Artificial Intelligence

Lecture # 1
Instructor:
Anam Khalid
Defining Intelligence
Intelligence
The concept of intelligence is complex and multifaceted, with
various definitions and perspectives. While experts
acknowledge the lack of a precise definition for "human-level
artificial intelligence," they often rely on working definitions to
understand and discuss the topic. This presentation explores
different perspectives on intelligence, its inheritance, and the
concept of intelligence quotient.
Working Definitions of Intelligence

1 Knowledge and Skills

The ability to acquire and apply knowledge and skills is a
fundamental aspect of intelligence.

2 Environmental Interaction
Intelligence encompasses the ability to learn about, learn from,
understand, and interact with one's environment.

3 Faculty of Understanding
Intelligence is often described as the faculty of understanding, the
capacity to grasp and comprehend information.

4 Problem-Solving Skills
Intelligence is defined as general cognitive problem-solving skills,
encompassing reasoning, perceiving relationships, calculating, and
learning quickly.
Theories of Intelligence

Early theories suggested a single underlying factor, the "g-

factor," responsible for intelligence. However, later
psychologists recognised the complexity of intelligence and
proposed multiple components. Howard Gardner, for example,
identified seven components of intelligence: musical, bodily-
kinesthetic, logical-mathematical, linguistic, spatial,
interpersonal, and intrapersonal.
Musical Bodily-Kinesthetic Logical-
Mathematical

Linguistic Spatial Interpersonal

Intrapersonal
Intelligence and Heredity
While intelligence is generally accepted to be inherited, environmental factors also play a significant role. Studies
have shown that heredity contributes significantly to intelligence, but the environment influences its expression.
Recent research suggests that 70% of the differences in twins' IQ scores are attributable to inherited traits, a higher
percentage than previously estimated.

Genetic Influence Environmental Impact

Studies have shown a strong genetic influence on IQ, While environmental factors can influence intelligence,
verbal and spatial abilities, and grey matter volume, they are more likely to slow down mental processes
which is linked to cognitive performance. than enhance them. There is no evidence that the
environment can significantly increase intelligence.
Intelligence Quotient (IQ)
(IQ)
An intelligence quotient, or IQ, is a score derived from
standardized tests designed to assess intelligence. The term "IQ"
originates from the German "Intelligenz-Quotient," coined by
psychologist William Stern. In current IQ tests, the median raw
score of the norming sample is defined as IQ 100, with scores
increasing or decreasing by 15 IQ points for each standard
deviation.

Approximately 95% of the population scores an IQ between 70

and 130, within two standard deviations of the mean.
Intelligence and the Environment

While genes play a crucial role in determining intelligence, environmental

factors can influence its development and expression. These factors
include access to education, nutrition, and stimulating environments.
While the environment cannot significantly increase intelligence, it can
impact its development and utilization.
Early Childhood
Early childhood experiences, including access to quality
education and stimulating environments, can significantly
impact cognitive development.

Social and Cultural Factors

Social and cultural factors, such as access to resources and
opportunities, can influence the development and expression
of intelligence.

Lifelong Learning
Continuous learning and engagement in mentally stimulating
activities throughout life can contribute to cognitive health and
maintain intellectual abilities.
Artificial Intelligence: A
Multifaceted Field
Artificial Intelligence (AI) is a branch of science that focuses on
creating machines capable of solving complex problems in a way
that mimics human intelligence. This involves borrowing human
intelligence characteristics and translating them into algorithms
for computer processing. The level of flexibility and efficiency in
these algorithms can vary depending on the specific
requirements, influencing how "artificial" the intelligent behavior
appears.
Interdisciplinary Nature of AI
AI is often associated with Computer Science, but it draws heavily from other fields like Mathematics,
Psychology, Cognition, Biology, and Philosophy. Combining knowledge from these diverse disciplines is crucial
for advancing the development of intelligent artificial beings.

Mathematics Psychology
Provides the foundation for algorithms and computational Offers insights into human cognition and decision-making
models. processes.

Cognition Biology
Studies how humans acquire, process, and use knowledge. Provides understanding of biological systems and their
potential for inspiration in AI.
Defining Artificial Intelligence

Various definitions of AI exist, each highlighting different aspects of this

complex field. Some focus on the creation of intelligent machines, while
others emphasize the emulation of human intelligence or the ability to
solve problems that previously required human intelligence.

1 Intelligent Machines
AI is the science and engineering of making intelligent machines,
especially intelligent computer programs.

2 Emulating Human Intelligence

AI is related to the task of using computers to understand human
intelligence, but it doesn't have to be limited to biologically
observable methods.
3 Problem Solving
The use of computers to solve problems that previously could only
be solved by applying human intelligence.
AI: Thinking and Acting Like Humans

Some definitions emphasize the ability of AI systems to think or act like humans. These definitions highlight the
goal of creating machines that can reason, learn, and interact with the world in a way that resembles human
behavior.

Thinking Like Humans Acting Like Humans

AI is a study in which computer systems are made that AI is a study in which computer systems are made that
think like human beings. (Haugeland, 1985 & Bellman, act like people. (Kurzweil, 1990)
1978)
AI: Outperforming Human Capabilities

Another perspective focuses on AI's potential to surpass human capabilities in specific tasks. This definition
emphasizes the goal of creating machines that can perform tasks that are currently better done by humans.

AI is the study of how to make computers do things which at the moment people are better at. (Rich & Knight,
1991)
AI: Rational Thinking and Acting
Some definitions emphasize the development of AI systems that can think and act rationally. These definitions
focus on creating machines that can make logical decisions and take actions based on reason and evidence.

Rational Thinking Rational Acting

AI is a study in which computers that rationally think AI is the study in which systems that rationally act are
are made. (Charniac & McDermott, 1985) made. (Schalkeoff, 1990)
AI: Computational Processes
Processes for Intelligence
Intelligence
AI is often defined as the study of computational processes that
enable perception, reasoning, and action. This definition
emphasizes the underlying mechanisms that allow AI systems to
exhibit intelligent behavior.

AI is the study of computations that make it possible to perceive,

reason and act. (Winston, 1992)
AI: Automation of Intelligent
Intelligent Behavior

AI is considered a branch of computer science that focuses on

automating intelligent behavior. This definition highlights the
practical applications of AI in various domains, from robotics to
natural language processing.

AI is considered to be a branch of computer science that is

concerned with the automation of intelligent behavior. (Luger &
Stubblefield, 1993)
What’s Involved in
Intelligence?

Intelligence is a complex concept that encompasses a wide range

of abilities, including the ability to interact with the real world,
reason and plan, learn and adapt, and communicate. This
document explores these aspects of intelligence and examines
whether computers can replicate them.
Can We Build Hardware as Complex as the Brain?

The human brain is incredibly complex, with an estimated 10^12 neurons and 10^14 synapses. While computers
are becoming increasingly powerful, they still lag behind the brain in complexity and interconnectivity. However,
computers could soon have as many essential processing elements as the brain but with fewer interconnections
and faster update times.

Brain Computer

10^12 neurons 10^8 or more transistors per CPU

10^14 synapses Supercomputer: hundreds of CPUs, 10^12 bits of RAM

Cycle time: 10^-3 seconds Cycle times: order of 10^-9 seconds

Can Computers Talk?
Speech synthesis, the process of translating text into spoken language, is
a challenging task for computers. While systems can generate individual
words with high accuracy, producing natural-sounding sentences
remains difficult. The complexities of emphasis, emotion, and context
make it challenging for machines to sound truly human.

Phonetic Mapping
Translating text to phonetic form, e.g., "fictitious" -> fik-tish-es.

Pronunciation Rules
Mapping phonemes to actual sound, e.g., "tish" -> sequence of basic
audio sounds.

Difficulties
Sounds made by this approach sound unnatural. Sounds are not
independent, e.g., "act" and "action".
Can Computers Recognize Speech?

Speech recognition, the process of mapping sounds from a microphone

into a list of words, is a classic problem in AI. While systems can
accurately recognize single words from a small vocabulary, recognizing
normal speech is much more difficult due to the continuous nature of
speech, large vocabularies, and the presence of background noise and
accents.

1 Single Word Recognition

Systems can recognize single words from a small vocabulary with
high accuracy (order of 99%).

2 Normal Speech Recognition

Modern systems are only about 60-70% accurate due to the
complexities of continuous speech, large vocabularies, and
background noise.
Can Computers Understand Speech?

Understanding speech is a different challenge than recognizing it.

Computers can recognize words, but understanding their meaning
requires a level of commonsense knowledge and context that is still
beyond their capabilities. For example, the sentence "Time flies like an
arrow" has multiple interpretations, and a computer would struggle to
determine the intended meaning.
Time passes quickly like an arrow?
This is the most likely interpretation.

Command: time the flies the way an arrow times the

times the flies
This interpretation is less likely but still possible.

"Time-flies" are fond of arrows

This interpretation is unlikely and nonsensical.
Can Computers Learn and Adapt?

Computers can learn and adapt, but they require specific information and training.
Machine learning allows computers to learn to do things without explicit
programming, but it still requires some setup and is not a replacement for human
intelligence. For example, a computer can learn to drive on a freeway by being
trained on a dataset of driving scenarios and being corrected when it makes
mistakes.

1 Rule-Based Learning
Teaching a computer explicit rules about how to drive.

2 Machine Learning
Allowing a computer to learn from data and experience, such as
driving scenarios.

3 Adaptation
The computer adjusts its driving behavior based on feedback and
experience.
Can Computers “See”?
Computers can recognize certain types of objects under limited circumstances, but they struggle with the complexities of
visual understanding. The human brain can effortlessly map a 2D visual image to a 3D "map," allowing us to understand the
world around us. Computers, however, lack this ability and are limited to recognizing specific objects in controlled
environments.
Human Computer

Effortlessly recognizes objects Limited to recognizing specific objects in controlled

environments

Maps 2D visual image to 3D "map" Struggles with the complexities of visual understanding
Can Computers Plan and Make Optimal Decisions?

Planning and decision-making involve solving problems and making choices in a complex world. While computers can excel
in well-defined, constrained problems, they struggle with real-world planning that involves unpredictable factors, a vast
number of details, and the need for commonsense knowledge. For example, planning a vacation to Brazil involves a
sequence of decisions, plans, and actions that are influenced by factors like flight cancellations and traffic delays.

Flight Decisions Transportation to Airport Unpredictable Factors

Choosing flights, dates, and airlines. Planning how to get to the airport. Dealing with flight cancellations, traffic
delays, and other unexpected events.