Introduction to Artificial Intelligence:

Intelligent Agents
• Based on "Artificial Intelligence: A Modern
Approach, 3rd Edition"
• Authors: Stuart Russell and Peter Norvig
• Publisher: Pearson Education
 What is Artificial Intelligence
• Intelligence --- ability to acquire and apply knowledge and
skills.
• Artificial - not natural/not occurring naturally /made by
human beings
• Artificial Intelligence (AI) is a branch of computer science that
deals with creating systems or machines capable of learning
independently, reasoning, making decisions, and performing
complex tasks that typically require human intelligence.
• AI uses programming and advanced models to process large
amounts of data that humans cannot easily manage. Its
applications include data analysis, decision-making,
predictions, and forecasting.
 What is Artificial Intelligence?
• AI is the science and engineering of making
intelligent machines, especially intelligent
computer programs.
• Key Areas: Machine learning, natural language
processing, robotics, computer vision, and
expert systems.
 Overview of Intelligent Agents
• Agent –something that acts(from latin ager –to do)
• Rational agent-one that acts to achieve best
outcome, in case of uncertainty best expected
outcome
• An intelligent agent perceives its environment
and acts upon it to achieve specific goals.
• A rational agent is always goal-driven and makes
the best decision based on available information.
• An intelligent agent can learn, adapt, and evolve
over time to improve its performance.
Components:
- Perception
- Reasoning
- Learning
- Action
 Agent Function and Programs
• Agent Function: Maps percept sequences to actions.
(abstract mathematical description)
• Agent Program: The implementation of the agent
function.
• Relationship: Flow from perception to action through
the agent program.
• The program run on some computing device with
physical sensor and actuators –called architecture
• Agent=Architecture+ Program
Characteristics of Intelligent Agents
• - Autonomy: Operates without human
intervention.
• - Reactivity: Responds to environmental
changes in real time.
• - Proactivity: Takes the initiative to achieve
goals.
• - Social Ability: Communicates effectively with
other agents and humans.
 Types of Agents
• - Simple Reflex Agents: Act only on current
perceptions.
• - Model-Based Reflex Agents: Maintain internal
states.
• - Goal-Based Agents: Act to achieve specific
objectives.
• - Utility-Based Agents: Optimize performance based
on a utility function.
• - Learning Agents: Improve performance through
experience.
Simple Reflex Agents: These agents react to specific inputs with predefined
actions, without considering the past. Operates on condition-action rule

Example: A thermostat that adjusts the temperature based on the current room
temperature. It only reacts to temperature changes. Or room cleaner

Limitation-
They have very limited intelligence
Not adaptive to changes in the environment
Model-Based Reflex Agents: These agents store some history of the
environment and use that to make decisions, allowing them to react based on
more than just the current state.
Model-Based Reflex Agent is a self-driving car.
It doesn't just react to the current road situation (like a simple reflex agent would), but it uses a model of
its environment (e.g., maps, sensors, and traffic data) to keep track of the car’s state and the surrounding
conditions. It can adjust its actions based on both immediate inputs (e.g., objects in the road) and its
internal model (e.g., remembering road rules, speed limits, or previously encountered traffic patterns).
Another example Room cleaner that takes has understanding or model of the house and obstacles

model refers to the internal representation or understanding that

the agent has of the world (or environment) it operates in
A model typically includes knowledge about:
•States of the environment: What the environment looks like at
any given moment.
•Transitions: How the environment changes over time, based on
the actions the agent takes or external factors.
•Actions and consequences: How the agent’s actions will affect
the environment and how the agent can predict those outcomes.

Limitation
Limited by complexity and Accuracy of their internal
models, which may not always reflect the real world
perfectly. If the environment changes unexpectedly or the
model is outdated, the agent may make incorrect
decisions.
 Goal-Based Agents: These agents act with the purpose of achieving specific
goals. They can plan ahead and choose actions that lead to goal fulfillment.

Example: A GPS navigation system that finds the best route to a destination,
considering traffic and road conditions.

Limitation: limited by the difficulty of finding the best solution, especially in complex or
uncertain environments. They may also require a lot of time and resources to plan and
achieve their goals.
Utility-Based Agents: These agents aim to maximize a utility function (or
happiness score) to make decisions that lead to the highest overall
benefit.
A Utility-Based Agent makes decisions by choosing the option that gives it
the most benefit or satisfaction.

It calculates the "utility" of different actions(How mush it will satisfy)

Example: A movie recommendation system that selects films based on
your past ratings and preferences to maximize your satisfaction.
An example
Product recommendation system:

User preferences
Product ratings/reviews
Price range
Trending products
Utility score
Past purchases
Learning Agents: These agents improve their behavior over time by learning from
past experiences or feedback.

Example: A chatbot that gets better at answering questions after interacting with
users, learning from each conversation.
1. Learning Element: Improves the agent's performance over time by learning
from past experiences or feedback.

2. Performance Element: responsible for making decisions or taking actions

based on the current state of the environment. It uses the learned knowledge to
interact.
3. Critic: Provides feedback on the agent's actions to help it evaluate and learn
from its mistakes or successes.

4. Problem Generator: Suggests new actions or strategies for the agent to

explore, fostering better learning through experimentation.
 The Agent Environment
Agent Environment : Surrounding in which the agent is operating

• Properties of the Environment:

• - Fully or Partially Observable
• - Deterministic or Stochastic
• - Episodic or Sequential
• - Static or Dynamic
• - Discrete or Continuous
 Agent Function and Programs
• Agent Function: Maps percept sequences to
actions.
• Agent Program: The implementation of the
agent function.
• Relationship: Flow from perception to action
through the agent program.
 PEAS Framework
The PEAS framework is used to define the key components needed to
specify the design of intelligent agents

• Performance Measure: Criteria for success.

• Environment: The context in which the agent
operates.
• Actuators: Mechanisms for action.
• Sensors: Mechanisms for perception.
• Example: Self-driving car’s PEAS description.
 Self-driving car’s PEAS description.

Component Description

Safety (avoiding accidents), reaching destination efficiently, obeying

P - Performance Measure
traffic laws, passenger comfort, fuel efficiency.

Roads, traffic, pedestrians, weather conditions, other vehicles, road

E - Environment
signs, lane markings.

A - Actuators Steering, acceleration, brakes, indicator lights, windshield wipers.

Cameras (object detection), LiDAR (depth sensing), GPS (navigation),

S - Sensors
Radar (detects vehicles), Ultrasonic sensors (close obstacles).
Challenges in Building Intelligent Agents
• - Handling Uncertainty
• - Scaling Complexity
• - Learning from Limited Data
• - Ethical Considerations
• - Real-Time Decision-Making
 Applications of Intelligent Agents
• Examples:
• - Virtual Assistants (e.g., Siri, Alexa)
• - Autonomous Vehicles
• - Personalized Recommendations
• - Smart Home Devices
• - Game AI
 Summary
• Recap of Key Points:
• - Definition and characteristics of intelligent
agents.
• - Types of agents and their environments.
• - PEAS framework for designing agents.
• - Challenges and real-world applications.
• Closing Thought: "AI is not just about
technology; it’s about creating systems that can
perceive, reason, and act intelligently."
 Questions & Discussion
• Questions?