1.
Task Environments and AI Characteristics
Task Observable Agent Deterministic Episodic Static Discrete Instance of
Environment AI
Towers of Fully Single Deterministic Sequential Static Discrete Problem-
Hanoi Observable solving,
Planning
Robot Soccer Partially Multi Stochastic Sequential Dynamic Discrete Robotics,
Player Observable Multi-agent
Systems
Image Fully Single Deterministic Episodic Semi- Continuous Computer
Analysis Observable static Vision, Image
Processing
Interactive Partially Multi Stochastic Sequential Dynamic Continuous Educational
English Tutor Observable Technology,
NLP
Supermarket Fully Single Deterministic Episodic Static Discrete Automation,
Barcode Observable Object
Scanners Recognition
Voice-Activat Partially Single Stochastic Sequential Static Discrete Speech
ed Observable Recognition,
Telephone NLP
Menus
Spelling and Fully Single Deterministic Episodic Static Discrete NLP,Text
Grammar Observable Editing
Correction
(MS Word)
2. What are the agent types with example?
In AI, an agent is an entity that perceives its environment through sensors and acts
upon it using actuators. Agents can be classified into different types based on their
complexity and intelligence. Below are the main agent types with examples:
1. Simple Reflex Agents
Acts only based on the current percept, ignoring history. It follows condition-action
rules (if-then rules).
Example:
● Thermostat: Turns heating on/off based on temperature.
● Supermarket barcode scanner: Reads and identifies items based on scanned codes.
2. Model-Based Reflex Agents
Maintains an internal model of the world to handle partially observable environments.
Example:
● Self-driving cars: Maintain a model of traffic conditions, predicting movement.
● Chess-playing AI: Tracks board positions to decide the best move.
3. Goal-Based Agents
Acts to achieve specific goals, requiring planning and decision-making.
Example:
● GPS Navigation System: Plans the best route to reach a destination.
● Robotics Arm in Industry: Moves based on a goal, such as assembling parts.
4. Utility-Based Agents
Chooses actions based on a utility function, optimizing for the best outcome.
Example:
● Stock Trading AI: Chooses investments maximizing profit.
● Recommendation Systems (Netflix, Amazon): Suggests content/products optimizing
user satisfaction.
5. Learning Agents
Improves performance over time by learning from experience.
Example:
● Chatbots (e.g., ChatGPT): Improves responses based on past interactions.
● Google’s AlphaGo: Learns from playing millions of Go games.
3. PEAS Descriptors for Various Agents
Agent Type Performance Environment Actuators Sensors
Measure
Medical Diagnosis Healthy patient, Patient, Display of Keyboard entry of
System (Learning reduced costs hospital, staff questions, tests, symptoms,
Agent) diagnoses, findings, patient’s
treatments, answers
referrals
Satellite Image Analysis Correct image Downlink from Display of scene Color pixel arrays
System (Model-Based categorization orbiting categorization
Reflex Agent) satellite
Part-Picking Robot Percentage of Conveyor belt Jointed arm and Camera, joint
(Model-Based Reflex parts in correct with parts; hand angle sensors
Agent) bins bins
Refinery Controller Purity, yield, Refinery, Valves, pumps, Temperature,
(Utility-Based Agent) safety operators heaters, pressure,
displays chemical sensors
Interactive English Tutor Student’s score Set of Display of Keyboard entry
(Learning Agent) on test students, exercises,
testing agency suggestions,
corrections
Playing Soccer Goals scored, Soccer field, Legs for kicking, Camera, motion
(Model-Based Reflex game outcome players, arms for sensors, GPS
Agent) referee balance
Playing a Tennis Match Points scored, Tennis court, Arm movement Eye tracking,
(Goal-Based Agent) match outcome opponent, for racket swing motion sensors
referee
Performing a High Jump Height Track and field Legs for Motion sensors,
(Utility-Based Agent) achieved, stadium, jumping, arms bar height sensors
landing stability judges for balance
Chatbot (Simple Reflex User Online chat Text display Text input, speech
Agent) satisfaction, interface, recognition
response quality users
Music Composer AI Musical quality, Digital music Sound Music pattern
(Learning Agent) audience software, generation, note recognition, MIDI
feedback instruments selection input
4. Consider the following scenario and tell whether the agent is rational or not.
A vacuum cleaner (agent) in one the two rooms as shown in the figure. The rooms initially
contain dirt. The vacuum cleaner can do the following actions only: MoveLeft, MoveRight and
Suck. It has no memory.
The vacuum cleaner agent in this scenario is a Simple Reflex Agent because it makes
decisions based only on the current percept (whether there is dirt in its current
location). It does not have memory or a model of the environment.
Rationality of the Agent:
A rational agent is one that takes actions that maximize its performance based on the
given percepts and knowledge. In this case:
If the agent is in a dirty room, it should Suck to clean it.
If the room is clean, it should move Left or Right to check for dirt.
However, since it has no memory, it does not remember past actions, which may lead to
inefficient behavior (such as moving back and forth unnecessarily).
Conclusion:
● The agent is partially rational because it follows a reasonable cleaning strategy for
each percept but lacks memory to optimize its performance.
● If it continuously moves without knowing whether it has already cleaned an area, it
may not always perform optimally.
● A better rational agent would include memory to track which rooms have been
cleaned.
5.What are the key steps involved in transforming a system into an AI system?
To transform a system into an AI system, follow these key steps:
1. Problem Definition
○ Clearly define the problem the AI needs to solve.
○ Identify inputs, outputs, constraints, and success criteria.
2. Data Collection & Preprocessing
○ Gather relevant data for training or decision-making.
○ Clean and preprocess the data (handling missing values, normalization, etc.).
3. Selecting the Right AI Approach
○ Choose between Machine Learning (Supervised, Unsupervised, Reinforcement
Learning) or rule-based AI.
○ Consider deep learning for complex tasks like vision and NLP.
4. Model Development & Training
○ Develop AI models using techniques like neural networks, decision trees, or
reinforcement learning.
○ Train the model on the dataset.
5. Testing & Validation
○ Evaluate the AI system’s performance using metrics like accuracy, precision,
recall, and F1-score.
○ Perform real-world testing with unseen data.
6. Deployment & Optimization
○ Integrate AI into the system and optimize for efficiency.
○ Ensure scalability and reliability.
7. Continuous Learning & Feedback
○ AI should adapt based on new data and improve over time.
○ Implement mechanisms for continuous learning.
6. What are the essential components of a well-defined problem in AI?
A well-defined problem in AI consists of five essential components:
1. Initial State: The starting point of the agent, representing the state of the environment
at the beginning.
2. Actions: The set of possible actions available to the agent in a given state.
3. Transition Model: A function that defines the outcome of applying an action to a state,
determining the next state.
4. Goal Test: A function that checks whether a given state satisfies the goal conditions.
5. Path Cost: A numerical value that evaluates the cost of a sequence of actions, helping
to determine the best solution.
1. Route Finding (Shortest Path Problem)
● Initial State: A starting location (e.g., New York).
● Actions: Moving from one city to another via available roads.
● Transition Model: If you drive from City A to City B, the new state is being in City B.
● Goal Test: Reaching the destination city (e.g., Los Angeles).
● Path Cost: Distance, time, or fuel consumption.
2. Chess (Game Playing AI)
● Initial State: Standard chessboard setup.
● Actions: Legal moves allowed for each piece.
● Transition Model: Applying a move changes the board state.
● Goal Test: Checkmate the opponent’s king.
● Path Cost: Number of moves taken, probability of winning, or board evaluation scores.
3. Puzzle Solving (e.g., 8-Puzzle or Sudoku)
● Initial State: A shuffled board (e.g., a scrambled 8-puzzle or unsolved Sudoku).
● Actions: Moving a tile in the 8-puzzle or placing a number in Sudoku.
● Transition Model: The board configuration changes after each move.
● Goal Test: The board reaches a solved configuration.
● Path Cost: Number of moves made.
4. Medical Diagnosis AI
● Initial State: Patient symptoms and medical history.
● Actions: Ordering medical tests, suggesting treatments.
● Transition Model: Test results refine diagnosis, treatments affect patient health.
● Goal Test: Identifying the correct disease.
● Path Cost: Cost of tests, risk of treatment, time taken for diagnosis.
5. Robot Navigation (Autonomous Vehicles/Robots)
● Initial State: The robot’s starting position in an environment.
● Actions: Moving forward, turning left/right, stopping.
● Transition Model: The robot’s new position after an action.
● Goal Test: Reaching a predefined destination.
● Path Cost: Energy consumption, distance traveled, time taken.
Ln-2 ->Informed and uninformed search technique
Read these.. all for
unit 2
