Wide W e b have facilitated the creation of very large data
sets—a p h e n o m e n o n sometimes known as big data.
Deep learning (2011-present)
⚫ T h e term deep learning refers to machine learning using multiple layers of
simple, adjustable computing elements. Experiments were carried out with such
networks as far back as the 1970s, and in the f o r m o f convolutional neural
networks they f o und s o m e success in handwritten digit recognition in the
49
1990s.
 The State of the Art
● R O B O T I C VEHICLES
● LEGGED L O C O M O T I O N
● A U T O N O M O U S P L A N N I N G A N D S CHED U LING
● M A CHI N E T R A N S L A T I O N
● SPEECH R E C O G N I T I O N
● RECOMMENDATIONS
● G A M E PLAYING
● IMAGE U ND ERS TA ND I N G
● MEDICINE
● C L I M A T E SCIENCE

50
 Applications of AI
Solving problems that required thinking b y humans:
● Playing games (chess, checker, cards, ...)
● Great advances: the AI software D eep Blue beats h u m a n expert
Kasparov.
● Proving theorems (mathematical theorems, laws o f physics, …)
● Classification o f text (Politics, Economic, Social, Sports, etc,)
● Information filtering and summarization o f text
● Writing story and poems; solving puzzles
● Giving advice in Medical diagnosis, Equipment repair, Co mpute r
configuration, Financial planning,

51
 How to make computers act
like humans?
T h e following sub-fields have emerged
● Natural Language processing (enable computers to communicate in
hu man language, English, Amharic, ..)
● Knowledge representation (schemes to store information, both facts
and inferences,…)
● Automated reasoning (use stored information to answer questions
and to draw new conclusions)
● Machine learning (adapt to new circumstances and accumulate
knowledge)
● Computer vision (recognize objects based o n patterns in the same
way as the hu man visual system does)
● Robotics (produce mechanical device capable of controlled motion
with the ability to move, see, hear, and accordingly take actions in the

52 world, possibly responding to new perceptions)

 AI vs. HI?
T h e following sub-fields have emerged
● Artificial Intelligence (or AI) is the field that explores to
develop a system that think in the same sense as humans do.
● R e m e m b e r computer based chess program (Deep Blue) that beats
hu man expert (Gary Kasparov). What do y o u understand from
this?
● Does AI equal h u m a n intelligence?
● Is it possible to create a computer system called mind?
● What is our concern in designing an Intelligent agent?
● Is it to replace h u m a n beings or to support and give leverage
to t he m so that the humans can engage themselves in expert
works?

53
 Can we build hardware as
complex as the brain?
● H o w complicated is our brain?
● a neuron, or nerve cell, is the basic information processing unit in our
brain
● estimated to be o n the order o f 10 p o w 11 neurons in a h u m a n brain
● m a n y m o r e synapses (10pow14) connecting these neurons
● cycle time: 10 p o w - 3 seconds (1 millisecond)
● H o w complex can we make computers?
● 106 or m o r e transistors per CPU
● supercomputer: hundreds o f CPUs, 10pow9 bits o f RAM
● cycle times: order o f 10pow- 8 seconds
● Conclusion
● POSSIBLY: in the future w e m a y have computers with as m a n y basic
processing elements as our brain, but with far fewer interconnections
(wires or synapses) than the brain and m u c h faster updates than the brain
● but building hardware is ve r y different f r o m making a computer behave
54
like a brain!
 Can Computers Talk?
● This is known as “speech synthesis”
● tra nslate text to phonetic f o r m
● e.g., “fictitious” -> fik-tish-es

● us e pronunciation rules to m a p p h o ne mes to actual sound

● e.g., “tish” -> sequence of basic audio sounds

● Difficulties
● sounds m a d e b y this “lookup” approach sound unnatural
● so unds are not independent
● e.g., “act” and “action”

● m o d e r n systems (e.g., at A T & T ) can handle this pretty well

● a harder problem is emphasis, emotion, etc
● humans understand what they are saying
● machines don’t: so they sound unnatural

● Conclusion: N O , for complete sentences, but YES for individual

words
55
 Can Computers Recognize
Speech?
● Speech Recognition:
● ma p p ing sounds f r o m a microphone into a list of words.
● Hard problem: noise, m o r e than one person talking,
occlusion, speech variability,..
● Even if w e recognize each word, w e m a y not understand
its meaning.

● Recognizing single words f r o m a small vocabulary

● systems can do this with high accuracy (order of 99%)
● e.g., directory inquiries
● limited vocabulary (area codes, city names)
● co mp uter tries to recognize y o u first, if unsuccessful hands y o u over to a human

● operator
saves millions o f dollars a year for the p h o ne companies

56
 Recognizing human speech
(ctd.)
● Recognizing normal speech is m u c h m o r e difficult
● speech is continuous: where are the boundaries between words?
● e.g., “John’s car has a flat tire”

● large vocabularies
● can be m a n y tens of thousands of possible words

● we can use context to help figure out what s om e on e said

● try telling a waiter in a restaurant:
“I would like s om e cream and sugar in m y coffee”
● background noise, other speakers, accents, colds, etc
● o n normal speech, modern systems are only about 60% -70%
accurate
● Conclusion: N O , normal speech is too co mp l e x to accurately
recognize, but YES for restricted problems
● (e.g., recent software for P C use b y IBM, Dragon systems, etc)

57
 Can Computers Understand
speech?
● Understanding is different to recognition:
● “ T ime flies like an arrow”
● assume the computer can recognize all the words. but h o w could it
understand the sentence
● 1. time passes quickly like an arrow?
● 2. command: time the flies the wa y an arrow times the flies
● 3. command: o nly time those flies which are like an arrow
● 4. “time-flies” are f o nd o f arrows
● only 1. makes any sense, but ho w could a computer figure

● out?
this clearly humans use a lot of implicit co mmonsense knowledge in communication

● Conclusion: N O , m u c h of what w e say is b e y o n d the

capabilities of a computer to understand at present

58
 Can Computers Learn and
Adapt ?
● Learning and Adaptation
● consider a computer learning to drive o n the freeway
● w e could code lots of rules about what to do
● and/or w e could have it learn f r o m experience
Darpa’s Grand Challenge. Stanford’s
“Stanley” drove 150 km without
supervision in the Majove dessert

● 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

59
 Can Computers “see”?
● Recognition v. Understanding (like Speech)
● Recognition and Understanding of Objects in a scene
● look around this r o o m
● y o u can effortlessly recognize objects
● h u m a n brain can m a p 2D visual image to 3D “map”
● W h y is visual recognition a hard problem?

Conclusion: mostly NO : computers can only “see” certain

types of objects under limited circumstances: but YES for
certain constrained problems (e.g., face recognition)
60
 Must an Intelligent System be
Foolproof?
● A “foolproof” system is one that never makes an error:
● T y p e s of possible computer errors
● hardware errors, e.g., m e m o r y errors
● software errors, e.g., coding bugs
● “human-like” errors

● Clearly, hardware and software errors are possible in practice

● what about “human-like” errors?
● An intelligent system can make errors and still be intelligent
just like humans w ho apparently are not right all of the time
● we learn and adapt f r o m making mistakes
● e.g., consider learning to surf, ski, or ride a bike
● we improv e b y taking risks and falling

● an intelligent system can learn in the same way

● Conclusion:
● NO: intelligent systems will not (and need not) be foolproof
61
 Programming paradigms
● Each pro g ramm i n g paradigms consists of two aspects:
● Methods for organizing data/knowledge,
● Methods for controlling the flow of computation
● Traditional paradigms:
Programs = data structure + control
● AI programming paradigms:
Programs = knowledge structure + inference

62
 Risks and Benefits of AI
● LETHAL AUTONOMOUS WEAPONS
● SURVEILLANCE A N D PERSUASION
● BIASED DECIS I ON M A K I N G
● IMPACT O N EMPLOYMENT
● SAFETY -C R I T I C A L A P PL IC ATION S
● CYBERSECURITY

63
The End of Topic 1