MODULE 2

Overview of Artificial Intelligence(continuation of chapter 1)

1.7 BRIEF HISTORY OF AI

Disciplines Contributing to AI are many and can be considered to be the

foundations of AI. AI developed from several fields they are:
Philosophy:​ Introduced idea that mind works like a machine (as early as 400 B.C.)
Mathematics: Developed logic and algorithms for reasoning and uncertainty
handling
Economics:​ Studied decision-making and how to choose best actions
Psychology:​Modeled human and animal information processing
Linguistics:​ Explained how language conveys meaning and logic
Computer Engineering: Provided hardware and software to make AI applications
possible.
Control Theory: Helped machines act optimally based on feedback
Development of AI
AI combines all these disciplines to make machines intelligent — capable of
learning, reasoning, and adapting.
Important AI people associated with AI are,
Alan Turing
Warren McCulloch
Marvin Minsky
Allen Newell
Herbert Simon
John McCarthy
Important Event:
Dartmouth College Summer Workshop, June 1956 –Considered the official birth
of Artificial Intelligence as a field of study. However we ignore many people, ideas
and events that are also important and focus on only three things, the Dartmouth
conference and Chinese room are discussed in brief whereas Turing-test will be
discussed in detail.
The Dartmouth Conference (1956): Held in summer 1956 at Dartmouth College.
Organized as a two-month workshop.
Key attendees:
John McCarthy (creator of LISP)
Marvin Minsky (AI researcher)
Claude Shannon (Nobel Prize winner)
Plus seven others.
Outcome:
No major breakthrough but established communication between leading
researchers. The term “Artificial Intelligence” was officially coined.

1.7.1 Turing Test

Origin
Proposed by Alan Turing, a British mathematician and one of the founders of
computer science.
Introduced in his 1950 paper “Computing Machinery and Intelligence.”
This test provides answer to the Central question: “Can machines think?”
Concept:
The Turing Test is designed to determine if a machine can exhibit human-like
intelligence.
If a computer’s responses are indistinguishable from those of a human, it is said to
display intelligence.
The test he performed, now known as TURING TEST, was performed in two
phases.
Phase 1: (Fig. 3a):

●​ An interrogator communicates separately with a man and a woman through a

text-based medium (e.g., teletype).
●​ No visual or audio contact is allowed.
●​ The interrogator must determine who is the man and who is the woman.

Phase 2 (Fig. 3b):

●​ One human (e.g., the woman) is replaced by a computer.
●​ The interrogator interacts with both, unaware which is the machine.
●​ If the interrogator cannot distinguish between the computer and the human,
the machine passes the test.

Interpretation
 ●​ If a machine’s conversation is indistinguishable from a human’s, it is said to
think.
●​ The test evaluates intelligent behavior, not internal processes.

In “Computing Machinery and Intelligence”, Turing predicted that by the year

2000, computers would be able to pass the Turing Test at a reasonably
sophisticated level.

How It Works:
●​ A human interrogator communicates with both a human and a machine
through text (keyboard).
●​ If the interrogator cannot tell which one is the computer after several
questions, → then the machine is said to be intelligent.

Important Features of the Turing Test

1. Provides an objective standard to determine intelligence.
2. Focuses on results (answers), not on how the machine works internally.
3. Avoids bias — doesn’t require the machine to look or sound human.
4. Measures communication and reasoning ability.

Capabilities Needed to Pass the Turing Test, A machine must perform these tasks:
1. Natural Language Processing – Understand and communicate in human
language.
2. Knowledge Representation – Store what it knows or hears.
3. Automated Reasoning – Use stored information to answer questions.
4. Machine Learning – Adapt and learn from new situations.
5. Computer Vision – Detect and perceive objects.
6. Robotics – Move and manipulate objects physically.

Criticisms of the Turing Test have come under severe criticism for the following
reasons:
1. It tries to fit machine intelligence into a human framework.
→ Machine intelligence might be different from human intelligence.
2. Passing the test doesn’t prove true intelligence, only that the programmer made
it mimic human-like responses.

CHINESE ROOM EXPERIMENT

Proposed by: John Searle (1980)
Purpose: To challenge the idea that passing the Turing Test means true
understanding.
Thought Experiment:
Imagine a person (who doesn’t know Chinese) inside a room.
The person follows English instructions to match and respond to Chinese symbols.
To outsiders, it seems like the person understands Chinese — but actually, they
don’t.
Conclusion:
Machines can process symbols and give correct outputs without understanding
their meaning.
Hence, passing the Turing Test doesn’t prove real intelligence or understanding.

CURRENT STATUS OF AI
●​ Still, no computer has fully passed the Turing Test.
●​ AI systems can perform limited intelligent tasks (e.g., speech, vision, pattern
recognition).
●​ Every year, the Loebner Prize is awarded to the program closest to passing
the Turing Test.

●​ A chain of algorithms named HAL is being developed in Israel by Dr.

Anant Treister-Goren, a neurolinguist.

●​ HAL is being trained like a child, through experiential learning, similar to

how human children learn language.

●​ HAL can currently understand about 500 words and use 50 words to form
simple, infant-like sentences.

●​ Dr. Goren interacts with HAL in the same way a mother interacts with her
child — teaching it about colors, food, and animals.

●​ The Israeli hi-tech company developing HAL aims to make it an “adult”

computer within 10 years, capable of passing the Turing Test.

●​ If successful, this will blur the distinction between human intelligence and
machine intelligence.

●​ However, current AI computers are still far less intelligent than human
beings.
Intelligence Level​Entity

Low​ Conventional Computers

​ AI Computers
​ Some Animals
High​ Human Beings
​ Superior Aliens

1.8 BRANCHES OF AI
AI is a vast field with many interconnected areas.
Below are the major branches / subfields of AI:
1.​ Natural Language Processing (NLP):​ Understanding and generating human
language (speech/text).
2.​ Knowledge Representation: Storing and organizing facts so that machines
can use them.
3.​ Planning: Deciding steps to reach a goal from a current situation.
4.​ Machine Learning: ​ Enabling computers to learn from experience or
data.
 5.​ Pattern Recognition: Identifying patterns (e.g., recognizing faces,
handwriting).
6.​ Speech Recognition: Converting spoken words into text.
7.​ Search: Finding the best solution among many possibilities.
8.​ Representation: Representing real-world facts using symbols and logic.
9.​ Inference: Drawing new conclusions from known facts.
10.​Common Sense Reasoning: Using everyday logic that humans apply
automatically.
11.​Learning from Experience: Improving behavior based on previous results.
12.​Epistemology: Study of knowledge and how machines can use it.
13.​Ontology: Study of things that exist and their relationships.
14.​Heuristics: Rules of thumb to find quick, good-enough solutions.
15.​Genetic Programming: Using evolution-inspired methods to improve
programs.
16.​Bayesian Networks: Probabilistic reasoning using statistics.
17.​Neural Networks: Simulating the way the human brain processes
information.

1.8 NATURAL LANGUAGE

Definition:
●​ Natural Language refers to the languages spoken, written, or signed by
humans for general-purpose communication (e.g., English, Hindi, Tamil).
●​ It is different from programming languages or formal languages used in
mathematics and logic.
●​ The goal of AI in this domain is to create computer programs capable of
understanding and generating human language, imitating one of the most
fundamental human intelligence abilities.
Importance in AI
●​ Understanding and using natural language is a core aspect of human
intelligence.
●​ The ability of a computer to process language effectively is a major step
toward human-like AI.
●​ The field responsible for this is called Natural Language Processing (NLP)
— a subfield of AI and linguistics.
Applications of Natural Language in AI
●​ Text mining
●​ Search engines
●​ Automatic summarization
●​ Information retrieval
●​ Speech synthesis and recognition
●​ Machine translation
●​ Chatbots and conversational systems
Characteristics of Natural Languages
1. Evolved naturally – Not constructed deliberately, but developed through social
interaction over time.
2. Used by native speakers – Primarily learned naturally during childhood.
3. Complex and ambiguous – Contain idioms, metaphors, and multiple meanings.
4. Infinite vocabulary – Can form countless new sentences from limited words and
rules.
5. Context-dependent – Meaning often depends on situation, tone, and shared
understanding.
Linguistic Aspects of Natural Language
Natural languages are studied in terms of:
Aspect​ Description

Syntax​ Structure or grammar of sentences.

Semantics​ Meaning of words and phrases.
Phonetics/Phonology​ Sound patterns of language.
Pragmatics​Meaning in context (use of language in real situations).

Universal Grammar Theory (Noam Chomsky

Suggests that all natural languages share underlying grammatical rules.
These rules constrain how words and sentences can be structured.
Constructed or artificial languages (like programming languages) do not follow
these natural constraints.

Difference Between Natural & Constructed Languages

Aspect​ Natural Language​ Constructed Language

Origin​ Evolved naturally over time​ Man-made (e.g., programming or

formal languages)
Learning​ Learned spontaneously (e.g., during childhood)​ Intentionally
learned or designed
Complexity​Ambiguous, context-sensitive​ Precise and rule-based
Examples​ English, Hindi, Tamil​ Python, C++, Predicate Logic

Written Language vs. Spoken Language

 ●​ Written languages are not the same as natural languages.

●​ For centuries, many people were fluent speakers but illiterate (unable to
read/write).

●​ Writing is an intentional skill, while speech is spontaneous and innate.

Natural Language vs Programming Language

Aspect​ Natural Language​ Programming Language
Ambiguity​ High​ None
Vocabulary​Infinite and evolving​ Fixed and limited
Grammar​ Complex, irregular​ Strict and well-defined
Learning curve​ Learned from birth​ Requires explicit teaching
Purpose​ Human communication​ Machine instruction

Natural Language Processing (NLP)

●​ NLP enables computers to understand, interpret, and generate human

language.

●​ It involves both syntax (structure) and semantics (meaning).

●​ Researchers have struggled with both aspects:

●​ Early focus: Syntax (grammar).

●​ Later realization: Semantics (meaning) is equally critical.

Challenges in NLP
1. Ambiguity – Words and sentences can have multiple meanings.
2. Context dependence – Meaning depends on surrounding text or situation.
3. Cultural variations – Same word may differ in meaning across cultures.
4. Incomplete understanding of human cognition – Computers lack true
comprehension.

1.10 AUTOMATED REASONING

Definition:
●​ Automated Reasoning (AR) is a subfield of AI focused on enabling
computers to reason automatically.
●​ It involves building systems that can prove theorems, derive conclusions,
and solve logic-based problems without human intervention.

Objective:To mechanize reasoning, just as early computers mechanized arithmetic.

To formalize and automate deduction processes in mathematics, logic, and AI.

Key Components of Automated Reasoning

1. Problem Definition

Identify the class of problems the program will solve.

Define the problem domain (set of known facts or axioms).

2. Knowledge Representation

Choose a language or formalism for representing problems and facts (e.g.,

First-order Logic).

3. Inference Mechanism

Specify how the system will deduce new information or prove theorems.
Example methods: resolution, deduction, logical inference.

4. Computation Efficiency

Optimize reasoning so that the system can perform in a feasible time.

Common Logic Formalisms Used

●​ Propositional Logic
●​ First-order Logic
●​ Higher-order Logic
●​ Type Theory
●​ Lambda Calculus
●​ Horn Logic (used in Prolog)
●​ Non-classical Logics (intuitionistic, modal, etc.)
Important Considerations in AR System Design

1. Problem Domain Definition:

Define which problems belong to the domain and what axioms define it.

2. Internal Representation:

How problems and proofs are stored and processed.

3. Solution Presentation:

How completed proofs or solutions are displayed to the user.

4. Deductive Efficiency:

Ensuring fast and accurate inference mechanisms.

Applications of Automated Reasoning

●​ Mathematical theorem proving
 ●​ Formal verification of software and hardware
●​ Logic programming (e.g., Prolog)
●​ AI planning and scheduling
●​ Circuit design verification
●​ Knowledge-based expert systems

1.11 VISUAL PERCEPTION

Definition:
●​ Visual perception refers to the ability of an organism (or a machine) to
interpret and understand visual information from the environment.
●​ It’s the process of detecting, interpreting, and organizing light stimuli to
form a mental image or understanding of the surrounding world.

Psychological & Biological Perspective

●​ In psychology: visual perception is the end process of vision.
●​ In biology: it involves the visual system (part of the nervous system) that
interprets light data to build a representation of the external world.

Key Functions
●​ Detect and interpret visual stimuli.
●​ Reconstruct a 3D world from 2D retinal images.
●​ Handle binocular vision and correspondence problems (matching images
from both eyes).
Unconscious Inference (Helmholtz’s Theory)
●​ Proposed by Hermann von Helmholtz, who viewed vision as a form of
unconscious inference.
●​ The brain infers probable states of the world based on sensory input.
 ●​ Humans unconsciously apply prior assumptions (e.g., light comes from
above) to interpret scenes.
●​ Recently connected to Bayesian inference models (probabilistic reasoning).

Gestalt Theory
●​ Developed by Gestalt psychologists (1930s–1940s).
●​ Focuses on how people perceive patterns and wholes, not just parts.
●​ Describes laws of visual organization — how we naturally group visual
elements.
Gestalt Laws of Organization
1. Proximity: Objects close together are perceived as a group.
2. Similarity: Similar-looking objects form a group.
3. Closure: The mind fills in missing parts to complete shapes.
4. Symmetry: Symmetrical patterns are seen as belonging together.
5. Common Fate: Items moving in the same direction are grouped together.
6. Continuity: Once a pattern is seen, perception continues it smoothly.

Ecological Psychology (James J. Gibson)

●​ Emphasizes direct perception of the environment.
●​ Believes enough visual information is available in natural settings for
accurate perception.
●​ Focuses on interaction between perception and environment (not internal
models).

Factors Affecting Visual Perception

●​ Personality and cognitive style
●​ Occupation and education
 ●​ Age, culture, motivation, and beliefs
●​ Impairment in vision or spatial skills

Types of Visual Perception

1. Black & White Vision
2. Colour Vision
3. Object Perception
4. Motion Perception
5. Depth Perception
6. Multimodal Perception (integrating multiple sensory inputs)

1.11.2 AUTOMATING VISUAL PERCEPTION

Challenges:Much harder than natural language understanding.
It Requires:
●​ Knowledge of objects and physical properties.
●​ Understanding of basic optics and lighting.
●​ Ability to handle variations in viewpoint, lighting, and shadow.
Approaches
(1) Practical Approach: Focuses on industrial applications.
●​ Aim: recognize objects under simple and controlled conditions.
●​ Example: identifying simple planar surfaces under fixed lighting.

(2) Modeling Human Vision: Developed by David Marr.

●​ Focuses on how humans perceive and represent visual information.
 ●​ Marr separated theory (how vision should work) from implementation (how
to build it).
●​ His computational theory of vision remains influential in computer vision
and AI.

1.12 INTELLIGENT AGENTS

Definition of an Intelligent Agent

An agent is an entity that perceives its environment through sensors and acts upon

that environment through actuators.

●​ Human Agent: Eyes, ears (sensors) and hands, mouth (actuators).

●​ Robotic Agent: Cameras, infrared sensors (sensors) and motors, effectors
(actuators).
●​ Software Agent: Receives keystrokes, files, network packets (inputs) and
acts by displaying, sending data, or modifying information.
Characteristics of Intelligent Agents
An Intelligent Agent (IA) is a software entity that:

Senses its environment and performs actions on behalf of a user or program with
some degree of autonomy.
Key characteristics:

1.​ Autonomy: Works on its own without human intervention.

2.​ Reactivity: Perceives its environment and responds to changes.
3.​ Proactiveness: Takes initiative, not just reacts.
4.​ Persistence: Operates over long time periods.
5.​ Adaptability: Learns and adapts to changes in the environment.
6.​ Goal-oriented: Takes on user’s goals and works toward achieving them.
7.​ Mobility: Can move from one system/network to another.
8.​ Communicative: Interacts with humans and other agents.
9.​ Learning Ability: Improves performance through experience.

1.12.1 Agents and Environments

Definition:
An agent acts within an environment.​

Each agent perceives its environment and performs actions that affect it.

●​ An environment is everything that the agent interacts with.

●​ A software agent “lives” in an operating system, network, or database.
●​ A robotic agent “lives” in the physical world.

Key Points:

●​ Every agent is situated in some environment.

 ●​ Agents operate continuously and autonomously.
●​ Each agent acts in pursuit of goals or objectives.
●​ The agent’s actions affect its environment (and vice versa).
Example:​

A chatbot interacts with users (environment), receives input text (percept), and

produces replies (actions).

1.12.2 Concept of Rationality

Rationality refers to doing the right thing — choosing actions that maximize

performance or goal achievement.

Advantages of Rational-Agent Design

1.​ Rational design enables correct inference and decision-making.

2.​ Provides a clear, general standard for evaluating agent behavior.

Factors Affecting Rationality:

1.​ Performance Measure: Defines success (e.g., accuracy, efficiency).

2.​ Percept Sequence: All percepts received so far.
3.​ Agent’s Knowledge: What the agent knows about the world.
4.​ Actions: Possible operations the agent can perform.

Rational Agent Definition:

A rational agent acts to achieve the best outcome, or when there is uncertainty, the
best expected outcome.

1.12.3 Classification of Agents

Agents can be classified by properties or by complexity of intelligence.
A. Property-Based Classification
Property Other Names Meaning

Responds quickly to
Reactive Sensing and Acting
environment changes.

Exercises control over its

Autonomous —
own actions.

Acts to achieve specific

Goal-Oriented Proactive, Purposeful
goals, not just react.

Continuously operates over

Temporally Continuous —
time.

Interacts with other agents

Communicative Socially able
or humans.

Improves behavior from

Learning Adaptive
experience.

Moves between machines

Mobile —
or environments.

Displays believable
Flexible —
personality and emotion.
B. Complexity-Based Classification

1.​ Reactive Agents (Stimulus-Response):

○​ Stateless; have no memory.
○​ Respond directly to environmental stimuli.
○​ Example: A thermostat that reacts to temperature.
2.​ Goal-Based Agents:
○​ Plan actions by predicting outcomes.
○​ More flexible than reactive agents.
○​ Example: A chess program predicting opponent’s moves.
3.​ Utility-Based Agents:
○​ Use a utility function to measure the “goodness” of states.
○​ Choose actions that maximize expected utility.
○​ Example: Self-driving car optimizing comfort and safety.

1.12.4 Application Areas of Intelligent Agents

Intelligent agents are widely used across domains:

1.​ Information Access and Navigation:

○​ Used in search engines for information retrieval.
○​ Example: Google Search Agent, web crawlers.
2.​ Decision Support and Empowerment:
○​ Assist knowledge workers in making data-driven decisions.
○​ Example: Business analytics systems.
3.​ Repetitive Office Activities:
○​ Automate routine clerical or administrative work.
○​ Example: Email sorting, scheduling, data entry bots.
 4.​ Mundane Personal Activities:
○​ Simplify daily tasks through voice or personal assistants.
○​ Example: Siri, Alexa, Google Assistant.
5.​ Search and Retrieval:
○​ Handle large databases and automate searching/filtering.
○​ Example: E-commerce recommendation agents.
6.​ Domain Experts:
○​ Simulate human expertise in specific areas.
○​ Example: Medical diagnosis systems, legal expert systems.
7.​ Management Activities:
○​ Support organizational management tasks.
○​ Example: Agents that advise, organize, negotiate, and monitor.

1.13 MAJOR COMPONENTS OF INTELLIGENT SYSTEM

An Intelligent System is a system capable of perceiving its environment, reasoning

about it, and taking appropriate actions to achieve specific goals.​

Such systems can learn from experience, adapt to new situations, and improve their

performance over time.

1.13.1 Major Components

Any intelligent system has four major components, namely:

1.​ Knowledge Representation

2.​ Inference Process or Control Strategy
3.​ Programming Languages
4.​ AI Computers (Hardware and Software)
(i) Knowledge Representation

Definition:
Knowledge representation is the method used to store and organize knowledge so

that an AI system can use it effectively to solve problems.

Functions:

●​ Represents facts and rules about the world.

●​ Defines what can be done to solve a problem.
●​ Specifies how to solve it efficiently.

Types of Knowledge:

1.​ Essential Knowledge: Defines what must be done to solve a problem.

2.​ Heuristic Knowledge: Provides strategies or guidelines for solving problems
efficiently (based on experience or rules of thumb).

Examples of Knowledge Representation Schemes:

●​ Semantic networks
●​ Frames
●​ Production rules (If–Then rules)
●​ Predicate logic
●​ Ontologies

Importance:The performance of an AI system depends heavily on how well

knowledge is represented — it must be:

●​ Comprehensive (cover all necessary information)

●​ Efficient (quickly retrievable)
●​ Structured (easy to modify and interpret)
(ii) Inference Process or Control Strategy

Definition:
The inference process is the reasoning mechanism that uses stored knowledge to

derive new conclusions or solve problems.

It determines what to do next by applying logical rules to the available knowledge

base.

Functions:

●​ Guides the AI system toward solutions.

●​ Evaluates multiple possibilities to find the best path.
●​ Applies logical or heuristic reasoning techniques.

Types of Inference/Control Strategies:

1.​ Brute-force Search:

○​ Explores all possible solutions until one is found.
○​ Used in games and problem-solving but may be inefficient.
2.​ Heuristic Search:
○​ Uses rules or experience-based shortcuts to find solutions faster.
○​ Example: A* algorithm in pathfinding.

Example Applications:

●​ Expert systems use inference engines to reason about facts.

●​ AI planning systems use control strategies to generate sequences of actions.
(iii) Programming Languages

Definition:
Programming languages are tools for expressing knowledge and reasoning in a

form that computers can understand and execute.

AI programming languages provide:

●​ Syntax and structures for symbolic reasoning.

●​ Facilities for handling data structures like trees and lists.
●​ Built-in functions for pattern matching and knowledge manipulation.

Common AI Languages:
Language Used For

Symbolic processing, logic

LISP
representation

PROLOG Predicate logic, rule-based reasoning

Python Machine learning, data science

Java / C++ AI tools, neural networks, robotics

Functions of AI Programming Languages:

●​ Represent and manipulate knowledge.

●​ Implement inference and learning algorithms.
●​ Integrate software and hardware components for AI applications.

(iv) AI Computers (Hardware and Software)

Definition:
AI computers are specialized hardware and software systems designed to execute

AI algorithms efficiently.

They include both:

●​ Hardware architecture: processors, memory systems, and specialized chips.

●​ Software systems: operating systems, AI toolkits, and libraries.

Hardware Components:
Architecture Description

Uniprocessor Single CPU for simple computation.

Multiple processors handling array data

Array Processor
simultaneously.

Special Purpose Chips Custom-built for specific AI tasks.

Super Computers Handle large-scale AI computations.

Execute multiple processes

Parallel Computers
simultaneously.

Uses integrated circuits for AI data

VLSI Array Processor
handling.

Used in neural networks and deep

Parallel Distributed Processing Systems
learning.

Software Components:
 Software Type AI Concept Implemented

Machine Language Numerical Processing

Assembly Language Symbolic Processing

High-Level Languages General Problem Solving

LISP / PROLOG Logic Programming

4th Generation Languages Heuristic Search

Object-Oriented Languages Computational Linguistics

Distributed Languages Natural Language Processing

Natural Language Interfaces Expert Systems, Neural Networks

1.13.2 Significance of Major Components

The integration of these four components forms the backbone of any AI system:

Component Role

Knowledge Representation Stores world knowledge.

Inference Process Applies logic to reach conclusions.

Programming Languages Implements algorithms and reasoning.

AI Computers Provides processing power and tools.

Modern Trends:
With advancements in AI:

●​ Systems are now modular and distributed.

●​ Parallel computation enhances speed and efficiency.
●​ AI is increasingly used in psychological systems, data analysis, and
automation.

1.14 IMPORTANT DEFINITIONS AND CONCEPTS

This section provides the fundamental AI terms and key concepts that form the

foundation for understanding topics such as search, logic, reasoning, learning, and

problem-solving.

(1) A* Algorithm

Definition:
A* (A-star) is a search algorithm used to find the shortest path from a start node to

a goal node through a search space, using a heuristic function.

Key Features:

●​ Combines the advantages of Uniform-Cost Search (g) and Greedy Best-First

Search (h).
●​ Guarantees the optimal path if the heuristic is admissible.

(2) Admissibility

Definition:
An admissible search algorithm is one that always finds the optimal solution

(shortest or least-cost path) if one exists.

In A* search:

●​ A heuristic is admissible if it never overestimates the true distance to the

goal.

Example:
For a pathfinding problem, the straight-line distance (Euclidean distance) is an

admissible heuristic because it never exceeds the real travel distance.

(3) Backward Chaining

Definition:
In a logic system, backward chaining means reasoning from a goal (query) back to

the data or known facts.

Process:

1.​ Start from the goal or hypothesis.

2.​ Identify rules that can lead to the goal.
3.​ Check if premises of those rules are true using known facts.

Example:
To prove “It is raining,” check rules like “If clouds are dark, then it is raining.” If

“clouds are dark” is true, the conclusion is confirmed.

Used In:

●​ Expert systems
●​ Logic programming (e.g., PROLOG)
(4) Belief Network (Bayesian Network)

Definition:
A Bayesian Network is a probabilistic graphical model that represents relationships

among variables using nodes (variables) and edges (dependencies).

Purpose:

●​ To represent and reason under uncertainty.

●​ To perform probabilistic inference efficiently.

Applications:

●​ Medical diagnosis
●​ Spam filtering
●​ Fault detection

(5) Breadth-First Search (BFS)

Definition:
An uninformed search algorithm where the shallowest (or lowest-depth) node is

expanded first.

Characteristics:

●​ Explores all nodes level by level.

●​ Guarantees the shortest path (if all step costs are equal).
●​ Uses queue (FIFO) data structure.

Drawbacks:

●​ High memory usage because all nodes at each level are stored.
(6) Data Mining

Definition:
Data Mining (or Knowledge Discovery in Databases – KDD) is the process of

extracting hidden, useful, and previously unknown information from large datasets.

Key Techniques:

●​ Machine Learning
●​ Statistical Analysis
●​ Data Visualization

Applications:

●​ Market basket analysis

●​ Fraud detection
●​ Customer segmentation

(7) Depth-First Search (DFS)

Definition:
An uninformed search algorithm that expands the deepest non-terminal node first.

Characteristics:

●​ Explores one branch fully before backtracking.

●​ Uses stack (LIFO) structure.
●​ Requires less memory but may get stuck in deep or infinite paths.
(8) Forward Chaining

Definition:
A reasoning approach that starts from known facts and applies inference rules to

derive new conclusions.

Process:

1.​ Begin with initial facts.

2.​ Apply rules that match those facts.
3.​ Continue until the goal or conclusion is reached.

Example:
If “It is raining” → “The ground is wet,”​

and we know “It is raining,”​

then we conclude “The ground is wet.”

(9) Fuzzy Logic

Definition:
A type of logic that allows truth values to be any real number between 0 and 1,

rather than just true or false.

Purpose:
To handle uncertainty and vagueness in reasoning (e.g., "very hot", "somewhat

cold").

Applications:
 ●​ Control systems (washing machines, air conditioners)
●​ Robotics
●​ Expert systems

(10) Machine Learning

Definition:
A subfield of AI that focuses on programs that learn from data and improve

automatically through experience.

Types:

●​ Supervised Learning: Learning from labeled data.

●​ Unsupervised Learning: Learning from unlabeled data.
●​ Reinforcement Learning: Learning through reward and punishment.

Applications:

●​ Image and speech recognition

●​ Recommendation systems
●​ Predictive analytics

(11) MiniMax Algorithm

Definition:
A game-playing algorithm used in games with perfect information (e.g., chess,

tic-tac-toe).​

It minimizes the possible loss for a worst-case scenario.

Working Principle:

●​ The maximizing player tries to get the highest score.

●​ The minimizing player tries to reduce it.
●​ The algorithm searches the game tree alternately between both players.

(12) Modus Ponens

Definition:
An inference rule in propositional logic that says:​

“If x → y (if x then y) and x is true, then y must also be true.”

Example:

●​ If it rains, the ground gets wet.

●​ It is raining.
●​ ∴ The ground is wet.

(13) Planning

Definition:
A branch of AI concerned with constructing a sequence of actions to achieve

specific goals in the real world.

Components:

●​ Start state: The initial condition.

●​ Goal state: The desired outcome.
●​ Plan: The ordered sequence of actions to reach the goal.
Applications:

●​ Robotics
●​ Automated scheduling
●​ Logistics and navigation

(14) Problem Space (State Space)

Definition:
The representation of an AI problem in terms of:

●​ States (start, goal, intermediate), and

●​ Operators (actions that move between states).

Purpose:
The problem-solving process involves searching this space to find a path from start

state to goal state.

(15) Search

Definition:
The process of finding a path or sequence of actions from the initial state to the

goal state.

Types:

●​ Uninformed Search: No domain knowledge (e.g., BFS, DFS).

●​ Informed Search: Uses heuristics (e.g., A* search).

(16) Strong AI
Definition:
The claim that computers can be made to actually think and possess human-like

understanding.

Concept:

●​ Machines could have consciousness, emotions, and true intelligence.

●​ A strong AI system would not just simulate thinking but actually understand.

(17) Unification

Definition:
In logic, unification is the process of finding a substitution of variables that makes

two logical statements identical.

Example:
If we have:

●​ P(x, y) and P(a, b)​

Then unification substitutes x=a and y=b.

Used In:

●​ Logic programming (e.g., PROLOG)

●​ Theorem proving
(18) Weak AI

Definition:
The idea that computers are useful tools for simulating human activities but do not

possess real intelligence.

Example:

●​ Chatbots, expert systems, and recommendation engines.

●​ They mimic intelligent behavior but do not truly understand.