Introduction to Artificial Intelligence &

Machine Learning
(AI & ML)
Subject Code: CS301, 5th,CSE, Credit: 5

Priya Rao
Assistant Professor , Department of CSE
CV Raman Global University, Bhubaneswar
Email-priya.rao@cgu-Odisha.ac.in
 Learning Assessments

Bloom’s Level of Cognitive Teacher Assessment / Formative Assessment (40 %) Summative

Task Assessment (60 %)

Quiz Assignment Experiential Attendance Mid Sem End Sem

(10%) (10%) learning * (10%) (20%) (40%)
(10%)
Level-1 Remember 40% 40% 20% 40% 30%
understand
Level-2 Apply 60% 40% 40% 60% 50%
Analyze
Level-3 Evaluate 20% 40% 20%
Create
Total 100% 100% 100% 100% 100%

2
 Vision & Mission

Vision of the C. V. Raman Global University: To emerge as a global leader in the area of higher education through the
pursuit of excellence with future of skills and innovation to match the ever changing global scenario.
Vision of the Department of CSE : To become a leader in providing high quality education and research in the area of
Computer Science, Information Technology, and allied areas.

Mission of C.V. Raman Global University :

 Providing State-of-the art education both at undergraduate, postgraduate and research.
 Working collaboratively with technical Institutes / Universities/ Industries of National and International repute
Keeping abreast with latest technological advancements with a view to enhancing future of skills, R&D and start-up activities at
large;
Realising its’ goals and objectives in a time bound phased manner.
Mission of the Department of CSE:
M1: To develop human resource with sound theoretical and practical knowledge in the discipline of Computer Science & Engineering.
M2: To work in groups for Research, Projects, and Co-Curricular activities involving modern methods, tools and technology.
M3: To collaborate and interact with professionals from industry, academia, professional societies, community groups for
enhancement of quality of education.

Dept. of CSE

3
 Program Outcomes
(POs)
Engineering knowledge: Apply the knowledge of mathematics, science, engineering fundamentals, and an engineering specialization for the solution of complex
engineering problems.

Problem analysis: Identify, formulate, research literature, and analyze complex engineering problems reaching substantiated conclusions using first principles of
mathematics, natural sciences, and engineering sciences.

Design/development of solutions: Design solutions for complex engineering problems and design system components or processes that meet the specified
needs with appropriate consideration for public health and safety, and cultural, societal, and environmental considerations.

Conduct investigations of complex problems: Use research-based knowledge and research methods including design of experiments, analysis and interpretation
of data, and synthesis of the information to provide valid conclusions.

Modern tool usage: Create, select, and apply appropriate techniques, resources, and modern engineering and IT tools, including prediction and modeling
to complex engineering activities, with an understanding of the limitations.

The engineer and society: Apply reasoning informed by the contextual knowledge to assess societal, health, safety, legal, and cultural issues and the consequent
responsibilities relevant to the professional engineering practice.

Environment and sustainability: Understand the impact of the professional engineering solutions in societal and environmental contexts, and demonstrate
the knowledge of, and need for sustainable development.

Ethics: Apply ethical principles and commit to professional ethics and responsibilities and norms of the engineering practice.

Individual and team work: Function effectively as an individual, and as a member or leader in diverse teams, and in multidisciplinary settings.

Communication: Communicate effectively on complex engineering activities with the engineering community and with the society at large, such as, being able
to comprehend and write effective reports and design documentation, make effective presentations, and give and receive clear instructions.

Project management and finance: Demonstrate knowledge and understanding

of the engineering and management principles and apply these to one’s own work, as a member and leader in a team, to manage
projects and in multidisciplinary environments.

Life-long learning: Recognize the need for, and have the preparation and ability to engage in independent and life-long learning in the broadest context of
technological change.

4
 PROGRAM EDUCATIONAL OBJECTIVE (PEO)

PEO1- To provide the fundamental knowledge in mathematics, science and engineering concepts for the
development of engineering system (Fundamental Knowledge).

PEO2- To apply current industry accepted computing practices and emerging technologies to analyze,
design, implement, test and verify high quality computing systems and computer based solutions to real
world problems (Design and development).

PEO3- To enable the use of appropriate skill sets and its applications towards social impacts of computing
technologies in the career related activities (Skill Set) and to produce Efficient team leaders, effective
communicators and capable of working in multi-disciplinary environment following ethical
values(Communication).

PEO4- To practice professionally and ethically in various positions of industry or government and/or
succeed in graduate (Professionalism) with lifelong learning and to make substantial contributions to the
society (Societal Contribution)

5
 PROGRAM SPECIFIC OUTCOMES (PSO)

Graduates have the ability to

PSO1- The ability to understand, analyze and develop computer programs in the
area of computer science and to solve computer software and hardware
related engineering problems
PSO2- The ability to develop software systems to allow convenient use of
computing system and posses professional skills and knowledge of software
design processes.
PSO3-The ability to gain knowledge in diverse area of computer science and apply
for successful career entrepreneurship and higher studies.
PSO4-Ability to use knowledge of ethical and management principles required to
required to work in a team as well as to lead a team.

6
 Detailed Syllabus
Introduction: Objective, scope and outcome of the course. AI techniques, Level of the Model, AI problems, foundation of AI and history of AI
intelligent agents: Agents and Environments, the concept of rationality, the nature of environments, structure of agents, problem solving agents,
Unit – I problem formulation, defining the problem as a State Space Search, problem characteristics, Issues in the designing of Search Programs.

Knowledge representation: Knowledge representation and mapping, Knowledge issues, knowledge representation using propositional and
predicate logic, logical consequences, logic programming, semantic nets- frames and inheritance, constraint propagation, representing knowledge
using rules, rules-based deduction systems, Conceptual Dependency Scripts, CYC.
Unit – II
Predicate Logic: Predicate Representing Instance and ISA Relationships, Computable Functions and Predicates, Resolution, Natural Deduction,
syntax and semantics of an expression, semantic Tableau. Forward and backward reasoning.
Quantifying Uncertainty: Acting under uncertainty, basic probability notations, inference using full joint distributions, independence,
Bayes’ rule and its use.
Probabilistic Reasoning: Representing knowledge in an uncertain domain, the semantics of Bayesian Networks, Efficient representation of
Unit – III conditional distributions, Exact and approximate inference in Bayesian Networks, Relational and First-Order Probability models.
Making Simple Decisions: Basics of utility theory, utility functions, multi-attribute utility functions, decision networks, the value of information,
decision-theoretic expert systems. Making Complex Decisions: Sequential decision problems, value iteration, policy iteration, Partially
Observable MDPs, Decision with multiple Agents: Game theory.
Introduction to Machine Learning: Definition, History, applications and need, Learning: what is learning, Types of learning, Examples,
Unit – IV Data/Features, Hypothesis space and inductive bias, Bias-variance Trade off, Loss function, Evaluation matrix. Basic statistical concepts: Mean
Median, Mode, Variance, Co-variance, Correlation, dispersion matrix.
Supervised Learning: Supervised learning setup (training, testing). Classification and Regression, Regression: Linear regression. Logistic
regression. Classification: Minimum distance classifier, k-nearest neighbour classifier, Perceptron (single layer/multi-layer). Decision trees,
Model selection, Over-fitting, Cross validation.
Unit – V
Unsupervised Learning: Clustering. Similarity measures, Case Study for K-means algorithm, Hierarchical clustering - Divisive algorithm,
Density based clustering (DBSCAN), Inductive and deductive learning.

7
 Course Outcomes
Upon successful completion of this course, students will be able to:

CO-1 Demonstrate fundamental understanding of the history of artificial intelligence (AI)

and its foundations to intelligent agents.
CO-2 Apply selected basic AI techniques; judge applicability of more advanced techniques.
Participate in the design of systems that act intelligently and learn from experience.

CO-3 Apply basic probability notations and inference techniques to reason under
uncertainty. Represent uncertain knowledge using Bayesian networks and perform
exact and approximate inference in these networks. Quantify and optimize utility
functions for decision-making tasks.

CO-4 Learn the basics and use mathematical concepts required for machine learning.

CO-5 Identify and differentiate different types of supervised and unsupervised learning
 Textbooks

• The course lecture are based on the following books

1. Suryani, A. (2025). ARTIFICIAL INTELLIGENCE FOR BEGINNERS: A QUICK GUIDE TO

UNDERSTANDING AI: Your Essential Guide to AI Fundamentals, Real-World Applications, and
Future Skills. GuinEvel Editions.
2. Dhinakaran, M., Deepthi, P., Mary, S. S. C., & Namdev, M. A. (2023). Machine Learning for
Absolute Beginners. Booksclinic Publishing.
3. Mehrotra, D. (2019). Basics of artificial intelligence & machine learning. Notion Press.
 Reference Books

1. S. Russel and P. Norvig, Artificial Intelligence: a Modern Approach, Pearson.

2. D. W. Patterson, Introduction to Artificial Intelligence and Expert Systems, Prentice Hall of India.
3. “Machine Learning”, Tom Mitchell, First Edition, McGraw- Hill, 1997.
4. “Introduction to Machine Learning”, Ethem Alpaydin, Second Edition, 2010.
5. “Machine Learning: An Algorithmic Perspective”, Stephen Marsland, Second Edition, 2014.

Open Sources (Websites References )

1. https://nptel.ac.in/courses/106105077
2. https://nptel.ac.in/courses/106106139/
3. https://onlinecourses.nptel.ac.in/noc23_cs18/preview

10
What is AI?
 What is AI?

AI: Artificial Intelligence

Man-made Thinking Power

• Making machines think & act

like humans.
• Uses logic, data & learning
• Examples: Google Maps, Siri,
Self-driving cars

Definition: Artificial intelligence [1] is a domain of computer

science focused on the examination and development of computer
systems that demonstrate some kind of intelligence.
1. Vacuum Cleaner 2. Phone Keyboard
🧾 Traditional: You push and move it 🧾 Traditional: Just type — no help.
manually. 🧾 AI: Predicts your next word, auto-
🧾 AI: Robot vacuum moves on its corrects based on your habits.
own, avoids furniture, learns the room.

3. Camera 4. Video Games

🧾 Traditional: Manual focus and 🧾 Traditional: Enemies behave the
settings. same every time.
🧾 AI: Auto-focuses, detects smiles, 🧾 AI: Enemies learn your moves and
and improves photo quality. become smarter.

Traditional AI
Fixed rules Learns from data
No understanding Understands input
Same every time Adapts and improves
Needs perfect input Works with messy input
Applications of AI:
• Business (financial
strategies)
• Engineering
• Fraud detection
• Education

• Medicine: monitoring,
diagnosing
• Object identification
• Information retrieval
Building AI Systems:

perception reasoning

learning acting

Example: Building a robot that finds the shortest path in a maze.

 It needs perception: detect walls.
 It needs reasoning: plan a route.
 It needs learning: improve over time.
 It needs acting: move correctly.
Intelligent Systems:
In order to design intelligent systems, it is important to categorize them into four
categories (Luger and Stubberfield 1993), (Russell and Norvig, 2003)
1. Systems that think like humans
2. Systems that think rationally
3. Systems that behave like humans
4. Systems that behave rationally

1. Systems that think like humans

•These systems try to mimic the way humans think.
•Based on cognitive science: how the brain processes information, solves problems,
learns, etc.
🔍 Goal: Understand and simulate human thought processes.
Example:
•A system that uses neural networks to solve a puzzle by learning like a child.
Systems that think rationally
•These systems aim to think logically and correctly, using rules and reasoning.
•They don’t try to copy human thinking exactly, but instead focus on formal logic.
🔍 Goal: Follow the laws of rational thought.
Example:
•An expert system that uses logical rules like "IF symptoms = X, THEN diagnosis = Y".
•A robot that makes decisions using mathematical logic.
3. Systems that behave like humans
•These systems try to act like humans, even if they don’t “think” the same way.
•The focus is on making machines that appear intelligent to people.
🔍 Goal: Imitate human behavior, especially observable actions.
Example:
•A chatbot that holds a conversation like a human (e.g., Siri, Alexa).
•A humanoid robot that can walk, talk, and respond to questions.

4. Systems that behave rationally

These systems focus on doing the right thing to achieve the best result.
They don’t need to think like humans — they just need to make good decisions.
🔍 Goal: Choose actions that lead to the best outcome, based on the current situation.
Example:
A self-driving car that avoids traffic and follows safety rules.
A chess AI (like AlphaZero) that always chooses the best move to win the game.
 Intelligent systems

Think Like Humans Think Rationally

(Human-like Thought) (Logical Reasoning)

Cognitive AI Rule-Based AI

Ex: Neural networks modeling brain Ex: Expert systems, logic solvers

Behave Like Humans Behave Rationally

(Mimic Human Actions) (Best Decisions)

Ex: Chatbots, robots Ex: Self-driving cars, Game-playing AIs

History of Artificial Intelligence
Year/Period Milestone/Event Key Points
Alan Turing asks “Can machines
1950 Turing Test
think?”

Term Artificial Intelligence coined

1956 Dartmouth Conference
— birth of AI as a field

Logic Theorist, General Problem

1950s–1960s Early Programs
Solver — first AI programs

Basic games (chess), natural

1960s–1970s Early Developments
language (ELIZA)
Funding cuts due to limited
1970s–1980s AI Winter
progress

Practical industry applications

1980s–1990s Expert Systems
revived interest

IBM’s Deep Blue defeats chess

1997 Deep Blue
champion Garry Kasparov
Modern Era of AI
Year/Period Milestone/Event Key Points
Rise of large datasets and improved
2000s Big Data & Machine Learning
algorithms
AlexNet wins ImageNet — huge leap
2012 Deep Learning Breakthrough
for neural networks
Google DeepMind’s AlphaGo defeats
2016 AlphaGo
Go champion Lee Sedol
AI models like GPT, DALL·E —
2020s Generative AI
language, images, video
Virtual assistants, self-driving cars,
Today Everyday AI
recommendation systems
 Agents & Environments
An agent [2] is defined as an entity that perceives its surroundings via sensors and interacts
with that environment through actuators.

senses thinks acts

 Intelligent (AI) Agents
Aspect Agent Intelligent Agent
An agent that acts rationally to
A system that perceives its
Definition achieve goals based on percepts and
environment and acts on it.
learning.
Makes decisions based on
Decision May act based on fixed rules or direct
performance measures, goals, and
Making input.
past experience.
Usually does not learn from Can learn and adapt from
Learning Ability
experience. environment and previous actions.
Goal
Not necessarily goal-driven. Operates with specific goals in mind.
Orientation
Operates independently after initial
Autonomy May require continuous human input.
design.
May or may not respond to Reacts to changes in the
Reactivity
environment changes. environment.
A smart thermostat: learns user
A thermostat: senses temperature and
Example preferences and adjusts settings
turns heater on/off.
automatically.
TYPES OF AI AGENTS
Types of AI Agents:
1. Simple Reflex
• Simple reflex agents act only on the basis of the current perception and ignore
the rest of the previous state in which the system was.

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

• No internal state (memory)

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

Example: Thermostat: "If temp < 20°C, then turn on heater"

 • Limitations:
• Very limited intelligence
• No knowledge about the non-perceptual parts of the state.
• Operating in a partially observable environment, infinite loops are
unavoidable.

AGENT
Sensors

E
What is the N
Percepts
situation right now V
I
Condition R
action O
What action
rules N
should I take
M
now?
E
N
T
Actuators Action
2. Model-based reflex agents:

• It works by finding a rule whose condition matches the current situation.

• The Model-based agent can work in a partially observable environment, and
track the situation.
• A model-based agent has two important factors:
• Model: It is knowledge about "how things happen in the world," so it is
called a Model-based agent.
• Internal State: It is a representation of the current state based on percept
history.

•Maintains internal state using a model of the world

•Can handle partial observability

Example: Self-driving car remembering past locations of obstacles

Model-based reflex agents:
3. Goal-based agents:
• A goal-based agent has an agenda.
• It operates based on a goal in front of it and makes decisions based on how
best to reach that goal.
• A goal-based agent operates as a search and planning function, meaning it
targets the goal ahead and finds the right action in order to reach it.
• Expansion of model-based agent.

•Acts to achieve defined goals

•Evaluates future actions and their outcomes
•Example: A Navigation system finding the best route
4. Utility-based agents:
• A utility-based agent is an agent that acts based not only on what the
goal is, but the best way to reach that goal.
• 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 term utility can be used to describe how "happy" the agent is.

•Considers goals and preferences (utility)

•Chooses action that maximizes utility
•Example: Delivery drone selecting fastest and safest route

5. Learning Agents
•Improves performance by learning from experience
•Components:
• Learning Element
• Performance Element
• Problem Generator
•Example: Game AI adapting to player behavior
Real-Life Applications
•Self-driving cars → Utility-Based & Learning Agents
•Chatbots (Siri, Alexa) → Model-Based & Goal-Based
•Game AI → Learning Agents
•Smart Thermostat → Simple Reflex

Type How it works Example

Simple Reflex Agent Acts only on current Room heater turns

perception. ON/OFF.
Model-Based Agent Remembers past states Self-driving car maps
(has memory). road.
Goal-Based Agent Chooses actions to reach Robot arm moves
goals. objects.
Utility-Based Agent Chooses best among Smart assistant
options by comparing scheduling.
utilities.

Learning Agent Learns and improves AI that gets better at

from experience. games.
Components of AI agents
AI agents - perceive their environment
- process information
- decide, collaborate, take meaningful actions
- learn from their experience
Components
1. Perception and input handling
2. Planning and task decomposition
3. Memory
4. Reasoning and decision making
5. Action and tool calling
6. Communication
7. Learning and adaptation
Architecture of AI Agent
• the underlying framework that determines how it processes information, makes
decisions, and interacts with its environment.
• components enable it to function autonomously and adapt to changing inputs.
Step Robot Example Ticket Booking Example

Sensors detect dirty living User says: “Book me a train

Perception room, vase nearby, battery at ticket from Delhi to Mumbai
30%. tomorrow morning.”

Decide to search trains for

Decide to clean living room,
Reasoning tomorrow morning,
avoid vase, charge later.
compare, select best.
User prefers AC Chair Car,
Map of house, vase location,
World Knowledge Memory knows booking process,
cleaning history.
payment details.
Learns living room gets dirty Learns user prefers fast
Learning
in evenings. trains.
Start cleaning, avoid vase, Book ticket, process
Action
send update. payment, confirm.

Sends message: “Your ticket

Sends app notification:
Interaction is confirmed for 7:05 AM
“Living room cleaned.”
Rajdhani Express.”
Step Robot Example Ticket Booking Example
User says: “Book me a train
User says: “Clean the living
User → Agent ticket from Delhi to Mumbai
room, avoid the vase.”
tomorrow morning.”
“You are a travel assistant.
“You are a cleaning robot.
Find and book the best
Navigate to the living room
Prompt Template available train from Delhi to
and clean it, avoiding the
Mumbai tomorrow morning,
vase near the north wall.”
AC Chair Car preferred.”
Plans: search trains → filter
Plans route, avoids vase,
LLM morning → pick fastest →
decides cleaning steps.
book.
Motor control, vacuum, Train booking API, payment
Tools
obstacle sensors. gateway API.
Past preferences, payment
Memory House layout, vase location.
details.
“Booking completed:
“Living room cleaned Rajdhani Express, 7:05 AM,
Response
successfully, vase avoided.” AC Chair Car. PNR:
1234567890.”
Architecture of AI Agent
• 1. Perception Layer (Input)
 Purpose: sense and understand its environment by collecting raw data from various
sources.
 Components: Sensors (physical world), data streams (digital world), cameras,
microphones, biometric sensors, IoT devices, and GPS.
 Function: Data collected through these sources allows the agent to form a
representation of its environment.
• 2. Processing Layer (Data Interpretation and Understanding)
 Purpose: Data must be cleaned, processed, and structured to make sense of it.
 Components: Feature Extraction Modules, Data Preprocessing, Natural Language
Processing (NLP).
 Function: Digital Agents, Physical Agents
Architecture of AI Agent
• 3. Decision-Making Layer
 Purpose: informed decisions based on the processed data.
 Components Rule-Based Systems, Heuristic Algorithms, Machine Learning
Models, Reinforcement Learning
 Function: Based on the data input and the goals set, the AI agent makes decisions
that are contextually appropriate. Rule-Based Systems:, Machine Learning
Systems
4. Action Layer (Output)
 Purpose: This layer allows the AI agent to interact with its environment by
executing decisions that have been made.
 Components: Actuators, Digital Triggers, Control Systems
 Function: The agent takes the decision it has made and turns it into action
Architecture of AI Agent
• 5. Learning and Feedback Layer
 Purpose: AI agents to update their behavior based on past actions and new data,
enabling adaptive, dynamic decision-making.
 Components: Machine Learning Algorithms, Feedback Loops, Memory Systems
 Function: The AI agent updates its models, Reinforcement Learning, Supervised
Learning
6. Communication Layer
 Purpose: interactions between the AI agent and external systems, other agents, or
users.
 Components Networking and API Interfaces, Multi-Agent Systems
 Function: Multiple AI agents might collaborate to achieve a common goal.
Architecture of AI Agent
• 5. Learning and Feedback Layer
 Purpose: AI agents to update their behavior based on past actions and new data,
enabling adaptive, dynamic decision-making.
 Components: Machine Learning Algorithms, Feedback Loops, Memory Systems
 Function: The AI agent updates its models, Reinforcement Learning, Supervised
Learning
• 6. Communication Layer
 Purpose: interactions between the AI agent and external systems, other agents, or
users.
 Components Networking and API Interfaces, Multi-Agent Systems
 Function: Multiple AI agents might collaborate to achieve a common goal.
Working Principle of AI Agents

Sensors → Perception → Knowledge Base → Reasoning → Learning

→ Actuators
1.Perception:
 The agent uses sensors to perceive its environment and collect relevant data (called
percepts).
2. Storage of Percepts:
 The agent may store percept history to help in decision-making (especially in model-based
or learning agents).
3. Decision Making:
 Based on percepts and internal knowledge (e.g., rules, goals, utility functions), the agent
decides on the best action.
4. Action Execution:
 The chosen action is performed through actuators, which interact with the environment.
5. Goal Fulfillment or Utility Maximization:
 The agent acts in a way to achieve its goals or maximize its performance measure.
6. Feedback and Learning (only in learning agents):
 The agent monitors the outcome of its actions and learns from experience to improve future
performance.
Environment:
An environment is everything in the world that surrounds the agent, but it is not a
part of the agent itself.

An environment can be described as a situation in which an agent is present.

An environment is the setting in which an agent operates.

Types:
1. Fully observable vs Partially Observable
2. Deterministic vs. Stochastic
3. Episodic vs. non-episodic (Sequential)
4. Static vs. dynamic
5. Discrete vs. continuous
6. Known vs. Unknown
1. Observable Environments
•Fully Observable: Agent has access to the complete state of the environment
• Example: Chess game
•Partially Observable: Agent has incomplete or noisy perception
• Example: Driving in fog
2. Deterministic vs. Stochastic
•Deterministic: Next state is completely determined by current state and
action
• Example: Puzzle-solving robot
•Stochastic: Involves randomness or uncertainty
• Example: Stock market prediction

3. Episodic vs. Sequential

•Episodic: Agent's actions do not depend on previous ones
• Example: Image classification
•Sequential: Current action affects future actions
• Example: Autonomous driving

4. Static vs. Dynamic

•Static: Environment does not change while the agent is deciding
• Example: Crossword puzzle
•Dynamic: Environment changes over time
• Example: Multiplayer online game
5. Discrete vs. Continuous
•Discrete: Finite number of distinct states and actions
• Example: Board games
•Continuous: Infinite states or actions
• Example: Robot arm movement

6. Known vs. Unknown

•Known: Agent knows the rules and outcomes
• Example: Chess
•Unknown: Agent must learn how the environment works
• Example: New software game

Real-Life Examples of Environments

•Self-driving car → Partially observable, dynamic, stochastic, continuous
•Tic-tac-toe → Fully observable, deterministic, discrete, known
7. Single-agent vs. Multi-agent:
Single-agent: Only one agent is present in the environment.
Example Solving a maze with a single robot is an example.
Multi-agent:
Multiple agents interact within the environment.
A game of chess with two players is a multi-agent
environment
Concept of Rationality
A rational agent
 is an entity that acts in a way that maximizes its chances of
achieving its goals.
 agent’s decisions are logically aimed at achieving its goals, given
the knowledge it has and the constraints it faces.
It perceives its environment,
processes information,
 and then takes actions
 that are likely to lead to the best outcome
 based on its objectives and available knowledge.
Concept of Rationality
Key Characteristics of a Rational Agent:
 Autonomy: It can act independently without direct
human intervention.
 Goal-oriented: It has specific objectives it tries to
achieve.
 Rationality: It makes decisions that are likely to lead to
the best outcome based on its goals and available
information.
 Adaptability: It can learn and adjust its behavior based
on new information or changing circumstances.
Concept of Rationality
Examples
1. Self-Driving Cars:
Perception:
uses sensors like cameras, lidar, and radar to perceive its
surroundings, detecting traffic, pedestrians, and road signs.
Reasoning:
Based on this perception, the car's AI analyzes the data to
make decisions about speed, steering, and braking.
Action:
The car then acts on these decisions, navigating safely and
efficiently to its destination.
Concept of Rationality
Examples
2. AI Assistants (like Siri or Alexa):
Perception:
receive input from users through voice commands or text.
Reasoning:
They then process this input, understanding the user's intent and
accessing relevant information from databases or other sources.
Action:
They respond to the user, providing information, playing music,
setting reminders, or controlling smart home devices.
 What is Agentic AI?

• Agentic AI is a class of artificial intelligence that focuses on autonomous

systems that can make decisions and perform tasks without human
intervention.
• The independent systems automatically respond to conditions, to produce
process results.
• The field is closely linked to agentic automation, also known as agent-
based process management systems.
 Examples of Agentic AI
1. Autonomous Vehicles
 Agentic AI powers self-driving cars, enabling them to:
 Detect obstacles and pedestrians.
 Navigate complex road networks.
 Adapt to unpredictable traffic conditions.

https://so-development.org/how-agentic-ai-works-a-deep-dive-into-autonomous-intelligence/
Examples of Agentic AI
2. Robotics and Automation
 Industries are employing Agentic AI in robotics to automate tasks such as:
 Warehouse and inventory management.
 Precision manufacturing.
 Medical diagnostics and robotic surgery.
Examples of Agentic AI
3. AI powered Personal Assistants
 Advanced digital assistants like ChatGPT, Auto-GPT, and AI-driven
customer service bots leverage Agentic AI
4. Research and Discovery
 Agentic AI assists researchers in fields like biology, physics, and materials
science by:
 Conducting simulations.
 Generating hypotheses.
 Analyzing vast datasets
Examples of Agentic AI
5. Health Care
AI Agents and Agentic AI
The AI Agent acts as a deterministic component with limited scope, while Agentic AI
reflects distributed intelligence, characterized by goal decomposition, inter-agent
communication, and contextual adaptation, hallmarks of modern agentic AI frameworks.
AI Agents and Agentic AI

https://3scsolution.com/insight/agentic-ai-vs-ai-agents
 AI agent vs Agentic AI
Characteristics AI Agents Agentic AI

rule-based or relies on They can evaluate multiple options, make

1. Decision-Making pattern recognition within a complex, goal-driven decisions, and adapt
specific, well-defined task. their strategies based on the situation and
feedback.
operate with a degree of can operate independently, make
2. Autonomy autonomy within a specific decisions, and take actions with minimal
task or domain, but often human intervention
require human input
they are programmed to more proactive in goal setting and pursuit
3. Goal-Orientation achieve specific objectives

have a structured architecture build upon AI agent structure with

4.Structure of Agents with components like additional capabilities for learning and
sensors, a perception adaptation.
system, a decision-making
module, and actuators.
Overview of Large Language Models (LLMs)
LLM-AI models trained on massive datasets of text and code, enabling them to understand,
generate, and manipulate human language.
Key capabilities:
1. Text generation:
LLMs can generate human-quality text, including articles, stories, poems, and even code.
2. Translation:
They can translate text from one language to another with a high degree of accuracy.
3. Summarization:
LLMs can condense large amounts of text into concise summaries.
4. Question answering:
They can understand questions and provide relevant answers based on their training data.
5.Code generation:
Some LLMs can even generate code in various programming languages.
Examples of LLMs

GPT-3, GPT-4 (OpenAI)

LaMDA (Google)

Llama 2 (Meta)

BERT (Google)

BLOOM
Applications of LLMs

LLMs are being used in a wide range of industries, including:

1. Content creation: Generating marketing copy, social media posts, and other forms of
content.
2. Customer service: Powering chatbots and virtual assistants for customer support.
3. Education: Assisting with research, writing, and learning.
4. Software development: Helping developers write and debug code.
5. Healthcare: Assisting with medical research, diagnosis, and patient care.
How of LLM works

An LLM is kind of giant “word prediction machine”

 Input: Break text into tokens
 Embedding: Turn tokens into number vectors
 Transformer Layers: Use self-attention + feed-forward steps to process meaning
 Output: Predict the most likely next token
 Repeat until done
 Retrieval Augmented Generation (RAG)

 is an architecture for optimizing the performance of an AI model by connecting it with

external knowledge bases ( internal organizational data, scholarly journals and
specialized datasets).
 RAG helps lLLMs deliver more relevant responses at a higher quality.
 By integrating relevant information into the generation process, chatbots and other NLP
tools can create more accurate domain-specific content without needing further training.
 Benefit of RAG
“Cost-efficient AI implementation and AI scaling”

https://blogs.nvidia.com/blog/what-is-retrieval-augmented-
generation/
https://www.ibm.com/think/topics/retrieval-augmented-generation
RAG Use Cases

Use case of Data-powered question-answering abilities of RAG systems

1. Specialized chatbots and virtual assistants
2. Research
3. Content generation
4. Market analysis and product development
5. Knowledge engines
6. Recommendation services
How does RAG work?
RAG systems follow a five-stage process:
1. The user submits a prompt.
2. The information retrieval model queries the knowledge base for relevant data.
3. Relevant information is returned from the knowledge base to the integration layer.
4. The RAG system engineers an augmented prompt to the LLM with enhanced context
from the retrieved data.
5. The LLM generates an output and returns an output to the user.
 The RAG Pipeline
1. User Query
You type a question:
"What’s the national flower of India?"
2. Query Embedding
Your question is turned into a vector (a list of numbers that represent its meaning).
This lets the system search for meaning, not just exact words.
3. Retrieval
The vector is sent to a vector database or search index.
The system finds the most relevant documents or text chunks from an external knowledge
source (could be PDFs, websites, company docs).
Example:
It retrieves a text snippet: "Lotus is the national flower of India."
 The RAG Pipeline
Example: Query about Indian History
User Question:
"When did the Quit India Movement start?"
1. User Query → Embed Query
The question is converted into a vector.
2. Retrieval
The system searches an Indian history database or a digital archive like the National
Archives of India, Wikipedia, or NCERT e-books.
It finds a snippet:
"The Quit India Movement was launched by Mahatma Gandhi on 8 August 1942 during
the Bombay session of the All-India Congress Committee."
 The RAG Pipeline
3. Augmentation
The snippet is added to the query:
Context: The Quit India Movement was launched on 8 August 1942 by Mahatma Gandhi
during the Bombay session of the All-India Congress Committee.
Question: When did the Quit India Movement start?
4. Generation
The LLM responds using both the retrieved snippet and its own training:
"The Quit India Movement began on 8 August 1942, when Mahatma Gandhi called for
an end to British rule during the Bombay session of the All-India Congress Committee."
5. Return to User
The final answer is displayed.
The RAG Pipeline
 Business Benefits of RAG

• Access to Up-to-Date Info:

Uses latest data from external sources, avoiding outdated model answers.
• Domain-Specific Accuracy:
Pulls from company databases, research, or legal docs for precise answers.
• Reduced Hallucination:
By grounding responses in real documents, lowers the chance of made-up facts.
• Faster Knowledge Access:
Saves employees time searching manuals, policies, or archives.
• Better Compliance:
Answers can be sourced from approved, compliant documentation.
• Cost-Effective:
No need to retrain the whole LLM every time data changes — just update the database.
 Tools Commonly Used for RAG

A. Vector Databases (for storing and retrieving embeddings):

- Pinecone
- Weaviate
- Milvus
- Vespa
- FAISS (open-source by Facebook AI)
B. LLMs (for generation):
- OpenAI GPT models
- Anthropic Claude
- LLaMA 2 / LLaMA 3
- Mistral
- Cohere Command R
 Tools Commonly Used for RAG

C. Frameworks & Orchestration:

- LangChain
- LlamaIndex
- Haystack
- Semantic Kernel
D. Data Sources:
- Company knowledge bases (Confluence, Notion, SharePoint)
- Legal databases
- Research repositories
- Websites, news feeds, APIs
 Use Cases of RAG

1. Customer Support:

Scenario: An Indian e-commerce site integrates RAG so its chatbot can instantly fetch the
latest return policies, shipping timelines, and discount rules from internal documents.
Benefit: Reduced customer wait time and consistent answers.

2. Legal & Compliance:

Scenario: A law firm in Mumbai uses RAG to answer legal queries based on recent
Supreme Court judgments and case files stored in a database.
Benefit: Lawyers get fast, context-specific summaries without manually searching through
thousands of pages.
 Use Cases of RAG

3. Healthcare:

Scenario: A hospital chain in India uses RAG to pull the latest treatment guidelines from
WHO and the Indian Council of Medical Research.
Benefit: Doctors get accurate, updated recommendations during consultations.

4. Market Research:

Scenario: A fintech company in Bangalore uses RAG to analyze competitor reports and
news feeds before making investment decisions.
Benefit: Decision-making is based on fresh, relevant market data.
Thank You!