AI-Module-1-Important-Topics

For more notes visit

https://rtpnotes.vercel.app

Playlist to refer

https://youtube.com/playlist?list=PLnzz0gSUYIN3mo-LB-
l2Vfadz6HiPi8Po&feature=shared

AI-Module-1-Important-Topics
1. History of AI
2. AI-Levels, Goals and Types
Levels of AI:
Goals of Artificial Intelligence (AI):
Types of AI:
3. Structure of an Agent
Structure of an AI Agent
1. Architecture:
2. Agent Function:
3. Agent Program:
Formula:

4. Intelligent and Rational Agent

Intelligent Agents
Rational Agents
5. PEAS description
Example 1:
Example 2
6. Properties of task environment
Task Environment Types
 Fully Observable vs Partially Observable
Single Agent vs Multi Agent
Deterministic vs Stochastic
Episodic vs Sequential
Static vs Dynamic
Discrete vs Continuous
Known vs Unknown
7. Types of Agents
What is an Agent?
Actuators and Effectors
1. Simple Reflex Agent
2. Model-based reflex agent
3. Goal-based agents
4. Utility Based Agent
5. Learning agent

1. History of AI
Here’s the timeline formatted as a table for a more structured presentation:

Year Event
1943 Early Beginnings: McCulloch & Pitts: Boolean circuit model of the brain.
1950 Turing's Vision: Turing's "Computing Machinery and Intelligence."
1956 Birth of AI: Dartmouth meeting; The term "Artificial Intelligence" is adopted.
1950s Initial Promise: Early AI programs, including:
- Samuel's Checkers Program
- Newell & Simon's Logic Theorist
1955-1965 “Great Enthusiasm”:
- Newell & Simon: General Problem Solver (GPS)
- Gelertner: Geometry Theorem Prover
- McCarthy: Invention of LISP
1966-1973 Reality Dawns: Realization that many AI problems are intractable and
limitations of existing neural network methods.
 Year Event
1969-1985 Adding Domain Knowledge: Development of knowledge-based systems and
rule-based expert systems, such as:
- DENDRAL
- MYCIN
- However, these systems were brittle and did not scale well in practice.
1986 Rise of Machine Learning: Neural networks return to popularity; major
advances in machine learning algorithms and applications.
1990 Role of Uncertainty: Introduction of Bayesian networks as a knowledge
representation framework.
1995 AI as Science: Integration of learning, reasoning, and knowledge
representation.
1995-2011 Intelligent Agents: Development of intelligent agents as a major focus in AI.
2011- Deep Learning, Big Data, and Artificial General Intelligence:
Present
- 2018 Achievements:
- Alibaba's language processing AI outperforms top humans in a Stanford
University test (82.44 vs 82.304).
- Google Duplex: AI assistant can book appointments over the phone with a
"nearly flawless" imitation of human speech.

2. AI-Levels, Goals and Types

Artificial Intelligence (AI) is the field of computer science focused on creating machines that
can think and act intelligently, similar to how humans do.

Levels of AI:

1. Narrow AI: Machines perform specific tasks better than humans, such as facial
recognition or playing chess. Current AI research is focused here.
2. General AI: A machine that can perform any intellectual task at the same level as a
human, though this has not been achieved yet.
3. Strong AI: Machines that surpass human intelligence in many areas, being able to
perform tasks better than humans across a wide range of domains.
Goals of Artificial Intelligence (AI):

1. Create Expert Systems: Develop systems that can perform tasks intelligently, learning,
explaining, and advising users, much like a human expert.
2. Implement Human Intelligence in Machines: Build machines that can think, learn,
understand, and behave like humans.

Types of AI:

General-purpose AI: Like the robots seen in science fiction, which can perform a wide
range of tasks. However, replicating the complexity of the human brain is extremely
difficult, and we still don't fully understand how it works.
Special-purpose AI: Focuses on specific tasks, which is more achievable. Examples
include AI used in chess or poker programs, logistics planning, voice recognition, web
search, data mining, medical diagnosis, and self-driving cars.

3. Structure of an Agent
Structure of an AI Agent

The structure of an AI agent explains how an agent works by combining two key parts: the
architecture and the agent program.

1. Architecture:

This is the hardware or platform on which the agent runs. It could be a physical machine
(like a robot) or a computer system. Think of it as the "body" of the agent that allows it to
function.

2. Agent Function:

The agent function is like the brain behind the agent's actions. It takes the percept (what
the agent senses from the environment) and decides what action to take in response. It's
represented as:
f: P -> A* (where "P" is the percept and "A" is the action).

3. Agent Program:
 The agent program is the actual code or software that makes the agent function work. It
runs on the architecture and translates the agent’s observations into actions.

Formula:

Agent = Architecture + Agent Program: An AI agent is a combination of the physical

structure (architecture) and the intelligent software (agent program).

4. Intelligent and Rational Agent

Intelligent Agents

An intelligent agent is like a smart, independent system that senses its surroundings, makes
decisions, and takes actions to achieve a goal. It uses sensors to gather information and
actuators to interact with the environment.

Autonomous: It operates on its own without constant human guidance.

Goal-oriented: It works toward specific goals, like keeping a room at the right temperature
or navigating a car safely.
Learning: An intelligent agent can learn from its environment and improve its decisions
over time.

Example: A thermostat
A thermostat is an intelligent agent that senses the room temperature (using sensors) and turns
the heater on or off (using actuators) to maintain the desired temperature.

Rational Agents

A rational agent is an agent that makes decisions with a clear goal in mind and tries to achieve
the best possible outcome based on what it knows.

Clear Preferences: A rational agent knows what it prefers or what it is trying to achieve
(its goal).
Models Uncertainty: It understands that it might not know everything, so it makes
decisions based on the best possible information.
Maximizes Performance: The agent aims to take actions that lead to the best results,
according to a performance measure. For example, in a game, this could mean winning,
 and in real life, it could mean solving a problem efficiently.
Acts Rationally: It always tries to take the best possible action to reach its goal, given the
situation it is in.

5. PEAS description
PEAS is a framework that helps describe how an AI agent works by breaking it down into four
parts:

P (Performance Measure): This is how we judge the success of the AI's actions. For
example, in an automated taxi, performance could be measured by how safely and
efficiently it gets passengers to their destination.
E (Environment): This is the world the AI operates in. For a taxi, this would be the city
streets, weather, traffic, and pedestrians.
A (Actuators): These are the parts of the AI that allow it to take action. In a taxi, the
actuators would be the steering wheel, brakes, and accelerator.
S (Sensors): These help the AI gather information about the environment. In the taxi's
case, sensors could include cameras, GPS, and radar.

Example 1:

Task of designing a self-driving car

Performance Measure
Safe, fast, legal, comfortable trip
Environment
Roads, other traffic, pedestrials
Actuators
Steering wheel, accelerator, brake, signal, horn

Sensors
Cameras, Speedometers, GPS, Odometer, engine sensors etc

Example 2
6. Properties of task environment
Task Environment Types

Fully observable vs partially observable

Single agent vs Multi agent
Deterministic vs stochastic
Episodic vs sequential
Static vs dynamic
Discrete vs continous
Known vs unknown
Here's a simplified version of the concepts:

Fully Observable vs Partially Observable

Fully Observable: If an agent's sensors provide complete information about the

environment at all times, it’s fully observable. The agent doesn't need to keep track of
anything internally since all relevant details are available.
Partially Observable: If the sensors are inaccurate, noisy, or miss some details, the
environment is partially observable. This could be because parts of the environment aren't
detected.
Unobservable: If there are no sensors, the environment is unobservable.

Single Agent vs Multi Agent

Single Agent: When there is only one agent in an environment, it's called a single-agent
environment.
Multi-Agent: When multiple agents are involved, it's a multi-agent environment. For
example, chess is competitive, while a taxi-driving scenario can be both competitive (for
parking) and cooperative (avoiding collisions).

Deterministic vs Stochastic

Deterministic: If the agent’s actions completely determine the next state of the
environment, it’s deterministic. There's no uncertainty.
Stochastic: If the outcome is uncertain and has probabilities, it's stochastic. If the
environment is partially observable, it can seem stochastic even if it’s not.

Episodic vs Sequential

Episodic: The agent's experience is divided into episodes where each one is independent
of the others. The agent makes decisions based on the current situation only.
Sequential: Actions have consequences for future decisions. In games like chess or taxi
driving, short-term choices affect long-term outcomes.

Static vs Dynamic

Static: The environment doesn’t change while the agent is thinking. Time passing doesn’t
affect the environment.
 Dynamic: The environment can change while the agent is thinking, so the agent has to
make decisions in real-time.
Semi-Dynamic: The environment doesn’t change, but the agent’s performance may vary
over time (e.g., in chess with a clock).

Discrete vs Continuous

Discrete: The environment has a limited number of distinct states, actions, and percepts.
Chess is an example of a discrete environment.
Continuous: The environment is smooth, with many possible states and actions, like in
taxi driving where the position and speed can change continuously.

Known vs Unknown

Known: The outcomes of actions are known, so the agent can make informed decisions.
Unknown: The agent has to explore and learn how the environment works to make good
decisions. Even in a fully observable environment, if the agent doesn’t know the rules (like
in a new game), it’s unknown.

7. Types of Agents
What is an Agent?
 Think of an "agent" as a robot or a computer program that can "see" what's happening
around it and "do" things to change or respond to its surroundings.
Operates in an environment:.
Perceives its environment through sensors
Acts upon its environment through actuators/effectors
Has goals
In simpler terms, an agent is like a smart helper that watches, thinks, and acts to
accomplish something in the world it operates in.

Actuators and Effectors

Actuators and effectors are devices or components that allow an agent (like a robot or a
system) to take action and interact with its environment.

Actuators: These are the mechanisms that physically make something happen. In robots,
for example, actuators are like muscles—they control movements such as turning wheels,
moving arms, or opening grippers.
Effectors: These are the parts of the agent that cause the actual change in the
environment. In a robot, effectors might be things like wheels, robotic arms, or grippers.
The actuators control the effectors. For example, an actuator moves a robot’s arm, and
the arm (the effector) is what interacts with the object or the environment.

1. Simple Reflex Agent

A Simple Reflex Agent is a type of artificial agent that works by reacting to its current situation
(called the "percept") without considering any history or future consequences.

Choose actions only based on the current percept: The agent decides what to do just
by looking at the current situation it's in, without worrying about past events or future
outcomes.
Rational only if a correct decision is made based on the current percept: The agent
can be called "smart" or "rational" only if it can make the right decision based solely on
what it currently observes.
Their environment is completely observable: The agent can see or sense everything it
needs to know about its environment right now.
Condition-Action Rule: The agent follows simple rules that connect what it sees
(condition) to what it should do (action). For example, "If the light is red (condition), then
 stop (action)."

2. Model-based reflex agent

A Model-based Reflex Agent is a more advanced type of agent that can make better decisions
by understanding the world and keeping track of changes

A model of the world to choose their actions: Instead of just reacting to what it sees in
the moment, this agent builds a mental picture (or "model") of how the world works to
guide its actions. It also keeps track of what's happening, even if it can't see everything at
once.
Model: This is the agent's understanding of how things usually happen in the world. For
example, it knows that if it drops something, it will fall.
Internal State: The agent remembers things that it can’t currently see. This memory is
built up from everything it has observed so far (called "percept history"). For example, if
the agent has walked into another room, it remembers where it came from, even if it can’t
see it now.
Updating the state: To keep this memory accurate, the agent needs to:
Understand how the world changes over time.
Know how its own actions will change the world around it.
3. Goal-based agents

Goal-based agents are a type of AI that make decisions to achieve specific goals. Unlike
reflex agents, they don't just react to situations; they use knowledge to plan their actions.
This makes them more flexible, as you can change their behavior by adjusting their goals.
 A "goal" simply describes the outcome the agent is trying to achieve.

4. Utility Based Agent

Utility-based agents decide what to do by ranking different outcomes, or states, based on

how useful or beneficial (utility) they are.
Unlike goal-based agents, they don’t just aim for one goal but balance multiple goals
that might conflict.
When some goals can't be fully achieved, utility-based agents choose the action that gives
the best possible result.
They also factor in uncertainty, considering both the chances of success and how
important each goal is.
5. Learning agent

A learning agent is an AI that improves its performance by learning from past experiences. It
starts with basic knowledge and then adapts automatically as it learns. It has four main parts:

1. Learning element: This part learns from the environment and makes improvements.
2. Critic: It provides feedback to the learning element, showing how well the agent is doing
compared to a set standard.
3. Performance element: This part chooses the actions the agent will take.
4. Problem generator: It suggests new actions that help the agent explore and gain more
useful experiences.