Artificial Intelligence:

AI is the field of (computer) science that focuses on building machines that can think, learn
and act intelligently like humans, meaning we are trying to build a machine that perform
tasks that normally require human intelligence. Unlike traditional computer programs that
follow rigid instructions, AI systems are designed to perceive, reason, learn and adapt. The
Ultimate Goal is to enable computers to think and act intelligently either through
cooperation or autonomously.

Key Aspects of AI are:

- Cognitive Modelling: Think like humans

- Turing Test Approach: Act like humans

- Logic-based reasoning: Think rationally

- Rational agent approach: Act rationally

So, what is an intelligent system that is often termed with the use of AI?

An intelligent system is any system that uses AI techniques to analyse data, make decisions
and improve over time. It typically follows this (ctrl + click).

To understand intelligent systems better we have to look at their main components:

1. Sensing(input):
The system collects data from its environment. This could be text, images, audio or
any other form of input.
2. Reasoning(processing):
The system then interprets the data and makes logical inferences or decisions and
uses techniques such as search algorithms or rule-based reasoning etc.
3. Learning(adaptation):
Intelligent systems improve performance over time. Machine learning allows them to
detect patterns and adjust decisions based on past experiences like a
recommendation system improving its suggestions as it learns user preferences.
4. Acting(Output):
The system performs actions or provides responses based on its reasoning which can
either be physical or informational.

Intelligent Systems are defined by their Autonomy where they can operate without constant
human intervention, Adaptability where they adjust to changes in the environment or input
data, Goal-oriented behaviour where their actions are directed toward achieving objectives,
Rationality where they make decisions that are expected to maximize success and
Interactivity which gives them the ability to interact with humans or other systems.
Examples like Siri and Alexa (Virtual Assistants), Netflix and YouTube suggestions
(recommendation systems) etc.

The most common intelligent systems that can be understood easily are Reactive Systems
and Knowledge-based Systems.

In reactive systems the system respond directly to stimuli like how our reflexes react, thus
the word reactive systems. It does not learn nor plan ahead, for example a thermostat that
switches heating on/off depending on the temperature it records at present.

Knowledge-based systems on the other hand store structured knowledge and reason with it.
They can solve problems more often than not but often lack learning ability like for example,
Expert systems for medical diagnosis using the patients symptoms, data from perception
through auditory and imagery to look for out of the ordinary findings and such.

(Out of Syllabus)

There are also two other systems that dive deeper into this branch called Learning Systems
(Deeper Intelligence) and Autonomous Intelligence Systems.

Learning Systems improve their performance over time by learning from data and are
capable of adaptation. They mainly use Neural Networks and fall under the category of Deep
Learning. Using models trained for specific use scenarios like image recognition.

Autonomous Intelligent Systems perceive, reason, learn and act with minimal human
guidance like self-driving cars.

(End of Out of Syllabus)

So all in all, An intelligent system is the embodiment of AI principles, a machine that can
perceive, reason, learn and act. While today’s systems often operate at surface levels
(reactive and knowledge driven) advancements in machine learning and robotics are
gradually pushing them toward deeper intelligence and autonomy. Understanding the basics
of intelligent systems lays the foundation for exploring more advanced topics in artificial
Intelligence.

One of the major contributions of studying AI is it helps us understand natural intelligence

too. For example, to program a self-driving car we must know how humans drive, first the
eyes perceive the roads and surroundings (vision, sensors, perception), reason with the
brain (decision making) and act with the body (motor control of hands and legs.

We also have differentiable characteristics between Human and Artificial Intelligence:

Human Intelligence Artificial Intelligence

 1. Learns from experience 1. Learns from data
2. Uses common knowledge 2. Encoded knowledge + algorithms
3. Adapts flexibly to new situations 3. Generalization limited and focused on
task-specific solutions
4. Uses natural language for communication 4. NLP systems try to mimic this
5. Combines reasoning with perception and 5. AI agents integrate sensors + processing +
action actuators

The AI Problem:

The ambition and central challenge of AI is to build machines that can behave intelligently,
perceive the world, reason about what they perceive, learn from experience, use natural
language and solve problems intelligently.

- Perceive to take in information from environment using vision, sound and sensors.
- Process information using reasoning to make decisions using logic and knowledge.
- Learn and adapts on its environment and improves with experience.
- Interact with the world through actions.

Early AI pioneers believed machines could soon replicate most human intelligence during
the 1950s – 1970s, tasks like playing chess against an opponent, solving math problems or
understanding languages were thought easy.

But the reality was many tasks required handling huge amounts of data and complex
choices, this led to Combinatorial Explosion. In many tasks the number of choices or
possible states grew exponentially, for example in chess the number of possible positions to
place the pieces exceeded the number of atoms in the universe. So as a result Brute force
method failed and smarter strategies like heuristics and pruning were needed.

Types of tasks in AI:

a) Mundane Tasks (everyday intelligence)

 Perception: vision, speech, pattern recognition

 Natural Language: understanding, generation, translation

 Commonsense reasoning

b) Formal Tasks (academic/intellectual)

 Games: chess, checkers

 Mathematics: logic, calculus, theorem proving

c) Expert Tasks (specialized, professional)

 Medical diagnosis

 Financial analysis

 Scientific research

What are commonsense reasoning and Heuristics?

The ability for AI to use background knowledge that humans take for granted like making
logical decisions based on everyday knowledge is termed commonsense reasoning. It allows
AI to interpret situations, understand hidden facts and avoid absurd conclusions.

Heuristics on the other hand is a rule of thumb or shortcuts to guide search and decision
making, applying practical methods or educated guesses. They help AI systems make
decisions faster when exact solutions are too complex or time consuming but they do not
guarantee optimal solutions but aim for good-enough results efficiently.

NLP:

Natural Language Processing is a branch of AI that deals with understanding, interpreting

and generating human language so that systems and computers can communicate with
humans. Examples like chatbots/voice assistants, language translation, sentiment analysis
etc.
it is one of the most difficult branches since human language can be very ambiguous,
context dependent and flexible. Example: Bank can mean either a bank or a river bank,
Sarcastic remarks can also be positive ones too depending on the tone like “That’s crazy”.

Key components in NLP are:

- Syntax: Structure or grammar of sentences like “I am going home tomorrow” is valid

but “ home I going tomorrow am” is not.
- Semantics: Meaning of words and sentences (Dog bites man vs Man bites dog)
- Pragmatics: Meaning in context ( “Can you open the door?” – not about ability, but a
polite request)
- Sentiment Analysis: Meaning in tone and emotion (“I love this place” – positive. “I
love being ignore”- negative (sarcastic))

Applications of NLP:
Language Translation → Google Translate
Chatbots → Customer service bots, virtual assistants

Speech-to-Text → Dictation, subtitles

Text Summarization → AI-generated summaries of articles

Sentiment Analysis → Social media monitoring

Problem Spaces:

A problem space is the set of all possible or potential states that can be reached while
solving a problem, along with the rules (operators) for moving between these states. It
represents the universe of possibilities an AI System can explore to find a solution.

Components of a Problem Space:

1. States: Represents every possible scenario or configuration that could arise in the system.

2. State Space: Collection of all possible states that is reachable by applying sequence of
operators from initial state.

3. Operators: Rules that define how to move from one state to another

4. Paths: Paths are sequence of states connected by operators that lead from starting state
to goal state.

Example for a maze solving robot:

State: specific location or configuration in the maze (x,y position)
State Space: Collection of all possible states (positions) the robot could occupy during
navigation by moving step by step.

Operators: Directions and how much it can move(action/transitions) between these states.
(Move x cell(s) north, south, east, west )

Path: Path traced by the robot from maze entrance to exit.

Therefore, solving the maze involves searching the state space using allowed operators
(action/moves), to find a path from the initial state to the goal state.

Components of a Problem in AI:

A problem can be formally defined by four components:
1. Initial State – where the agent starts.

2. Goal State – the desired outcome.

3. Operators (actions/transitions)- rules that define how to move from one state to another.

4. Path/Path Cost or Restrictions/Constraints

Inside problem space we have state space representation. The graph/tree formed by all
possible states and transitions is called state space. Each node is a state, and each edge is an
operator, traversing the state space to reach the goal is called the search process.

Well defined vs Ill Defined:

NQueens- for 4x4cell