CS 611: ARTIFICIAL

INTELLIGENCE

B. I. Ya’u, August 26, 2021

 Introduction
• Humans are good at solving problems and making decisions.

• A human being is capable of making decisions on his/her own.

• That is why we say that human beings are intelligent.

• This intelligent behavior of human beings can be borrowed and implemented in computers and machines.

• We can design and develop computer systems and machines capable of exhibiting the intelligent behavior of
human beings.

• This is known as artificial intelligence.

• Artificial intelligence has revolutionized a number of sectors across the world.

• Transportation

• Health

• Security

• Agriculture
 What is Artificial Intelligence?
• Artificial intelligence is made up of two words, artificial and intelligence.

• The word artificial means man-made, while the word intelligence means thinking power.

• From this, we can say that artificial intelligence is man-made thinking power.

• Artificial intelligence is a branch of computer science that deals with the development of computers or
machines that are as intelligent as human beings.

• Artificial intelligence involves studying how the human brain thinks, how humans learn, make
decisions and work when solving problems.

• The outcomes of such a study are then used for the development of intelligent software and systems.

• We say we have artificial intelligence when we have a machine that exhibits human-like characteristics
such as reasoning, learning, and problem solving.

• In artificial intelligence, a machine doesn’t have to be preprogrammed so as to do some task, but one
can create a machine with programmed algorithms that work based on own intelligence.

3
 Goals of AI
The following are the goals of artificial intelligence:
• To replicate human intelligence- AI is geared towards replicating human intelligence into machines for problem
solving.

• Solving knowledge-intensive tasks- human beings can be overwhelmed by tasks that are knowledge-intensive.
Such a task can be performed using intelligent machines and systems.

• Intelligent connection of perception and action- with artificial intelligence, we connect perceptions and actions.

• Building machines that can perform tasks that require human intelligence- artificial intelligence is a great tool as
it helps us develop machines that can be used to perform tasks that require human tasks to be done. Examples
of such problems include playing chess, solving a theorem, driving a car in traffic and performing surgery.

• Create systems capable of showing intelligent behavior- artificial intelligent-powered machines can show
intelligent behavior such as learning new things on their own and advising the user.
 Components of AI
Artificial intelligence is an inter-play of the following disciplines:

• Computer Science

• Mathematics

• Psychology

• Biology

• Sociology

• Neurons Study

• Statistics
Intelligent
Systems
 What are Intelligent Systems?
• Intelligent systems are machines that are technologically advanced to perceive and
respond to their surrounding environment.

• Intelligent systems have revolutionised a number of industries such as security,

manufacturing, transportation, and logistics.

• They have helped in improving the quality, energy efficiency, and flexibility of the systems.

• One of the ways through which intelligent systems perceive their environment is through
vision.

• The study of how computers can interpret visual information started in the 1950s. Since
then, it has become a very powerful technology and has been implemented in commercial,
industrial and government sectors

• The field of intelligent systems is also concerned with studying how the systems interact
with humans to bring changes to the dynamic social and physical environments.
 Applications of Intelligent Systems
Intelligent systems are applied in the following fields:
• Factory automation

• Assistive robotics

• Field and service robotics

• Military applications

• Education

• Entertainment

• Medical care

• Visual inspection

• Character recognition

• Visual surveillance

• Human identification using biometric modalities such as the face, iris, fingerprint, hand.

• Intelligent transportation.
 Challenges Facing Intelligent Systems
• Uncertainty- The physical sensors/effectors only provide limited, inaccurate
and noisy information/action.

• Dynamic world- the world we live in changes dynamically.

• Time-consuming computation- the process of searching for the optimal path

that leads to the goal involves an extensive search done within a large space.

• Mapping- much information is lost during the transformation from a 3D to a

2D world. Computer vision has to deal with problems such as changes in

• perspectives, background clatter or motion, lighting and scale, and the

groupings of items based on intra/inter class variation.
Intelligent Agents and
Environments
 What is an AI Agent?

• An artificially intelligent system is made up of both the

agent and the environment.

• Every agent acts in its own environment, and this

environment may have other agents.

• An agent refers to anything capable of perceiving its

environment through sensors and acting the same
environment through effectors.

• For example, a human agent has sensory organs like

ears, eyes, tongue, nose and organs that are parallel to
the sensors like the skin, hands, legs etc. A robotic
agent has a camera and infrared range finders as
sensors and motors and actuators as the effectors.
 Types of AI Agents

• Simple Reflex Agent

• Model-based Reflex
Agent

• Goal-based Agents

• Utility-based Agent

• Learning Agent
 Simple Reflex Agent
• These are the simplest forms of agents. Their
decisions are based on the current percepts
and ignore percept history.

• These types of agents can only survive in a

fully observable environment. However, when
making decisions or taking actions, it does
not consider the history of percepts. It works
based on the condition-action rule, meaning
that it simply maps the current state to its
corresponding action.

• Simple reflex agents are very limited in

knowledge and they are not adaptive to the
environment.
 Model-based Reflex Agents
• This type of agent is capable of working in a partially
observable environment and track the situation. The
model-based reflex agent is made up of the following
two important factors:

• Model- This is knowledge regarding how things

are done in the world, hence, it is referred to as
a model based agent.

• Internal state- this represents the internal state

based on the history of percepts.

• The agents have a model representing the knowledge

of the world. Actions are performed based on this
model. For the agent state to be updated, the
following information is needed:

• How the world evolves.

• The effect of the agent’s actions on the world.

 Goal-based Agents
• Knowledge about the current state of the
environment is not enough for an agent to decide on
what to do.

• The agent should know its goal which states its

desirable situations.

• A goal-based agent expands the ability of a model-

based agent by including the goal information. They
choose the action to perform based on a need to
achieve a goal.

• In some cases, these types of agents have to

evaluate a long sequence of actions so to know
whether a goal will be achieved or not.

• Different scenarios have to be considered, and this

process is known as searching and planning and it
makes an agent be proactive.
15
 Utility-based Agents
• These types of agents are similar to the goal-
based agents but they provide an extra
component of utility measurement.

• This makes them unique in that they have a way

of measuring success at any given state.

• This means that utility-based base their actions

on how to achieve a goal as well as how best to
achieve the goal.

• It is good to type of agent when there is a

number of alternatives, and the agent has to
make a choice so as to perform the best action.

• The utility function works by mapping every

state to a real number in order to check how
efficiently each action achieves the goal.
 Learning Agents
• In artificial intelligence, a learning agent refers to an agent
capable of learning from its experience, or it has learning
capabilities. The agent begins to act depending on basic
knowledge, then it begins to learn and act based on learning.
It is made up of the following conceptual components:

• Learning element- this is the element that is

responsible for making improvements by learning from
the environment.

• Critic- The learning element has to receive feedback

from the critic which is the component that tells how
well the element is doing with respect to a particular set
performance standard.

• Performance Element- this is the component that

selects the external action.

• Problem Generator- this is the component that

suggests actions that may create new experiences.
 Agent Environment
• An environment refers to everything in the world that surrounds an agent, but it’s not part of the
agent. An environment is a situation in which an agent is present. It is where the agent lives, where
it operates and it provides the agent with something to sense and act upon.

• Fully observable vs Partially Observable

• Static vs Dynamic

• Discrete vs Continuous

• Deterministic vs Stochastic

• Single-agent vs Multi-agent

• Episodic vs Sequential

• Known vs Unknown

• Accessible vs Inaccessible
Fully observable vs Partially Observable

• A fully observable environment is the one in which an agent can access or

sense the complete state of the environment at every point of time.
Otherwise, the environment is partially observable.

• A fully observable environment is easy since there is no need to keep the

internal state and keep a track history of the world.

• If an agent does not have sensors in all environments, the environment is said
to be unobservable.
 Deterministic vs Stochastic

• A deterministic environment is the one in which the current state and

selected action of an agent can completely determine the nest state of the
environment.

• This is not the case with a stochastic environment as it is random and is

not completely dependent on an agent.

• If the environment is deterministic and fully observable, the agent doesn’t

need to be worried about uncertainty.
 Episodic vs Sequential

• An episodic environment is made up of a series of one-shot actions, and

the agent requires only the current percept for the action.

• In a sequential environment, the agent is in need of the memory of its past

actions to determine its next actions.
 Single-agent vs Multi-agent

• A single agent environment is one in which we have only a single agent

operating by itself.

• A multi-agent environment is one in which we have multiple agents

operating in the environment.

• The agent design problems are different in the two environments.

 Static vs Dynamic
• A dynamic environment is one that can change itself while the agent is
deliberating. Otherwise, the environment is referred to as be static.

• It is easy to deal with static environments because an agent is not required

to be looking or observing the environment while choosing an action.

• However, for the case of dynamic environments, the agent will have to
keep on looking at the environment at every action.

• An example of a static environment is the Crossword puzzle.

• An example of a dynamic environment is driving a taxi.

 Discrete vs Continuous

• A discrete environment is one in which there is a finite number of percepts

and actions that can be performed within it.

• Otherwise, it is called a continuous environment.

• For example, a chess game is an example of a discrete environment as

there is only a finite number of moves that we can make.

• However, a self-driving car is a good example of a continuous

environment.
 Known vs Unknown

• These two are not features of the environment, but they refer to the state
of knowledge of the agent to perform a certain action.

• In a known environment, the agent knows the results of every action. In an

unknown environment, the agent is expected to learn how to work in order
to perform an action.

• It is possible for a known environment to be partially observable and an

unknown environment to be fully observable.
 Accessible vs Inaccessible
• An accessible environment is the one in which an agent is capable of
obtaining complete and accurate information about the state of the
environment.

• Otherwise, the environment is known as inaccessible.

• A good example of an accessible environment is an empty room whose

state we can define by its temperature.

• A good example of an inaccessible environment is information regarding

an event on the earth.
End of the Slides