Thema: Artificial Intelligence

Module 12.1: Constructing Intelligence

Topic 1. Introduction to Artificial Intelligence
Artificial Intelligence (AI) can be described as the development of digital systems endowed with
the intellectual processes that are characteristic to humans, such as the ability to reason, and
learn from experience to perform reiterative and time-consuming tasks with ease.

From a coined term as a component of science fiction to a mainstream in technology in the

present-day, AI is an integrated part of the common every-day human activities, especially in
urban cities. AI’s can be used to prevent cybercriminal theft, robotics for surgery and
manufacturing industries, and even improve diagnostics for pathologists when analyzing
tissues. At present, there is still a great potential yet to be unveiled.

This topic will introduce students to the concepts of Artificial Intelligence and highlight its
growing influence in the modern technological age of today. We will also we will define and
contrast Human Intelligence and AI, the types and applications of AI, and the basics of AI
development.
Key Points:
❖ Artificial Intelligence is the construction of Intelligence with the aim to replicate or
“mimic” some or almost all aspects of Human Intelligence.
❖ AI primarily aid in assisting humans to perform simple to complex tasks such as
problem identification and solving.
❖ Machine Learning is an important branch of AI that is commonly used and are of
three types: Supervised, Unsupervised and Reinforcement Learning.

1.1 Definition
1.1.1 Human Intelligence
Intelligence (Human Intelligence) is generally referred to as the ability to understand
and adapt to the environment by using inherited abilities and learned knowledge. It
refers to the intellectual capability that allow us to think, learn and adapt to situations
and experiences, understand complex concepts, apply logic and reason, solve
mathematical problems, recognize patterns, and communicate amongst fellow human
beings. This is what enables mankind as a species to be self-sufficient and has allowed
us to thrive for generations on end.
1.1.2 Artificial Intelligence
Artificial Intelligence is the construction of Intelligence with the aim to replicate or “mimic” some
or almost all aspects of Human Intelligence. This is done by coding and/or programming
machines, imbuing them with the ability to learn from experience and historical data, analyze
their surrounding environments, and execute the most appropriate and effective course of
action to accomplish a set of given tasks or instructions. In essence, these machines are
created to “think intelligently” like a human. For this reason, AI can also be referred to as
Computational Intelligence.

Intelligence, whether human or artificial, is goal-oriented: the behavior of which to achieve

something. Through performing and achieving a set of tasks or instructions, intelligence is
utilized.

Figure 1. Human Intelligence (left) and Artificial Intelligence (right)

 Activity 1: Classroom Discussion
1. Differentiate Human Intelligence verses Artificial Intelligence and list some pros and
cons.
1.2 Goals for Artificial Intelligence
The overall research goal of AI is to create technology that allows computers and
machines to function in an intelligent manner. To achieve this goal, AI studies patterns
of the human brain and analyzes the cognitive process to develop intelligent software
and systems. These patterns are generally divided into particular traits and capabilities
that AI developers and researchers expect an intelligent system to display. These
include some of the following:

1.2.1 Perception
Perception is the ability of the computer or machine to make decisions based on the
different percepts that allow for it to interact with its surroundings. Essentially, any
intelligent agent - whether artificial or biologically created - perceives its environment
through sensors and acts upon that environment through actuators. This can be
explained in Figure 2 below.
Figure 2. Basic illustration of how agents interact with the environment through sensors and actuators
(Source: Artificial Intelligence: A Modern Approach [4th Edition] by Stuart Russel, Peter Norvig, Chapter
2: Intelligent Agents).

A human agent has eyes, ears, a tongue and other organs for sensors and hands,
legs, vocal tract, and others for actuators. A robotic agent might have cameras and
infrared range finders for sensors and various motors for actuators. A software agent
receives file contents, network packets, and human input
(keyboard/mouse/touchscreen/voice) as sensory inputs and acts on the environment
by writing files, sending network packets, and displaying information or generating
sounds.
1.2.2 Knowledge Representation
AI research revolves around knowledge representation and knowledge engineering.
Many of the problems that machines are supposed to address should encapsulate a
broad understanding of the world and how it relates all these different types of
knowledge. It will try to relate objects and their structures, attributes, categories, and
interactions between them; situations, events, states, and time; causes and
consequences; knowledge about knowledge (what we know about what other people
know); and many more.

1.2.3 Planning and Navigation

Intelligent agents must be capable of setting and achieving objectives. They require a
means of visualizing the future or a representation of the state of the world that allows
them to make predictions about how their actions will alter it. This will enable better
decision making that maximize the utility (or "value") of available options.

This can be as simple as travelling from an initial point to a destination (Point A to Point
B) such as Google’s self-driving Toyota Prius or as difficult as calculating, predicting,
and analyzing a human body’s entire protein shape, folding and structure.
1.2.4 Logical Reasoning
AI programs enable computers to perform tasks that solve simple to complex
problems. Algorithms are required to imitate the step-by-step process and reasoning
that human use when trying to solve puzzles or make logical deductions. Algorithms for
tough problems can require large computing resources in which most face a
"combinational explosion". This term is used when the amount of memory or computer
time required for issues of a given size becomes astronomical.

On February 10, 1996, IBM's Deep Blue computer defeated former world champion
Garry Kasparov in a chess match, displaying the “combinational explosion” capabilities
of AI technology.
1.2.5 Natural Language Processing
Natural Language Processing (NLP) is a fundamental branch of AI that forms the
relationship between computers and human language. More specifically, NLP is how
computers understand, analyze, manipulate, and/or create the natural language of a
human.

Natural language basically refers to speech analysis in both audible and written
form/speech such as spoken words or inscribed texts of any human language. NLP
systems capture meaning from them as input of words (i.e., sentences, paragraphs,
pages, etc.) in the form of a structured output (which varies greatly depending on the
application).

The main goal of NLP is for computers to achieve human-like comprehension of

texts/languages. This will allow computer systems to understand, draw inferences from,
summarize, translate, and generate accurate and natural human texts and languages,
ultimately bridging understanding between humans and computer systems. (Editors of
expert.ai, 2016)
1.2.6 Emergent Intelligence
Emergent Intelligence is the intelligence that is not explicitly programmed but emerges
from the rest of the specific AI features. This goal is purposed for machines to exhibit
emotional intelligence and moral reasoning. This enables them to become socially
intelligent to recognize, understand, process, and imitate human emotions.

This type of Intelligence also attempts to enable them to possess “Artificial Creativity”
or “Intuition” where it is both theoretical (philosophical and psychological perspective)
and practical (specific implementation of systems that generate novel and useful
outputs).

Activity 2: Student Task

1. Draw an illustration that links the following traits and characteristics to artificial
intelligence and how they related to human intelligence: (1) Perception, Knowledge
Representation, Planning and Navigation, Logical Reasoning, Natural Language
Processing, and Emergent Intelligence.
1.3 Types of Artificial Intelligence
There are two types of AI which are divided based on their capabilities and functions.
(See Figure 3).

Figure 3: A flow diagram hat explains the types of AI

1.3.1 Artificial Intelligence Type 1: Capabilities
1.3.1.1 Weak/Narrow
This type of Artificial Intelligence is able to perform a dedicated task with intelligence. It
is the most common and currently available type of AI. It is trained for performing mostly
one specific type of task and therefore cannot perform beyond its field or limitations.

Narrow AI can fail in unpredictable ways if it goes beyond its limits. For this reason, it is
termed as a weak AI.

Virtual Assistant AI’s such as Apple’s Siri is a good example of a Narrow AI, but it
operates with a limited pre-defined range of capabilities. Other examples include playing
chess, purchasing suggestions on e-commerce site, self-driving cars, speech
recognition, and image recognition.
 Activity 3: Student Task
1. List some 5 more examples from your understanding of Weak AI
2. List 3 weak AI that you have been using daily and explain how they have made
tasks easier.
1.3.1.2 General
This type of Artificial Intelligence is an intelligence that could perform any intellectual
task with efficiency like a human.

The basic understanding behind General AI is to make an intelligent system that could
be smarter than the Weak AI and think like a human on its own. Currently, there is no
such system that exists which could be classified as general AI and can perform any
task as perfect as a human.

AI researchers are now focused on developing machines with General AI. This will
require tremendous effort and time to develop such a system.
1.3.1.3 Super
Super AI is perceived to be the pinnacle of all intelligent systems where a machine
surpasses human intelligence, performing any task better than a human in practically
every field. It is an outcome of general AI.

Super AI’s are basically described as the better versions of humans, possessing the
ability to think, to reason, solve complex problems, make judgments, plan, learn, and
communicate on its own.

Super AI is still a hypothetical and science fictional concept where its development for
the real world is still a science fiction.
Where are we?
Generally, in terms of Artificial Intelligence development, we have only mastered
developing Weak AI (See Figure 4). However, there have been

Figure 4. The stage at which we are with AI technology development.

 Activity 4. Classroom Discussion
1. Group Students
2. Have each Group discuss whether or not we should be advancing AI technology
from Weak AI to General AI, and eventually to Super AI.
3. Have each Group discuss what to look forward to and look out for regarding Super
AI. Give an examples with reasons.
4. Have students declare their discussions in class.
1.3.2 Artificial Intelligence - Type 2: Functions

1.3.2.1 Purely Reactive AI

Purely reactive AI’s only focus on current scenarios and react to achieve the best
possible action or outcome. It the most basic and common type of Artificial Intelligence.
These machines do not possess any memory or ability to store data from past
experiences to work with and therefore specialize in just one field of work.

Example: IBM’s Deep blue system is an example of a purely reactive AI as it observes

the moves made and makes the best possible decision to win in chess.
1.3.2.2 Limited Memory
These machine and computer systems collect previous data and continue adding it to
their memory. Although they have enough memory or experience to make proper
decisions, their memory is at minimal. For example, this machine can suggest a
restaurant based on the location data that has been gathered.

Limited memory machines can store past experiences or some data for a short period
of time. Some examples include predictive texts in smartphones, and suggestions
made by google when searching based what’s trending or on your browsing history.

Example: Self-driving or autonomous cars are another example of a Limited Memory

system. These cars can store recent speed of nearby cars, the distance of other cars,
speed limit, and other information to navigate the road.
1.3.2.3 Theory of Mind
Theory of Mind AI should be able to understand the human emotions, people, beliefs,
and be able to interact socially like humans. A machine or computer system equipped
with Theory of Mind AI is believed to better understand the entities it interacts with.

This type of AI can be incorporated into any developed AI technologies such as speech
and facial recognition technologies that can identify a worker’s social wellbeing:
whether he is stressed, emotionally unstable, finds the work environment unfavorable
for working, and/or other social factors that can affect his current rate of productivity.
These detections can be identified by the system to correct and alert proper authorities
to take appropriate actions that influence and improve the employee’s productivity.

This type of AI machines is still in development, but researchers are making lots of
efforts and improvement for developing such AI machines.
Self-Aware
Self-awareness AI is referred to as the future of Artificial Intelligence. These machines
will be super intelligent, possessing their own consciousness, exhibiting sentiments or
emotions. In essence, they will become self-aware and smarter than the human mind.
Self-Awareness AI does not exist yet and remains a hypothetical concept.
1.4 Branches of Artificial Intelligence
AI is composed of six main branches: Machine Learning, Deep Learning, Robotics,
Expert Systems, Fuzzy Logic, and Natural Language Processing

Figure 4. Branches of Artificial Intelligence

1.4.1 Machine Learning
Machine Learning enables computers and machine systems to translate, process,
analyze and interpret data with the aim of providing solutions for real-life problems.
Essentially, Machine Learning allows computers to operate without the need for
programming: enabling them to learn from past occurrences/experiences as well as a
sufficient amount of stored and presented data. By learning from their experiences,
they use algorithms that predict and improve the probabilities of outcomes in the
execution of tasks. For this reason, Machine Learning is the most demanding branch of
AI.

Technology discoveries such as web search, face recognition, and virtual personal
assistants such as Amazon’s Alexa and Apple’s Siri are all results of Machine Learning.

There are primarily three types of Machine Learning: Supervised, Unsupervised, and
Reinforcement Learning.
1.4.1.1 Supervised Learning
This type of Machine Learning uses labeled data to train Machine Learning models. In
labeled data, the output is already known. The model essentially maps the inputs or
features to the respective outputs. This means that you already know the target
variable.

An example of Supervised Learning is to train a system that identifies whether an e-

mail is a spam or not by identifying the key words and pop-ups advertisements
contained within the e-mail.

Supervised Learning algorithms are generally used for solving classification and
regression problems. Few of the top Supervised Learning applications are weather
prediction, sales forecasting, and stock price analysis.
1.4.1.2 Unsupervised Learning
The Unsupervised Learning technique does not require any supervision to build and
train its models. Essentially, they learn on their own and predict the output.

An example of an Unsupervised Learning technique is to train the model that uses

images of vehicles to classify if it is a bus or a truck. The model learns by identifying the
parts of a vehicle, such as a length and width of the vehicle, the front, and rear end
covers, roof hoods, the types of wheels used, etc. Based on these features, the model
classifies if the vehicle is a bus or a truck.

Unsupervised Learning is used for solving clustering and association problems. One of
the applications of Unsupervised Learning is customer segmentation. Based on a
customer’s behavior, likes, dislikes, and interests, you can segment and cluster similar
customers into a group. This can allow you to be more efficient in your product sales or
marketing strategies.
1.4.1.3 Reinforcement Learning
Reinforcement Learning trains a machine to perform certain suitable actions that
maximize its outcomes in a particular situation. It uses an agent that has an initial or
start and ending state. However, there are different paths to reach the end state, much
like a maze. Reinforcement Learning methods are similar to Unsupervised Learning
methods, being that they do not require any external supervision to train its models.

An example of Reinforcement Learning is to train a machine that can identify the

different shapes, given a list of different objects. The model tries to predict the shape of
the object, which can be either a square, rectangle, triangle, hexagon etc.

Reinforcement learning algorithms are widely used in gaming industries to build games
and for GPS. It is also used to train robots to do certain human tasks.
1.4.2 Deep Learning
Deep Learning is a branch of AI and is a sub-field of Machine Learning. It is an
extremely sophisticated approach that overcomes the “machine-like behavior” of
Machine Learning by constructing an artificially complex and multi-layered “neural
network”. This complex network allows for data to pass through nodes like neurons in
the human brain. Deep Learning draws its inspiration from the human brain in
resemblance to the function of the neurons inside and for this reason, it is often
referred to as Artificial Neural Networks.

Deep Learning is a key technology behind driverless cars. Powered by GPU

technology on the cloud and converging with massive reserves of data, neural
networks are used today by Internet companies such as Amazon to provide product
recommendations, and by Facebook to identify faces in photos.
 Figure 5. Artificial Intelligence, Machine Learning and Deep Learning
(Source: https://flatironschool.com/blog/deep-learning-vs-machine-learning)
1.4.3. Robotics
Robotics is the branch of AI that employs scientific and engineering methods to
design, construct, and operate intelligent robots. Robots are often deployed for
conducting tasks that might be difficult for humans to perform steadily. They often
imbue Machine Learning and Deep Learning models to function.

The applications of Robotics are versatile. Some applications include the assembly for
automobile manufacturing, underwater and space expeditions (see applications of AI
below), and agricultural practices such as seeding, weed and pest control, and
harvesting. Robots like drones can also be used in military applications such as in
surveillance recons for attacking or supplying Medicare to friendly forces.
1.4.4. Expert Systems
As its name suggests, this branch of AI refers to computer systems that mimic the
decision making intelligence of a human expert. Essentially, expert systems are
interactive decision-making systems that contain a wealth of knowledgeable database.
This database is pertained to a particular domain to solve complex decision-making
problems (in that domain) that would require human experts to resolve.

Expert Systems can find applications in healthcare where it can identify various
bacteria that could cause acute infections and recommend drugs based on the
person’s weight. It can also predict the degree and type of lung cancer.

Other applications include mineral, oil and gas prospecting, genetic engineering, and
even in businesses for corporate planning, scheduling and financial management.
1.4.5. Fuzzy Logic
Fuzzy logic is a branch of AI that employs a technique that represents and modifies
uncertain information by measuring the degree to which the hypothesis is true.
Essentially, it uses standard logic to identify whether a condition or concept exhibits a
degree of truth: If is completely true, standard logic is 1.0, if it is completely false, the
standard logic is 0.0. However, there are cases where a concept can either be partially
true or partially false.

Fuzzy logic is convenient and flexible to implement Machine Learning techniques and
assist in imitating human thought logically. In essence, an intelligent agent or computer
system is given the ability to resolve the fuzzy logics that humans face in their day-to-
day activities with the aim of finding a solution.

In these cases, fuzzy logic can have application in automobile industry particularly in
automatic gearboxes, healthcare for medicine decision making and even washing
machines where they can predict how heavy the wash load is and determine the
correct amount of detergent and water, speed of agitation, and duration of wash cycles.
1.4.6. Natural Language Processing
A computer communicates information using binary digits. When it wants to
communicate and interact with a human, it uses Natural Language Processing (NLP).

NLP basically conducts searches, analysis, understands and derives information from
the text form of data. In order to teach computers how to extract meaningful information
from the text data, NLP libraries are used by programmers.

A common example of NLP is spam detection where computer algorithms can check
whether an email is a scam or not just by looking at the subject of a line, or text of an
email.

NLP is also useful in companies that handle with texts everyday such as Facebook or
Twitter. Twitter uses NLP technique to filter terroristic language from various tweets.
Amazon implements NLP for interpreting customer reviews and enhancing their e-
shopping experience. Other applications involve text translation, sentiment analysis,
and speech recognition.
 Activity 5: Student Task

1. Students are to draw a visual representation (Flow Diagram, schematics, etc) of all
six branches of AI and how they all work together.
1.5 Applications of Artificial Intelligence
There are many applications of AI that is utilized today. Some of these are listed below:

1.5.1 Navigation - Autonomous Vehicles

AI can be used to build self-driving vehicles which will revolutionize the modern modes
of transport systems either for land, sea, or air.

For land vehicles, AI can be used along with the vehicle’s camera, radar, cloud
services, GPS, and other control signals to operate the vehicle. Together, they can
improve the in vehicle experience and provide additional systems like emergency
braking, blind-spot monitoring, and driver-assist steering.
1.5.2 Health Care
AI is also used in the healthcare sector for many diverse applications. In this industry,
the AI utilizes a sophisticated machine imbued with a combination of historical data and
medical intelligence that can analyze and detect diseases or signs of chronic
conditions that ensure early diagnosis, identify cancer cells, and even discover new
drugs. The reliance upon AI technology in the medical healthcare sector is saving a lot
of lives.

Cambio Health Care, a healthcare organization based in Linköping, developed a

clinical decision support system for stroke prevention. It can signal a warning to the
physician when a patient is at risk of having a heart stroke. Another company called
Coala life created a digitalized device that can discover and detect cardiac diseases.
1.5.3 Robotics
Robotics is a commonly used field for AI applications. Robots powered by AI use real-
time updates to sense obstacles in its path and pre-plan its journey instantly.
Applications for robotics are vast.

They can be used for carrying goods in hospitals, factories, and warehouses, cleaning
offices and large equipment as well as Inventory management for the aforementioned
industries.

Activity 6: Group Discussion

1. Discuss in groups which AI applications would be important in PNG.

1.5.4 AI Development
Regardless of AI being made from either machines or human developers, there are six
general steps to developing an AI:

Step 1: Identify the Problem

AI is purposed to perform and automate an intellectual task to solve a problem. It is
therefore essential to firstly identify the problem. The most appropriate questions to ask
are:
• What are you attempting to solve for?
• What is the desired outcome?
Step 2: Prepare the Data
In order to solve the issue at hand, data is vital in order to execute the most appropriate
course of action. Data is generally divided into two categories: structured and
unstructured.

Structured data contains a rigid format to ensure consistency in processing and also
ease of analytics. This can include information such as customer record with a first
name, last name, date of birth, address, etc.

Unstructured data is everything else that is not mentioned: it is data that is maintained
in the not uniformed pattern. It can include audio, pictures, words, and infographics.
Such examples include emails, a phone conversation, and a WhatsApp message.
Step 3: Choose an Appropriate Algorithm
The next step involves choosing the method of learning or Algorithm for the AI system.
The types of algorithms would be different, and we could classify them in several
different categories such as clustering where the algorithm tries to group objects
together, association when it finds links between objects, dimensionality reduction
where it reduces the number of variables to decrease the noise and many others.

There are many different types of algorithms that are dependent on the type of
learning that you choose. Once we have identified the problem, we can select the
most appropriate algorithm.

For example, if it is a classification problem, Supervised Learning algorithms are used.

Fundamentally, classification is about predicting and assigning a label. Classification
problems include e-mail spam detection, plant species classification and medical
diagnostics tests.
Step 4: Train the Algorithms
Once an appropriate algorithm is chosen, the model then needs training where a
series of data are placed as inputs into the model. In this step, model accuracy is
required. While there is not a widely accepted or internationalized threshold, it is vitally
important to establish model accuracy within your selection framework. Setting a
minimum acceptable threshold and applying a great statistical discipline is key and
fine-tunes the model.

Step 5: Choose a particular Programming Language

There are many programming languages that can create your AI of choice. C++ or C,
Java, Python and R are some examples.

Python and R are the more popular coding languages as they offer a strong set of
tools including extensive Machine Learning libraries to the users.
Step 6: Run on a Selected Platform
The final step is to choose a platform that provides all of the services instead of buying
your own service, database, etc. Platforms typically help with such issues as data
preprocessing, model training, and evaluation prediction.

There are many platforms that are readily available, particularly Machine Learning as a
service, and has been one of the most useful pieces of infrastructure that has assisted
in the widespread of Machine Learning. This is because these platforms are built to
simplify and facilitate Machine Learning and offer cloud-based advanced analytics
which can be used with and incorporate multiple algorithms and multiple languages.

The most popular platforms include Microsoft Azure Machine Learning, Google Cloud
Prediction API, and TensorFlow.
 Activity 7. Group Debate
Students to debate the dilemma of the following topic: Artificial Intelligence – is it good
or bad for society?
Activity 8. Class Assessment - Quiz
Refer to the Handout