Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

Module 1 – Introduction to
Artificial Intelligence and
Machine Intelligence

1. Introduction to Artificial Intelligence

1.1 What is Artificial Intelligence?
Simple idea:
Artificial Intelligence (AI) is about making computers and machines behave intelligently –
they can learn, reason, and make decisions like humans for specific tasks.

Formal definition (John McCarthy, 2004):

“Artificial Intelligence is the science and engineering of making intelligent machines,

especially intelligent computer programs.”

So, AI is both:

 a science – studies principles of intelligence,

 an engineering field – builds intelligent systems.
Everyday examples:

 Google Search suggestions

 Face unlock in mobiles
 Siri / Alexa / Google Assistant
 YouTube / Netflix recommendations
 Self‑driving cars
These systems sense their environment, process data, and take actions to achieve goals.

1.2 How Does AI Work?

AI systems generally work in four stages:

1. Data Collection

o Collect large amounts of data from sensors, databases, logs, images, text, etc.
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

2. Learning (Training)

o Use algorithms to find patterns in data.

o Example: learn the difference between “spam” and “not spam” emails.
3. Reasoning and Decision Making

o Based on learned patterns, the AI selects actions: classify, predict,

recommend, or control.
4. Self‑Correction / Feedback

o Compare output with correct answers (ground truth).

o Adjust internal parameters to reduce error (improve accuracy over time).
Key concepts:

 Learning – AI modifies its behaviour using data.

 Reasoning – AI chooses suitable rules/algorithms for a task.
 Self‑correction – AI refines models continuously to become more accurate.

1.3 Advantages of Artificial Intelligence

1. Works well with huge data
o Can process millions of records faster than humans.
2. High speed and efficiency
o Performs complex calculations in milliseconds.
3. Consistency and accuracy
o Does not get tired or distracted, gives uniform performance.
4. 24 × 7 availability
o Chatbots, monitoring systems, online services run continuously.
5. Resource optimization
o AI can schedule production, allocate resources, and reduce wastage.
6. Automation of complex tasks
o Robotics, autonomous vehicles, fraud detection, etc.
7. Predictive maintenance
o Predict machine failures and reduce downtime.
8. Better products and services
o Personalised recommendations, better diagnosis, smarter devices.

1.4 Disadvantages / Limitations of AI

Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

1. High cost
o Requires powerful hardware, large storage, and skilled developers.
2. Need for technical expertise
o Designing, training, and deploying models requires special skills.
3. Lack of trained professionals
o Shortage of experts can slow deployment.
4. Dependence on data quality
o Incomplete, biased, or noisy data can produce wrong or unfair results.
5. Limited generalisation
o Most current AI is narrow – performs only specific tasks; fails outside
training scenarios.
6. Ethical and social concerns (covered in later modules)
o Job loss, bias, privacy issues, misuse of AI.
Exam tip: be ready to list any 4–6 advantages and 4–6 disadvantages with short
explanations.

1.5 History of Artificial Intelligence (Timeline

View)
Year Event / Milestone
1943 Warren McCulloch & Walter Pitts propose first neural network model
1950 Alan Turing proposes the Turing Test (“Can machines think?”)
1956 John McCarthy coins the term “Artificial Intelligence” at Dartmouth
Conference
1997 IBM Deep Blue defeats world chess champion Garry Kasparov
2011 Apple Siri launched – popular voice‑based personal assistant
2010s Rapid progress in deep learning, image and speech recognition
2020s Large Language Models (LLMs) like ChatGPT, generative AI applications

Turing Test (basic idea):

If a human evaluator converses (through text) with both a human and a machine and cannot
reliably identify which is which, the machine is said to exhibit intelligent behaviour.

2. Types of Artificial Intelligence

VTU syllabus splits types of AI in two ways:
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

1. Based on capabilities – Weak AI vs Strong AI

2. Based on functionality – Reactive, Limited Memory, Theory of Mind,
Self‑Awareness

2.1 Based on Capabilities

(a) Weak AI (Narrow AI)
 Designed for one specific task only.
 Cannot perform tasks outside its training domain.
 Most real‑world AI today is weak AI.
Examples:

 Siri / Alexa / Google Assistant – understand limited commands.

 Spam filters – detect junk emails.
 Weather prediction systems.
 Stock price prediction models.
 Face recognition in photos.
Characteristics:

 High performance in a narrow task.

 No self‑awareness or general understanding of the world.

(b) Strong AI (Artificial General Intelligence – AGI / ASI)

 Aims to mimic human‑level intelligence.
 Can perform any intellectual task that a human can.
 Able to reason, understand, learn, and adapt in different domains.
Status:

 Still theoretical; we do not have true AGI yet.

 Many debates on its possibility, risks, and ethics.

2.2 Based on Functionality

(a) Reactive Machines
 Simplest type of AI.
 React only to current input; they do not store past experiences.
 No learning or memory.
Example:
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

 IBM’s Deep Blue (chess‑playing computer).

o Evaluates current board position and selects the best move.
o Does not remember previous matches.

(b) Limited Memory

 Can store and use recent past data for short periods.
 Learn from history but do not store lifelong experiences.
Examples:

 Self‑driving cars – remember positions and speeds of nearby vehicles for safe driving.
 AlphaGo – uses previous moves in a game, updates strategy while playing.
Important related concepts:

 Reinforcement Learning – learns by trial and error, receives

rewards/punishments, and improves.
 LSTM (Long Short‑Term Memory) – neural network model that remembers
recent information more strongly.

(c) Theory of Mind (future AI)

 Aim: machines that understand human emotions, beliefs, intentions, and
thoughts.
 Would adapt responses based on user’s emotional state.
 Still theoretical / research stage.

(d) Self‑Awareness (future AI)

 Highest level of AI.
 Machines would have consciousness and awareness of their own existence.
 Could understand their feelings and those of others.
 Currently does not exist – only concept for future discussions.

2.3 AI vs Augmented Intelligence vs Cognitive

Computing
1. Artificial Intelligence (AI):

o Tries to make machines autonomous – perform tasks that usually need

human intelligence.
2. Augmented Intelligence:

o Focuses on assisting humans rather than replacing them.

Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

o Example: system that highlights important lines in a long report so that the
human can read faster.
3. Cognitive Computing:

o Systems that simulate human thought processes (understand, reason,

learn) to support better decisions.
o Example: IBM Watson for medical diagnosis.
Exam point: be able to explain in 2–3 lines how augmented intelligence is assisting humans,
whereas strong AI aims to match/surpass humans.

3. Machine Learning and Deep Learning

3.1 What is Machine Learning (ML)?
Machine Learning is a subset of AI that focuses on:

“Teaching machines to learn from data and make decisions without being explicitly
programmed for every rule.”

Instead of writing rules manually, we:

 Provide data and examples.

 The algorithm learns patterns on its own.
How ML works:

1. Pattern finding – algorithms detect regularities in data.

2. Learning from experience – improve automatically from previous outputs.
3. Self‑correction – update internal parameters when wrong, to reduce error.
Real‑life analogy:

 Number series 10, 20, 30, 40 → machine learns pattern “+10” and predicts next as 50.

3.2 Relationship: AI, ML, Deep Learning, NLP

 AI – overall goal: make machines intelligent.
 ML – way to achieve AI by learning from data.
 Deep Learning (DL) – subset of ML using multi‑layer neural networks.
 Natural Language Processing (NLP) – enables machines to understand and
generate human language.
Think of it as:

AI (big circle) ⟶ contains ML ⟶ contains Deep Learning

Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

3.3 Traditional Programming vs Machine Learning

Traditional Programming
 Programmer writes explicit rules using languages like C, C++, Java, Python.
 Program = Rules (logic) + Data → Output
 Logic is fixed unless programmer modifies code.
Example:
To identify if an email is spam, programmer writes many if–else rules (words like “lottery”,
“prize”, etc.).

Machine Learning
 Programmer provides data + correct outputs (labels).
 Algorithm automatically learns rules/model from this data.
 Program = Data + Output → algorithm learns Rules/Model
Example:
Provide thousands of “spam” and “not spam” emails; model learns patterns and can classify
new emails.

Key differences:

Aspect Traditional Programming Machine Learning

Who creates Human programmer Algorithm learns from data
rules?
Adaptability Needs re‑coding for new Model can be retrained with new data
patterns
Use case Simple, well‑defined logic Complex tasks: vision, speech,
recommendations

3.4 Deep Learning (Basic Idea)

Deep Learning uses artificial neural networks with many layers.

 Inspired by the human brain.

 Learns hierarchical features:
o In images: edges → shapes → objects → scenes.
o In speech: raw audio → phonemes → words → meaning.
Deep learning is especially powerful for:
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

 Image recognition (face detection, medical imaging).

 Speech recognition (Siri, Google Voice).
 Text understanding and generation (chatbots, LLMs).

4. Machine Intelligence
4.1 Defining Intelligence
According to the VTU notes, intelligence can be seen in different domains (Howard
Gardner’s multiple intelligences idea, simplified):

 Linguistic intelligence – using language effectively (writers, speakers).

 Musical intelligence – understanding rhythm, tone, music (musicians).
 Logical‑mathematical intelligence – reasoning, problem solving (scientists,
engineers).
 Spatial intelligence – visualising space, rotating objects mentally (architects,
pilots).
 Bodily‑kinesthetic intelligence – using body skillfully (athletes, dancers).
 Intrapersonal intelligence – understanding one’s own feelings and motivations.
 Interpersonal intelligence – understanding others’ emotions and intentions
(teachers, counsellors).
Artificial Intelligence System:
If a computer system shows one or more of these capabilities (e.g., logic, language
understanding, perception), we say it has artificial intelligence.

4.2 Components of Intelligence

The core components (important for exam):

1. Reasoning
2. Learning
3. Problem Solving
4. Perception
5. Linguistic Intelligence

4.2.1 Reasoning
Definition:
Reasoning is the process of drawing conclusions and making decisions from available
information.

Two major types:

Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

1. Inductive Reasoning

o From specific observations → general rule.

o Example: See many swans which are white → conclude “All swans are white.”
o Risk: conclusions may be wrong if data is limited.
2. Deductive Reasoning

o From general rule → specific conclusion.

o Example:
 All swans are birds.
 This is a swan.
 Therefore, this is a bird.
AI systems often use logical reasoning to infer new facts from known rules.

4.2.2 Learning
Learning is acquiring new knowledge or skills via study, experience, or teaching.

Different types (with simple examples):

1. Auditory Learning – learning by listening

o E.g., listening to lectures or podcasts.

2. Episodic Learning – remembering ordered events

o E.g., recalling steps in a lab experiment.

3. Motor Learning – learning through movements

o E.g., learning to play a sport or type on a keyboard.

4. Observational Learning – learning by watching others

o E.g., a child imitates how parents do things.

5. Perceptual Learning – recognizing familiar situations/objects

o E.g., identifying road signs you have seen before.

6. Relational Learning – understanding relationships between things

o E.g., adjusting spices in cooking based on last time’s taste.

7. Spatial Learning – using visual maps/pictures

o E.g., mentally planning a driving route using a map.

8. Stimulus‑Response Learning – reacting to stimuli

o E.g., quickly pulling your hand from a hot pan.

AI systems also “learn” in different ways – supervised, unsupervised, reinforcement, etc.
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

4.2.3 Problem Solving

Definition:
Problem solving is the process of finding a solution to a challenge or obstacle.

Steps:

1. Identify the problem – understand what is wrong or what goal we must reach.
2. Generate possible solutions – think of different strategies.
3. Select the best solution – based on knowledge, constraints, and goals.
4. Apply and evaluate – implement solution and check result.
AI uses search algorithms, heuristics, and planning methods for problem solving.

4.2.4 Perception
Perception is the ability to:

 Receive information via senses (eyes, ears, etc.).

 Interpret that information to understand the environment.
Humans use sensory organs. AI systems use sensors:

 Cameras (vision)
 Microphones (sound)
 Infrared sensors, LIDAR, etc.
Perception allows systems like self‑driving cars and robots to sense their surroundings
and act accordingly.

4.2.5 Linguistic Intelligence

In AI context, linguistic intelligence is the ability to:

 Understand and generate human language in spoken or written form.

Examples:

 Chatbots and virtual assistants.

 Machine translation (Google Translate).
 Text summarization and question answering.

4.3 Human Intelligence vs Machine Intelligence

Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

A few important comparison points (exam oriented):

Aspect Human Intelligence Machine Intelligence

Perception Recognises patterns from Uses rules, models and data, may
partial info fail with noisy input
Memory & Recall Remembers by association Uses structured storage / search
and patterns algorithms
Handling missing Can guess/estimate missing Struggles; needs explicit handling
data information or imputation
Learning Slow but flexible, learns Needs large labelled datasets (in
from few examples general)
Generalisation Can apply knowledge across Usually narrow to trained tasks
domains
Creativity, Can create art, imagine, feel Limited; simulates creativity and
emotions emotions emotions through patterns

Key idea:
Humans are better at common sense, creativity, emotions, while machines are better at
speed, accuracy, and handling huge data.

5. Agent and Environment in AI

5.1 What is an Agent?
Agent:

Anything that can perceive its environment through sensors and act upon that
environment through actuators/effectors.

Examples:

 Human agent – sensors: eyes, ears; effectors: hands, legs, mouth.

 Robot agent – sensors: camera, proximity sensors; effectors: motors, wheels, arms.
 Software agent – sensors: input data, files, network; effectors: outputs, API calls.

5.2 Key Terms

 Performance Measure:

o A way to judge how successful an agent is (e.g., time taken, accuracy, safety).
 Behaviour of Agent:
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

o The action it performs after receiving percepts.

 Percept:

o The input received at a particular instant (e.g., camera image at time t).
 Percept Sequence:

o The complete history of all percepts received so far.

 Agent Function:

o A mapping from percept sequences to actions.

o Conceptual: f: PerceptSequence → Action
Rational Agent:

 Chooses the best possible action to maximize performance, based on:

a. Performance measure
b. Percept sequence
c. Prior knowledge
d. Possible actions

5.3 PEAS Representation

To design an intelligent agent, we use PEAS:

 P – Performance Measure – success criteria

 E – Environment – surroundings where agent acts
 A – Actuators – devices to take actions
 S – Sensors – devices to perceive environment
Example – Self‑driving taxi:

 P: Safety, speed, passenger comfort, legality

 E: Roads, traffic, pedestrians, weather
 A: Steering, accelerator, brake, horn
 S: Cameras, GPS, LIDAR, speed sensors

5.4 Types of Agents (3.4.4)

1. Simple Reflex Agent

o Chooses action only based on current percept (“if condition then action”).
o Works well only in fully observable environments.
o Can get stuck in loops in partially observable environments.
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

2. Model‑Based Reflex Agent

o Maintains an internal state (model of the world).

o Updates state based on:
 How the world evolves.
 How its own actions change the world.
3. Goal‑Based Agent

o Has explicit goals (desired states).

o Chooses actions that move it closer to goals.
o More flexible than reflex agents.
4. Utility‑Based Agent

o Uses a utility function to measure desirability of states.

o Can make trade‑offs when there are multiple conflicting goals.
5. Learning Agent

o Improves performance over time using experience.

o Has four components:
 Learning element – learns from environment.
 Critic – gives feedback about performance.
 Performance element – selects external actions.
 Problem generator – suggests new, exploratory actions.

5.5 Types of Environments (3.4.6)

Important environment properties:

1. Discrete vs Continuous

o Discrete – finite number of states/actions (e.g., chess).

o Continuous – infinite/large number of states/actions (e.g., self‑driving car).
2. Known vs Unknown

o Known – agent knows outcome of all actions.

o Unknown – agent must learn the outcome.
3. Observable vs Partially Observable

o Fully observable – agent can sense the entire state (e.g., chess board).
o Partially observable – agent can see only part (e.g., driving in fog).
4. Static vs Dynamic vs Semi‑dynamic
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

o Static – environment does not change while agent is thinking (e.g.,

crossword).
o Dynamic – environment changes over time (e.g., traffic).
o Semi‑dynamic – environment static but performance measure changes.
5. Single Agent vs Multi‑Agent

o Single – only one agent (e.g., vacuum cleaner robot alone).

o Multi – many agents interacting (e.g., game of chess, autonomous vehicles in
traffic).
6. Accessible vs Inaccessible

o Accessible – complete information available (e.g., empty room).

o Inaccessible – limited or noisy information (e.g., global stock markets).
7. Deterministic vs Non‑deterministic

o Deterministic – next state fully determined by current state and action (e.g.,
chess).
o Non‑deterministic – includes uncertainty (e.g., rolling dice).
8. Episodic vs Non‑episodic

o Episodic – each action is independent (past actions don’t affect future ones).
o Non‑episodic – current actions affect future ones (long‑term planning tasks).

6. Search in Artificial Intelligence

AI agents often solve problems by searching through possible states.

6.1 Components of a Search Problem

1. State Space – all possible configurations/states.
2. Start State – where search begins.
3. Goal Test – check whether current state is goal state.
4. Actions – operations that move from one state to another.
5. Path Cost – numerical cost associated with a path.
6. Solution – sequence of actions from start to goal.

6.2 Properties of Search Algorithms

 Completeness – guarantees to find a solution if one exists.
 Optimality – finds the best (minimum cost) solution.
 Time Complexity – how much time it takes.
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

 Space Complexity – how much memory it uses.

A “good” search algorithm should be reasonably fast and use acceptable memory.

6.3 Uninformed (Blind) Search Algorithms

These algorithms do not use any domain knowledge (no information on how close we
are to goal). They only use information from problem definition.

Important terms:

 Problem Graph – nodes (states) and edges (actions).

 Strategy – order in which nodes are expanded.
 Fringe (Frontier) – set of nodes generated but not yet expanded.
 Tree – expansion structure from start to all explored nodes.

6.3.1 Depth First Search (DFS) – Concept

 Explores as deep as possible along each branch before backtracking.
 Uses a stack (LIFO) for frontier.
Steps (simple):

1. Start from root node.

2. Expand one child and go deeper until you reach a leaf or dead‑end.
3. If goal not found, backtrack to last node with unexplored children.
4. Repeat until goal is found or tree is exhausted.
Advantages:

 Low memory consumption – stores only current path.

 May find solution quickly if it lies deep.
Disadvantages:

 Can get stuck in infinite paths (e.g., if graph has cycles).

 Not guaranteed to find optimal solution.
(Other uninformed algorithms listed in syllabus – BFS, Uniform Cost Search, IDDFS,
Bidirectional Search – can be introduced in class with diagrams.)

6.4 Informed (Heuristic) Search

Informed search uses additional heuristic information: an estimate of how close a state is
to the goal.
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

 Pure Heuristic Search – selects next node based purely on heuristic value
(smallest estimated distance to goal).
 Best‑First Search (Greedy Search) – always expands node with lowest heuristic
value first.
Heuristics make search faster, but may not always give optimal solution unless carefully
designed.

7. Knowledge Representation (Basics)

Knowledge Representation (KR) is how an AI system stores and uses knowledge about
the world so it can reason and act intelligently.

7.1 Knowledge‑Based Agent

 Has a knowledge base (KB) – collection of facts and rules.
 Uses inference to derive new facts from existing ones.
 Can answer queries and make decisions based on its KB.

7.2 Types of Knowledge (high‑level)

1. Declarative Knowledge – facts about the world

o Example: “Bangalore is in Karnataka.”

2. Procedural Knowledge – how to do things

o Example: steps of an algorithm.

3. Heuristic Knowledge – rules of thumb

o Example: “If patient has high fever and cough, suspect infection.”
4. Meta‑Knowledge – knowledge about knowledge

o Example: knowing which rules to use in which situation.

Knowledge representation formalisms (details in later modules/courses):

 Logic (propositional, first‑order logic)

 Semantic networks
 Frames
 Production rules

8. How to Use These Notes for VTU

Exams
Introduction to AI and Applications, 1BAIA103/203 Keertipriya Bellary

 Map topics to syllabus lines:

o “Introduction to Artificial Intelligence” – Sections 1, 2, 3 above.

o “Machine Intelligence” – Components of intelligence, human vs machine,
agents and environments.
o “Search Algorithms” – state space, properties, uninformed vs informed, DFS,
Best‑First.
o “Knowledge Representation” – basic ideas of KB, types of knowledge.
 For 2–3 mark questions:

o Memorise definitions: AI, ML, Deep Learning, agent, rationality, PEAS,

heuristic, etc.
o Be ready with differences tables (human vs machine intelligence,
traditional vs ML, weak vs strong AI).
 For 5–10 mark questions:

o Write short intro + neat headings + 4–6 points per heading.

o Use diagrams wherever possible:
 relationship between AI–ML–DL
 agent–environment diagram
 search tree sketch for DFS/BFS (rough).
 For 10–12 mark essay‑type questions:

o Cover definition, explanation, types, examples, advantages,

disadvantages.
o Conclude with applications or limitations to show understanding.