DEFINITION OF AI

AI is a branch of computer science that focuses on building intelligent

machines that function and respond just like people. It enables computers
and machines to simulate human intelligence and problem-solving
capabilities.

This generally involves borrowing characteristics from human

intelligence, and applying them as algorithms in a computer friendly way.

Artificial intelligence is about designing systems that are as

intelligent as humans. This view involves trying to understand human
thought and an effort to build machines that emulate the human thought
process.

It is also concerned with the design of intelligence in an artificial

device.

AI is generally associated with Computer Science, but it has many

important links with other fields such as Mathematics, Psychology,
Cognition, Biology and Philosophy, among many others.
DEFINITION OF AI
The term was coined by McCarthy in 1956. There are two ideas in the
definition.

 Intelligence
 Artificial device

What is intelligence? Is it that which characterize humans? Or is there

an absolute standard of judgment?

Accordingly there are two possibilities:

 A system with intelligence is expected to behave as intelligently as a

human
 A system with intelligence is expected to behave in the best possible
manner.

Secondly what type of behavior are we talking about?

 Are we looking at the thought process or reasoning ability of the

DEFINING THE SCOPE AND VIEW OF ARTIFICIAL INTELLIGENCE

Given this scenario different interpretations have been used by different

researchers as defining the scope and view of Artificial Intelligence.

Rationality refers to the ability to make decisions and take actions

based on logical reasoning or a purposeful thought process.
DEFINING THE SCOPE AND VIEW OF ARTIFICIAL INTELLIGENCE
Thinking humanly: The cognitive modeling approach

As the name suggests, this approach tries to build an Artificial

Intelligence model based on Human Cognition. Within computer science,
there is a field called Cognitive Modeling that deals with simulating the
human thinking process.

This view involves trying to understand human thought and an effort to

build machines that emulate the human thought process and behaviors, such
as perception, reasoning, and problem-solving.

Cognitive modeling is used in a variety of AI applications such as deep

learning, expert systems, Natural Language Processing, robotics, and so
on.
DEFINING THE SCOPE AND VIEW OF ARTIFICIAL INTELLIGENCE
Acting Humanly: The Turing Test

Alan Turing, proposed the Turing Test to provide a definition of

intelligence.
This approach aims to create machines that mimic human behavior (human
actions, emotions, and interactions).

He defined intelligent behavior as the ability to achieve human-level

intelligence during a conversation. In the Turing Test, a computer and
human are asked questions by a human interrogator to determine which is
human.

If a machine can engage in a conversation with a human without being

detected as a machine, it has demonstrated human intelligence.

The Turing Test approach has led to the development of a wide range of AI
technologies, including chatbots, virtual assistants, and recommendation
engines.
DEFINING THE SCOPE AND VIEW OF ARTIFICIAL INTELLIGENCE
Thinking rationally: The “laws of thought” approach

Rationality refers to doing the right thing in a given circumstance. A

system is rational if it does the right thing.

Thinking rationally, which follows the laws of thought approach aims to

develop AI systems that can reason logically and making correct
inferences based on logical notation or a set of predefined rules.

For example, if we know that Samson is a man and that all men are good,
we can conclude that Samson is a good man. This approach of using logical
notations and inferences is appealing because it mirrors how our own
minds work.

However, this approach faces two major obstacles. Firstly, it is

challenging to represent uncertain knowledge in formal terms required by
logical notation. Secondly, even with programs capable of drawing
conclusions from given premises, the computational resources required can
quickly become overwhelming.
DEFINING THE SCOPE AND VIEW OF ARTIFICIAL INTELLIGENCE
Acting Rationally: The Rational Agent Approach

An alternative approach to AI is acting rationally, which focuses on

building rational agents that operate autonomously, perceive from the
environment, persist over time, adapt to change, and pursue goals.

Acting rationally, also known as the rational agent approach, is an

approach to artificial intelligence (AI) and machine learning that
focuses on creating intelligent agents that can act in the world to
achieve their goals.

Rational action involves the translation of these rational thoughts into

physical actions that bring about the desired results.

This requires the AI system to have the necessary capabilities, such as

actuation, communication, and control, to interact with its environment
in a way that implements the decisions made during the rational thinking
process.
TYPICAL AI PROBLEMS
While studying the typical range of tasks that we might expect an
intelligent entity to perform, we need to consider both commonplace tasks
and expert tasks.

Examples of common-place tasks include

 Recognizing people, objects.
 Communicating (through natural language).
 Navigating around obstacles on the streets

These tasks are done matter of firstly and routinely by people and some
other animals.

Expert tasks include:

 Medical diagnosis.
 Mathematical problem solving
 Playing games like chess

These tasks cannot be done by all people, and can only be performed by
skilled specialists.
TYPICAL AI PROBLEMS

However, when we look at what computer systems have been able to achieve
to date, we see that their achievements include performing sophisticated
tasks like medical diagnosis, performing symbolic integration,
Mathematical problem solving and playing chess.

On the other hand it has proved to be very hard to make computer systems
perform many routine tasks that all humans and a lot of animals can do.
Examples of such tasks include navigating our way without running into
things, catching prey and avoiding predators.

Humans and animals are also capable of interpreting complex sensory

information. We are able to recognize objects and people from the visual
image that we receive. We are also able to perform complex social
functions.
INTELLIGENT BEHAVIOR
AI is a branch of computer science concerned with the programming of
computers so that they exhibit apparently intelligent behavior.

Intelligent behavior as an adaptation to the task environment, refers to

the ability of a system to learn from its environment, adapt its actions
based on past experiences, and make decisions that lead to successful
outcomes.

This discussion brings us back to the question of what constitutes

intelligent behavior. Some of these tasks and applications are:

One example of intelligent behavior in machines/computers is the ability

to recognize objects in images. This is accomplished through a technique
called computer vision, which uses algorithms to analyze and interpret
visual data.

Another example of intelligent behavior in computers is natural language

processing (NLP), which enables machines to understand and respond to
human language. This is used in virtual assistants like Siri and Alexa,
INTELLIGENT BEHAVIOR
Learning algorithms are also examples of intelligent behavior in
computers. These algorithms can learn from data and improve their
performance over time without being explicitly programmed. For example, a
spam filter that uses machine learning can improve its ability to
accurately classify emails as spam or not spam based on user feedback
over time.

Reasoning in Artificial Intelligence refers to the process by which AI

systems analyze information, draw inferences appropriate to the
situation, and draw conclusions to solve problems or make decisions. It
is a fundamental cognitive function that enables machines to mimic human
thought processes and exhibit intelligent behavior.

Problem-solving: Identify and solve problems in a goal-oriented way.

 Robotics
FAMOUS AI SYSTEM
IBM Deep Blue, a chess-playing supercomputer that defeated international
grandmaster Garry Kasparov.

Autonomous AI Agents, at their core, are intelligent entities capable of

decision-making and action execution without direct human intervention.
These agents leverage advanced algorithms and machine learning models to
analyze data, draw insights, and execute tasks. A robotic space probes
autonomously monitor their surroundings, make decisions and act to
achieve their goals.

Internet Agents: The explosive growth of the internet has also led to
growing interest in internet agents to monitor users' tasks, seek needed
information, and to learn which information is most useful.

Machine translation: A system capable of translations between people

speaking different languages will be a remarkable achievement.

Autonomous Land Vehicle in a Neural Network: This is a system which

learns to control vehicles by watching a person drive. It contains a
APPROACHES TO ARTIFICIAL INTELLIGENCE
Applied AI, also known as practical or operational AI, is a branch of
artificial intelligence focused on the practical implementation of AI
technologies in real-world scenarios. Its aims to produce commercially
viable "smart" systems such
as, security system that is able to recognize the faces of people who are
permitted to enter a particular building.

Cognitive AI deals with artificially intelligent systems that mimic human

behavior. It uses multiple advanced learning algorithms to learn from its
surroundings with time and improves itself by learning from its
experience.

For example, when a child is born, they don't know anything about the
world. However, over time, they learn and pick up human behavior from
their surroundings. If they perform an undesirable action, they are
punished. Likewise, if they do something good, they are rewarded.
Cognitive AI similarly aims for systems to learn from their environment.
APPROACHES TO ARTIFICIAL INTELLIGENCE
Strong AI, also known as artificial general intelligence or AGI, is
capable of behaving and performing actions in the same ways human beings
can. AGI mimics human intelligence, and is able to solve problems and
learn news skills in ways similar to our own.

The more an AI system approaches the abilities of a human being, with all
the intelligence, emotion, and broad applicability of knowledge, the more
‘strong’ the AI system is considered. Strong AI maintains that suitably
programmed machines are capable of cognitive mental states.

Weak AI is also known as specialized AI or narrow AI is a type of

artificial intelligence that is limited to a specific or narrow area
rather than possess full cognitive abilities like the human brain.

Weak AI holds that suitably programmed machines can simulate human

cognition.
PRACTICAL EXAMPLES OF WEAK/NARROW AI
Just about any AI that you can think of that is used today are examples
of narrow AI. Here are eight practical examples:

Digital voice assistants: A virtual assistant is a tool that uses voice

commands to complete tasks for the user. Siri and Alexa are both
examples.

Chatbots: A chatbot is an online tool that uses text-based communication

to answer questions, provide direction, or complete tasks.

Recommendation engines: Whether Netflix is telling you what movie you

should watch next or Amazon or other retail websites offer you helpful
advice about what you else you might be interested in purchasing, these
recommendation engines are examples of narrow AI.

Search engines: Google and other search engines are also examples of weak
AI. When you type in your question, the algorithm gets to work to run
that question through its vast database to classify it and come back with
answers.
PRACTICAL EXAMPLES OF WEAK/NARROW AI
Autonomous vehicles: AI that allows vehicles to operate without a human
driver is weak AI. The challenge, since this AI doesn’t possess full
cognitive abilities like a human brain, is to program and train the AI
regarding any potential road hazard or situation the vehicle might
encounter.

Image and speech recognition: A critical way narrow AI is making an

impact in healthcare is through image recognition by helping radiologists
detect disease in scans of patients. It’s also in play with speech
recognition and translation services like Google Translate.

Predictive maintenance and analytics: It uses data, algorithms, and

machine learning to examine historical data to construct a prediction of
a likely outcome in the future. In warehouses where heavy machinery is in
use, AI helps identify maintenance issues that need to be addressed
before a machine failure.

Robots: Currently, robots don’t have a mind of their own. Drones and
manufacturing robots operate with narrow AI and are able to complete a
LIMITATIONS OF AI
Today’s successful AI systems operate in well-defined domains and employ
narrow, specialized knowledge. Common sense knowledge is needed to
function in complex, open-ended worlds. Such a system also needs to
understand unconstrained natural language. However these capabilities are
not yet fully present in today’s intelligent systems.

Limited understanding of context: Machines are often trained on large

amounts of text data and are able to identify patterns and make
predictions based on that data. However, they lack the ability to
understand the nuances and subtleties of human language and
communication.

Lack of common sense: AI systems currently lack the ability to apply

common sense reasoning to new situations. They are only able to make
predictions and decisions based on the data they have been trained on,
meaning they are not able to apply their knowledge in a flexible way to
new situations

Lack of Emotion: AI systems do not have the ability to feel emotions.

LIMITATIONS OF AI
Real-Time Learning and Adaptability: The ability to learn and adapt in
real-time to dynamic environments is a distinctive human trait that AI
struggles to replicate. Human cognition allows for continuous learning
and adjustment, whereas AI often requires retraining and significant data
input for adaptation.

Vulnerability to Adversarial Attacks: AI systems are susceptible to

adversarial attacks, where intentional manipulation of input data can
mislead the system’s output. Safeguarding AI against such attacks is an
ongoing challenge, particularly in critical applications like autonomous
vehicles or cybersecurity.

Interpretability and Explainability: The “black box” nature of some AI

models poses challenges in interpreting and explaining their decisions.
Understanding how AI arrives at specific conclusions, especially in
critical areas like healthcare is crucial for acceptance and trust but
remains a complex task.

Data Dependency and Quality: AI’s effectiveness is heavily reliant on the

WHAT CAN AI SYSTEMS NOT DO YET?

List five tasks that computers are unlikely to be able to do in the next five years.

1. Understand natural language robustly (e.g., read and understand in

the newspaper)
2. Common sense reasoning
3. Surf the web
4. Interpret an arbitrary visual scene
5. Learn a natural language
6. Construct plans in dynamic real-time domains
7. Creativity and Innovation
8. Exhibit true autonomy and intelligence
9. Multitask: Most AI systems are highly trained to solve specific
problems.
WHAT CAN AI SYSTEMS DO
Facial recognition: Enabled by computer vision, facial recognition can
detect and identify individual faces from an image containing many
people's faces.

Robotics: We have been able to make vehicles that are mostly autonomous.
A robot is a programmable autonomous machine capable of sensing its
environment, carrying out computations to make decisions, and performing
actions in the real world. The term robotics describes the field of study
focused on developing robots and automation.

Machine Translations: In natural language processing, we have systems

that are capable of using artificial intelligence (AI) and machine
learning algorithms to automatically translate text or speech from one
language to another.

Medical expert systems: This are a class of AI systems that excels at

identifying complex patterns in medical data, leading to more accurate
diagnoses and personalized treatment plans as is done by a medical
expert.
NATURAL LANGUAGE PROCESSING (NLP)
Simply put, it allows computers to understand human language — speech or
text.

NLP is the ability to automatically receive, understand, and operate on

human language in the raw written or spoken form.

A useful feature of most email service providers is the ability to

automatically segregate spam emails away from regular emails. This is a
use case of a popular NLP task known as text classification.

Text classification is the task of assigning one or more categories to a

given piece of text from a larger set of possible categories. In the
email spam–identifier example, we have two categories—spam and non-spam—
and each incoming email is assigned to one of these categories.
INTRODUCTION TO AGENT
INTRODUCTION TO AGENT
 Anything that can be viewed as perceiving its environment through
sensors and acting upon that environment through actuators.

 Agents are entities capable of perceiving their environment using

sensors and taking actions within that environment using actuators.

 In artificial intelligence, an agent is a computer program or system

that is designed to perceive its environment, make decisions and take
actions to achieve a specific goal or set of goals. The agent operates
autonomously, meaning it is not directly controlled by a human
operator.

For example, an ATM is an agent. It either gives or doesn’t give cash

when it receives information about the account details.

Modern cars are equipped with a wide range of sensors that gather data
from the vehicle’s surroundings and internal systems. These sensors
provide crucial information to the car’s AI system, enabling it to make
informed decision. E.g.
INTRODUCTION TO AGENT
 An agent perceives its environment through sensors.
 The complete set of inputs at a given time is called a percept.
 The current percept can influence the actions of an agent.
 The agent can change the environment through actuators.
 Actuators in a car are responsible for
translating the decisions made by the
AI system into physical actions that
affect the car’s movement and
behavior.
 Sensor is a device which detects the
change in the environment and sends
the information to other electronic
devices. An agent observes its
environment through sensors.
 Agent Behavior: It refers to the
actions and responses it exhibits in
its environment based on the
An operation involving an Effector information
is called anitaction.
receives through
Actions canits
be
grouped into action sequences. The sensors.
agent can have goals which it tries to
INTRODUCTION TO AGENT
EXAMPLES OF AGENTS
Human Agent: A human agent refers to a human being who interacts with
their environment. Humans possess sensory organs such as eyes for visual
perception and ears for auditory input, which serve as sensors.
Additionally, hands, legs, and vocal tract function as actuators,
enabling humans to manipulate objects, move around, and communicate.
Human agents rely on their senses to perceive the environment and use
their limbs and voice to effect changes in the world.

Robotic Agent: A robotic agent is an autonomous machine designed to

perform tasks in an environment. It’s equipped with various types of
sensors to collect data from its surroundings. Cameras provide visual
information, infrared range finders measure distances, and Natural
Language Processing (NLP) capabilities allow the robot to interpret and
respond to human speech. Actuators in robots are usually motors that
enable movement, manipulation of objects, and sometimes even complex
tasks. Robotic agents can operate in physically challenging or hazardous
environments without direct human intervention.

Some examples of robots are Xavier from CMU, COG from MIT, Aibo from SONY
EXAMPLES OF AGENTS
Software Agent: A software agent is a program that operates in a digital
environment. Sensors for software agents can include keyboard input,
mouse movements, or even data read from files or network streams. These
agents use these inputs to make decisions and perform actions. The
outputs are usually displayed on a screen, sent over a network, or stored
in files.

Intelligent buildings: The AI-powered smart building will optimise energy

consumption using machine learned automations to adjust heating, lighting
and cooling systems based on real time usage and occupancy.

Autonomous spacecraft: Autonomy is an essential technology for multi-

spacecraft missions. It allows spacecraft to decide their next
activities, as opposed to having the spacecraft send their status to a
control station on the ground and await further instructions.

Medical expert systems: This are a class of artificial intelligence

systems that make treatment recommendations as is done by a medical
expert.
EXAMPLES OF AGENTS
Traffic management agents: These are agents that are designed to manage
traffic flow in cities. They can monitor traffic patterns, adjust traffic
lights, and reroute vehicles to minimize congestion.

Intelligent personal assistants: These are agents that are designed to

help users with various tasks, such as scheduling appointments, sending
messages, and setting reminders. Examples of intelligent personal
assistants include Siri, Alexa, and Google Assistant.

Gaming agents: These are agents that are designed to play games, either
against human opponents or other agents. Examples of gaming agents
include chess-playing agents and poker-playing agents.

Fraud detection agents: These are agents that are designed to detect
fraudulent behavior in financial transactions. They can analyze patterns
of behavior to identify suspicious activity and alert authorities.
Examples of fraud detection agents include those used by banks and credit
card companies.
AGENT PERFORMANCE
An agent function to implements a mapping that connects sequences of
perceptions to corresponding actions.

The behavior and performance of intelligent agents have to be evaluated

in terms of the agent function.

The ideal mapping specifies which actions an agent ought to take at any
point in time.

The performance measure is a subjective measure to characterize how

successful an agent is. The success can be measured in various ways.

 It can be measured in terms of speed or efficiency of the agent.

 It can be measured by the accuracy or the quality of the solutions

achieved by the agent.

 It can also be measured by power usage, money, etc.

INTELLIGENT AGENTS

 The fundamental faculties of intelligence are

 Acting
 Sensing
 Understanding, reasoning, learning

 In order to act intelligently, you must sense. Blind action is not a

characterization of intelligence

 Many robotic agents stress sensing and acting. Understanding not

necessary.

 Understanding is essential to interpret the sensory percepts and

decide on an action.
RATIONAL AGENT

A rational agent in AI is an agent that performs actions to achieve the

best possible outcome based on its perceptions and knowledge. It operates
under the premise of rationality, where it consistently makes decisions
that maximize its expected performance measure.

A rational AI agent is a system that makes optimal decisions to achieve

its goals, while an intelligent agent learns and adapts to its
environment.

Perfect Rationality assumes that the rational agent knows all and will
take the action that maximizes her utility. Human beings do not satisfy
this definition of rationality.

Rational Action is the action that maximizes the expected value of the
performance measure given the percept sequence to date.
RATIONALITY

Rationality is nothing but status of being reasonable, sensible, and

having good sense of judgment.

Rationality is concerned with expected actions and results depending upon

what the agent has perceived. Performing actions with the aim of
obtaining useful information is an important part of rationality.

The rationality of an agent is measured by its performance measure.

Rationality can be judged on the basis of following points:

 Performance measure which defines the success criterion.

 Agent prior knowledge of its environment.

 Best possible actions that an agent can perform.

 The sequence of percepts.

BOUNDED RATIONALITY

Bounded rationality is the idea that rationality is limited when

individuals make decisions, and under these limitations, rational
individuals will select a decision that is satisfactory rather than
optimal.

Limitation, include the difficulty of the problem requiring a decision,

the cognitive capability of the mind, and the time available to make the
decision. Because of the limitations of the human mind, humans must use
approximate methods to handle many tasks rather than an optimal solution.

In 1957, Simon proposed the notion of Bounded Rationality: that property

of an agent that behaves in a manner that is nearly optimal with respect
to its goals as its resources will allow.

This restriction causes us to make choices that are merely satisfactory

rather than continuing to search for the best option.
DESIGN PRINCIPLES OF RATIONAL AGENTS

1.Performance Measure
The performance measure evaluates how well the agent is achieving its
goals. It can be defined in various ways depending on the application.
For instance, in a self-driving car, the performance measure could
include safety, speed, and passenger comfort.

2. Rationality
Rationality is not just about achieving the best possible outcome but
also about acting optimally given the information and computational
resources available. An agent is considered rational if it does the
“right thing,” given what it knows.

3. Autonomy
A rational agent should operate autonomously to a certain extent, making
decisions and taking actions without human intervention. This involves
learning from the environment and updating its knowledge base.
AGENT ENVIRONMENT

Environments in which agents operate can be defined in different ways. It

is helpful to view the following definitions as referring to the way the
environment appears from the point of view of the agent itself.
AGENT ENVIRONMENT

Observable vs Partially Observable: A fully observable AI environment

(sensors in all environments) has access to all required information to
complete target task. Partially observable environments deal with partial
information in order to solve AI problems. The relevant features of the
environment are only partially observable.

Chess – the board is fully observable, and so are the opponent’s moves.
Autonomous vehicle – the environment is partially observable because
what’s around the corner is not known.

Static vs Dynamic: A static environment is one that does not change as an

agent makes a decision. For example, in chess, your opponent cannot make
a move while it is still your turn.

A dynamic environment, on the other hand, changes while the agent

deliberates.
For example, Traffic conditions on a city road network, where congestion
levels change dynamically due to factors like accidents, construction, or
rush hour.
AGENT ENVIRONMENT
Deterministic vs Stochastic: If the next state of the environment is
completely determined by the current state & the action executed by the
agent, then we say the environment is deterministic while in Stochastic –
For a given current state and action executed by agent, the next state or
outcome cannot be exactly determined e.g. Ludo.
Real-life Example: The traffic signal
is a deterministic environment where
current traffic signal (red) and the
action (stop) determine the next
traffic signal for a pedestrian
(Agent)

Strategic environment: The next state of the environment is completely

determined by the current state and the action executed by multiple
agent. This environment is strategic in Artificial Intelligence because
there will be a certain outcome, without any uncertainty. E.g. chess.
AGENT ENVIRONMENT
Episodic vs. sequential: The agent's experience is divided into atomic
"episodes" (each episode consists of the agent perceiving and then
performing a single action) and each state is independent of each other.

The action on a state has nothing to do with the next state. Real-life
Example: A support bot (agent) answer to a question and then answer to
another question and so on. An episode is a single question and answer.

However, in a Sequential environment, the agent engages in a series of

connected
episodes. An agent requires memory of past actions to determine the next
best actions. For example, chess is a complex game where the outcome of
each move depends on a sequence of previous moves.

Single-agent vs Multi-agent: Only one agent participates in the

environment is Single Agent. It’s a single AI agent handling tasks,
making decisions, and trying to be the jack-of-all-trades. E.g. medical
expert system. More than one agent interact the with the environment is
Multi Agent. Example: The game of football is multi-agent as it involves
AGENT ENVIRONMENT
Discrete vs Continuous: If the number of distinct percepts and actions
that can be performed within it, is limited, the environment is discrete,
otherwise it is continuous.

Chess is an example of a discrete environment. In chess, there are a

finite number of distinct chess pieces (e.g., pawns, rooks, knights) and
a finite number of squares on the chessboard.

The environment in which the actions are performed cannot be numbered

i.e. is not discrete, is said to be continuous.

Self-driving cars are an example of continuous environments as their

actions are driving, parking, etc. which cannot be numbered.
CLASSES OF INTELLIGENT AGENT
Table Based Agent: use a percept sequence/action table in memory to find
the next action. They are implemented by a (large) lookup table.

For example, customers can use a utility-based agent to search for flight
tickets with minimum traveling time, irrespective of the price.
CLASSES OF INTELLIGENT AGENT
Table-Driven Agent : The table driven agent program is invoked for each
new percept and returns an action each time using a table that contains
the appropriate actions for every possible percept sequence.

Real life example;

1. An assembler is a computer program that translates assembly language

code into machine code by looking up in a symbol table.

2. The vacuum cleaner agent is a Table-Driven Agent because the decision

is based only on the current location, and whether the place contains
dirt, and returns an action based on the dirt mapping in the table.
CLASSES OF INTELLIGENT AGENT
Percept based agent or simple reflex agent: A simple reflex agent is the
simplest of all the agent programs. Decisions are taken based on the
current percept, which does not depend on the rest of the percept
history.

The agent function is based on the condition-action rule. If the

condition is true, then the action is taken, else not. This agent
function only succeeds when the environment is fully observable.

Such agents are called reactive agents or stimulus-response agents

because of no notion of history. The rule-based agent aren’t suited for
complex tasks. However, they’re perfect straightforward tasks in a
predictable environment.

Real life example;

1. For example, a room thermostat is a Simple Reflex Agent. It turns the

air conditioner on or off based only on the current room temperature.
SIMPLE REFLEX AGENT ARCHITECTURES
In percept based agents,

1. Information comes from sensors - percepts

2. Changes the agents current state of the world
3. Triggers actions through the effectors

The actions are taken depending

upon the condition. If the
condition is true, the relevant
action is taken. If it is false,
the other action is taken.

The agent takes input from the

environment through sensors, and
delivers the output to the
environment through actuators.
CLASSES OF INTELLIGENT AGENT
The Subsumption Architecture: Subsumption architecture is a reactive
robotic architecture that is heavily associated with behavior-based
robotics which was very popular in the 1980s and 90s.

Brooks (mid 80s and 90s) view:

The philosophy behind Subsumption Architecture is that the world should
be used as its own model. According to Brooks, storing models of the
world is dangerous in dynamic, unpredictable environments because
representations might be incorrect or outdated.

What is needed is the ability to react quickly to the present. So, use
minimal internal state representation, complement at each time step with
sensor input.

Main idea is to build complex, intelligent robots by decomposing

behaviors
into a hierarchy of skills, each defining a percept-action cycle for one
very specific task.
CLASSES OF INTELLIGENT AGENT
The main features of Brooks’ architecture are;

 A layering methodology for robot control systems

 A parallel and distributed method for connecting sensor and actuator in
robots
 There is no explicit knowledge representation
 Behavior is distributed, not centralized
 Response to stimuli is reflexive
 The design is bottom up, and complex behaviors are fashioned from the
combination of simpler underlying ones.
At its core, the basic idea is to
have “levels” of behavior where
simpler behaviors get overridden by
more complex ones based on input from
the environment and higher ones
utilize the outputs of the lower
ones, but do not override them.
CLASSES OF INTELLIGENT AGENT
At its most basic level, the robot avoids collisions, even when still. If
someone approaches a stationary robot, it will back away. This lowest
level behavior can operate in isolation from the rest.

When you add a second level of behavior called wander, the robot will
wander about in random directions. This behavior “subsumes” or override
the lower-level “avoid collision” behavior. But critically, if an object
appears in its path, then “avoid collisions” kicks in again and the robot
will back off for a moment.

Adding a third level of behavior where the robot explores by picking a

distant point in the room and heading for it will subsume the two lower
layers, which can still take over when necessary.
The end result of the blending of these
three very simple behavior is a robot that
appears to explore its environment in a
highly intelligent manner. It has no map and
no 3D model, but it’s robust in a chaotic
and unpredictable environment. Its behavior
CLASSES OF INTELLIGENT AGENT
Real life example;

 Genghis, a six legged autonomous walking robot uses Subsumption

Architecture which was designed to navigate difficult terrain with many
obstacles and elevations. This kind of capability was made possible via
a finite state machine with technique called layered processing. In
Genghis the control system was organized into eight incremental layers.

 The robot, named Herbert employs subsumption architecture, its a Coke

Can Collecting Robot, Herbert searches desks and tables for empty soda
cans, which it picks up and carries away. The robot’s seemingly goal-
directed behavior emerges from the interaction of about 15 simple
behaviors.
CLASSES OF INTELLIGENT AGENT
Goal-based Agent: Often, to make the correct decision, the agent needs
some sort of goal information that describes situations that are
desirable. Its takes decisions on the basis of its goal or desirable
situations so that it can choose such an action that can achieve the
goal.

The agents follow predefined rules and often employ search and planning
algorithms to determine the most efficient path towards their goals.

Real life example;

1. GPS navigation system is an example of a goal-based agent. Given a
destination (goal), it evaluates various routes (actions) using its
world model (maps and traffic conditions) to recommend the fastest or
shortest path, adjusting as conditions change.

2. We can consider Google Bard as a goal-based agent. The specific

objective of Google Bard is to deliver top-quality responses to user
queries. It strategically selects its actions, focusing on actions
that are most likely to help users in their quest to discover the
GOAL-BASED AGENT ARCHITECTURES
Such agents work as follows:

 Information comes from sensors - percepts

 Changes the agents current state of the world
 Based on state of the world and knowledge (memory) and
goals/intentions, it chooses actions and does them through the
effectors.
Goal-based agent focuses only on
reaching the goal set and hence
the decision taken
by the agent is based on how far it
is currently
from the goal state.

Before selecting the right action,

the agent explores all possible
actions.
AGENT ARCHITECTURES
Utility-based Agents: These agents are similar to the goal-based agent
but provide an extra component of utility measurement which makes them
different by providing a measure of success at a given state.

Goals alone are not really enough to generate high-quality behavior. For
example, there are many action sequences that will get the taxi to its
destination, thereby achieving the goal, but some are quicker, safer,
more reliable, or cheaper than others.

Utility-based agent act based not only goals but also the best way to
achieve the goal. They choose actions with the highest expected utility,
which measures how good the outcomes are.

The Utility-based agent is useful when there are multiple possible

alternatives, and an agent has to choose in order to perform the best
action.

The utility function maps each state to a real number to check how
efficiently each action achieves the goals.
UTILITY BASED AGENT ARCHITECTURES
 Utility-based agents evaluate different courses of action based on a
utility function. This function assigns a specific numerical value to
each possible outcome, representing how desirable that outcome is for
the agent.

 The agent strives to maximize its overall score by choosing actions

that lead to outcomes with higher utility values
 Next, they consider different
possible actions to take. For
each action, the agent predicts
the potential outcomes that can
occur

 The utility function assigns a

score to each predicted outcome
based on how desirable it is for
the agent. Then, the agent
selects the predicted action to
lead to the outcome with the
CLASSES OF INTELLIGENT AGENT
Real life example of utility base agent;

1. Autonomous Vehicles: The agent balances factors like speed, safety,

passenger comfort and fuel efficiency to determine the optimal route
and driving behavior.

2. Netflix’s recommendation engine is a utility-based agent. The system

considers user preferences, viewing history, and other factors to
generate recommendations that aim to provide the most enjoyment to the
viewer.
CLASSES OF INTELLIGENT AGENT
Learning Agent: A learning agent refers to a system designed to
autonomously interact with its environment, acquire knowledge from these
interactions, and adapt its behavior to improve performance over time. It
starts to act with basic knowledge and then able to act and adapt
automatically through learning.

Unlike traditional AI systems that operate solely on pre-programmed rules

or models, Learning agents can dynamically change their decision-making
processes depending on experience rather than solely abide by previously
determined laws or instructions.

Real life example;

1. Platforms like Netflix and Amazon use a system equipped with a

learning agent to improve their recommendations for movies, shows, and
products.

2. Fraud Detection and Prevention in the finance sector, AI has

significantly improved the capability to detect and prevent fraud with
LEARNING AGENT ARCHITECTURES
A learning agent has mainly four conceptual components, which are:
 Learning element: It is responsible for making improvements by learning
from environment.
 Critic: Learning element takes feedback from critic which describes
that how well the agent is doing with respect to a fixed performance
standard.
 Performance element: It is responsible for selecting external action
 Problem generator: This component is responsible for suggesting actions
that will lead to new and informative experiences.
Hence, learning agents are able
to learn, analyze performance,
and look for new ways to improve
the performance.
CLASSES OF INTELLIGENT AGENT
State-Based Agent or Model-Based Reflex Agent: The agent should maintain
track of the portion of the world it can't currently see because this is
the most efficient method to handle partial observability. In other
words, the agent must maintain a certain internal state.

Based on the agent's model of how the world functions, it demonstrates

how the current percept is coupled with the previous internal state to
produce the updated description of the present state. A model-based agent
employs its internal state to assess condition during decisions and
action process.

Real life example;

1. Robotics: Robots often use model-based reflex agents to navigate

through dynamic environments, avoiding obstacles, and reaching
specific destinations.

2. Self-Driving Cars are excellent examples of model-based reflex agents.

These cars use a variety of sensors to feed percepts into the car's
MODEL-BASED REFLEX AGENT ARCHITECTURES
Thus a state based agent works as follows:

 Information comes from sensors – percepts

 Based on this, the agent changes the current state of the world
 Based on state of the world and knowledge (memory), and the condition-
action rule, it triggers actions through the effectors

The model-based reflex agent

operates in four stages:

Sense: It perceives the current

state of the world with its sensors.
Model: It constructs an internal
model of the world from what it
sees.
Reason: It uses its model of the
world to decide how to act based on
a set of predefined rules or
heuristics.
STATE SPACE SEARCH
State space search is a problem-solving technique used in Artificial
Intelligence (AI) to find the solution path from the initial state to the
goal state by exploring the various states.

The fundamental concept is to visualize the issue as a graph with nodes

standing in for states and edges for transitions. A set of all states
which can be reached, constitute a search space.
KEY CONCEPTS IN STATE SPACE SEARCH

A state is a representation of problem elements at a given moment. For

example, in a pathfinding problem, a state might represent the agent's
current location.

Initial state: It is the description of the starting configuration of the

agent.

Goal State: The ideal configuration denoting where the agent may end the
search. It is the partial description of the solution.

Intermediate states: These are states in between initial states and goal
states.

Actions and Transitions: They cause transitions from one state to another
which is called a successor state. Actions or transitions are also known
as operator. In a pathfinding problem, operators correspond to possible
movements like moving up, down, left, or right.

A plan is a sequence of actions. The cost of a plan is referred to as the

KEY CONCEPTS IN STATE SPACE SEARCH

Search is the process of considering various possible sequences of

operators applied to the initial state, and finding out a sequence which
match in the goal state.

The solution to the given problem is nothing but a sequence of operators

that map an initial state to a goal state. This sequence forms a solution
path.

A problem space is a set of states and a set of operators. The operators

map from one state to another state. There will be one or more states
that can be called initial states, one or more states which we need to
reach what are known as goal states and there will be states in between
initial states and goal states known as intermediate states.

What is the optimal solution? Obviously the shortest path from the
initial state to the goal state is the best one. Shortest path has only a
few operations compared to all other possible solution paths.

Problem formulation means choosing a relevant set of states to consider,

PROBLEM SPACE

What is problem space?

A problem space is a set of states and a set of operators. The operators
map from one state to another state. There will be one or more states
that can be called initial states, one or more states which we need to
reach what are known as goal states and there will be states in between
initial states and goal states known as intermediate states.

So what is the solution?

The solution to the given problem is nothing but a sequence of operators

that map an initial state to a goal state. This sequence forms a solution
path.

What is the best solution?

Obviously the shortest path from the initial state to the goal state is
the best one. Shortest path has only a few operations compared to all
other possible solution paths. Solution path forms a tree structure where
each node is a state. So searching is nothing but exploring the tree from
SEARCH PROBLEM

State the elements of a search problem?

The search problem is to find a sequence of actions which transforms the

agent from the initial state to a goal state.

A search problem is represented by a 4-tuple {S, s0, A, G}.

 S: the full set of states

 s0∈S: initial state
 A:S→S is a set of operators
 G is the set of final states. Note that G ⊆S
SEARCHING PROCESS

The generic searching process can be very simply described in terms of

the following steps:

Do until a solution is found or the state space is exhausted.

 Check the current state

 Execute allowable actions to find the successor states.
 Pick one of the new states.
 Check if the new state is a solution state If it is not, the new state
becomes the current state and the process is repeated.

The above example illustrates how we can

start from a given state and follow the
successors, and be able to find solution
paths that lead to a goal state. The grey
nodes define the search tree.
SEARCH ALGORITHM

Based on the search problems, we can classify the search algorithm as

follows;
UNINFORMED SEARCH OR BRUTE FORCE SEARCH

Uninformed search, also known as blind search, is a search algorithm that

explores a problem space without any domain knowledge such as closeness,
location of the goal state, etc. other than the initial state and the
possible actions to take.

In depth first search, newly explored nodes were added to the beginning
of your Open list. In breadth first search, newly explored nodes are
added to the end of your Open list.
DEPTH FIRST SEARCH
This is an algorithm for searching tree or graph data structures to find
the shortest path from the initial state to the goal state. The algorithm
starts at the root node and explores straight down into the tree as deep
as it can go, before backing up and trying different paths.

Like tree, we begin with the given source (in tree, we begin with root)
and traverse vertices level by level using a queue data structure. The
only catch here is that, unlike trees, graphs may contain cycles, so we
may come to the same node again. To avoid processing a node more than
once, we use a boolean visited array.
ARCHITECTURE OF THE BFS ALGORITHM

 The tree diagram contains three

layers, which are numbered from 0 to
2.

 Then we explore breadthwise and find

the nodes which are adjacently
connected to our source node.

 Then we must come down to layer two

and find the relative nodes that are
adjacent to the layer 1 nodes.
BFS ALGORITHM

BFS is a simple strategy in which the root node is expanded first, then
all the successors of the root node are expanded next, then their
successors, etc.
WHAT IS INFORMED SEARCH?

Informed search is a type of search algorithm in artificial intelligence

that uses additional information or heuristics to make more accurate
decisions about which paths to explore first.

These heuristics provide estimates of how close a given state is to the

goal, guiding the search toward more promising solutions.

Informed search is particularly useful in solving complex problems

efficiently, as it can significantly reduce the search space and improve
the speed of finding solutions.
STATE THE REAL DIFFERENCE BETWEEN ARTIFICIAL INTELLIGENCE (AI)
PROGRAM AND CONVENTIONAL PROGRAMMING?

AI (Artificial Intelligence) and traditional programming are two distinct

methods for solving problems in the field of computer science.

 In traditional computing, operations are bound by fixed instructions,

AI systems are dynamic.

 While conventional software is programmed to perform a task, AI is

programmed to learn to perform the task.

 Traditional programming is used for tasks with clear rules and defined
inputs and outputs. AI is used in situations where it would be too
complex or impractical to write traditional code, like natural language
processing, image recognition, and predictive modeling.

 AI algorithms require large amounts of data to train effectively.

Traditional programming is deterministic and can only solve the
problems it's specifically programmed to handle.
STATE THE REAL DIFFERENCE BETWEEN ARTIFICIAL INTELLIGENCE (AI)
PROGRAM AND CONVENTIONAL PROGRAMMING?

Traditional computing systems lack human-like reasoning, learning, or

comprehension capability. AI spans a spectrum of human-like capabilities.
STATE THE REAL DIFFERENCE BETWEEN ARTIFICIAL INTELLIGENCE (AI)
PROGRAM AND CONVENTIONAL PROGRAMMING?

A key attribute of an artificial intelligence system is typically one

that involves the ability to learn, reason, or solve problems
autonomously.
EXAMPLES OF HEURISTIC FUNCTION

The generic searching process can be very simply described in terms of

the following steps:

Do until a solution is found or the state space is exhausted.

1. Check the current state
2. Execute allowable actions to find the successor states.
3. Pick one of the new states.
4. Check if the new state is a solution state
If it is not, the new state becomes the current state and the process is
repeated.
EXAMPLES OF HEURISTIC FUNCTION

IT IS A SEARCH ALGORITHM USED TO FIND THE SHORTEST PATH

BETWEEN AN INITIAL AND A FINAL POINT .
EXAMPLES OF HEURISTIC FUNCTION
DIFFERENCE BETWEEN INFORMED AND UNINFORMED SEARCH IN AI

Parameters Informed Search Uninformed Search

It is also known as Heuristic It is also known as Blind

Known as
Search. Search.

It uses knowledge for the It doesn’t use knowledge for

Using Knowledge
searching process. the searching process.

Less efficient as it explores the

Generally more efficient due to
Search Efficiency search space systematically
heuristic guidance.
without guidance.
Breadth-First Search (BFS),
A* Search, Greedy Best-First
Examples Depth-First Search (DFS),
Search, IDA*.
Uniform Cost Search.

Often requires more memory to

Typically requires less memory
Memory Requirements store states and their
compared to informed search.
associated heuristics.

Better for simpler

Well-suited for complex
problems or when no
Problem Complexity problems with large state
domain knowledge is
spaces.
available.
WHAT IS INFORMED SEARCH

Heuristic means “rule of thumb”. Heuristics are criteria, methods or

principles for deciding which among several alternative courses of action
promises to be the most effective in order to achieve some goal”. In
heuristic search or informed search, heuristics are used to identify the
most promising search path.

Best-First Search: Best-first search is a search algorithm which explores

a graph by expanding the most promising node chosen according to a
specified rule.
WHAT IS INFORMED SEARCH

Greedy Search: The idea is to expand the node with the smallest estimated
cost to reach the goal. Greedy algorithms often perform very well. They
tend to find good solutions quickly, although not always optimal ones.

The A* search algorithm is a widely used graph traversal and pathfinding

algorithm in computer science. It's an extension of Dijkstra's algorithm
with an added heuristic, which makes it more efficient in finding the
shortest path in weighted graphs, particularly in games, robotics, and
other AI applications.

The previous Greedy Search considered the distance of the nodes from the
goal. A* uses the path of reaching to the current node from the starting
node, and the path of reaching the goal from the current node. So, the
heuristic function becomes:
f(n) = g(n) + h(n)
where:
f(n): cost of the optimal path from start to goal
g(n): shortest path of the current node from the start
HILL CLIMBING

This algorithm is a classic optimization method that mimics the process

of ascending a hill to reach the peak (the optimal solution). It
iteratively improves the current solution by making small changes, and
it's particularly suited for local optimization problems.

Its primary goal is to find the best solution within a given search space
by iteratively improving the current solution.

For example, hill climbing can be applied to the travelling salesman

problem.
HILL CLIMBING

What is Informed Search?

Informed search is a type of search algorithm in artificial intelligence
that uses additional information or heuristics to make more accurate
decisions about which paths to explore first. These heuristics provide
estimates of how close a given state is to the goal, guiding the search
toward more promising solutions.
KNAPSACK PROBLEM The knapsack problem is a classic
optimization problem in algorithms, where
you have to choose a subset of items with
different weights and values to fit in a
limited capacity knapsack, maximizing the
total value.

There are different versions of the

knapsack problem, such as the 0-1 knapsack
problem, where you can only take or leave
an item or the fractional knapsack problem,
where you can just take part of an item if
your bag overflows.

For the fractional knapsack problem, the

greedy principle can guarantee the optimal
solution, unlike the 0-1 knapsack problem.

At each stage of the problem, the greedy

algorithm picks the option that is locally
optimal. It does not revise its previous