AI Exam Question!

Q1. Describe the Artificial Intelligence with its Application and future Development Plans ?
Artificial Intelligence (AI):
Artificial Intelligence is the simulation of human intelligence in machines programmed to
think, learn, and make decisions. It is a multidisciplinary science involving machine learning,
deep learning, natural language processing, robotics, and computer vision. AI systems are
designed to perform tasks that typically require human intelligence, such as problem-
solving, pattern recognition, and language understanding.
Applications of AI:
1. Healthcare: AI assists in diagnostics, drug discovery, personalized medicine, and
robotic surgeries. Applications like IBM Watson provide support in diagnosing
diseases.
2. Finance: Fraud detection, algorithmic trading, and customer service through
chatbots are major AI applications in the financial sector.
3. Transportation: AI powers autonomous vehicles, traffic management systems, and
ride-sharing platforms.
4. Retail: Personalized shopping experiences, inventory management, and dynamic
pricing models.
5. Education: Adaptive learning platforms and AI tutors offer tailored education
experiences.
6. Entertainment: AI algorithms recommend content, enhance gaming experiences,
and facilitate media editing.
7. Agriculture: Crop monitoring, precision farming, and automated harvesting systems.
Future Development Plans in AI:
• Generative AI: Enhancing creative outputs, such as generating realistic text, images,
or videos.
• General AI: Transitioning from narrow AI to machines capable of general-purpose
reasoning.
• Ethical AI: Ensuring AI systems align with ethical principles and avoid biases.
• AI and Climate Change: AI-driven solutions for energy efficiency and environmental
monitoring.
• Integration with Quantum Computing: AI systems paired with quantum computers
for solving complex problems.
Q2. Elaborate BFS Algorithm with its characteristics ?
Breadth-First Search (BFS):
BFS is an algorithm for traversing or searching tree or graph structures. It starts from a given
node (source) and explores all its neighboring nodes at the current depth before moving on
to the next depth level.
Algorithm Steps:
1. Start at the root node (or any arbitrary node for graphs).
2. Add the starting node to a queue.
3. Dequeue a node from the queue and mark it as visited.
4. Enqueue all unvisited neighbors of the dequeued node.
5. Repeat until the queue is empty or the goal node is found.
Characteristics:
1. Completeness: BFS guarantees finding a solution if one exists.
2. Optimality: BFS is optimal if all edges have the same cost.
3. Space Complexity: O(bd)O(bd), where bb is the branching factor and dd is the depth
of the solution.
4. Time Complexity: O(bd)O(bd) since each node and edge is explored once.

Q3. Elaborate DFS Algorithm its Characterstics ?

Depth-First Search (DFS):
DFS is a graph traversal algorithm that explores as far as possible along each branch before
backtracking. It uses a stack (either explicitly or via recursion).
Algorithm Steps:
1. Start at the root node (or any arbitrary node for graphs).
2. Push the starting node onto a stack.
3. Pop a node from the stack and mark it as visited.
4. Push all unvisited neighbors of the popped node onto the stack.
5. Repeat until the stack is empty or the goal node is found.
Characteristics:
1. Completeness: DFS is not guaranteed to find a solution in infinite-depth spaces.
2. Optimality: DFS is not optimal as it may not find the shortest path.
 3. Space Complexity: O(b⋅m)O(b⋅m), where mm is the maximum depth.
4. Time Complexity: O(bm)O(bm), where mm is the depth of the search tree.

Q4. Concept of heuristic Algorithm and How it is best from other algorithm ?
Heuristic Algorithm:
A heuristic algorithm is a problem-solving technique that uses practical methods or rules of
thumb to produce solutions quickly. It focuses on expediency rather than exhaustiveness,
often trading optimality for speed.
Advantages Over Other Algorithms:
1. Speed: Heuristic algorithms prioritize faster results, ideal for real-time systems.
2. Practicality: Useful for complex problems where exact algorithms are
computationally infeasible.
3. Flexibility: Can be adapted to different problems with heuristic modifications.
4. Resource Efficiency: Requires fewer computational resources compared to
exhaustive methods.
Examples of Heuristic Algorithms:
• A* algorithm in pathfinding.
• Genetic algorithms in optimization problems.

Q5. Describe the A* algorithm with step by step solution for a problem.
A Algorithm:*
The A* algorithm is a pathfinding and graph traversal technique used to find the shortest
path from a start node to a goal node. It uses a heuristic
function f(n)=g(n)+h(n)f(n)=g(n)+h(n), where:
• g(n)g(n): Cost from the start node to nn.
• h(n)h(n): Estimated cost from nn to the goal.
Step-by-Step Example:
Suppose we want to find the shortest path from AA to EE on a graph.
1. Initialization:
• Add the starting node AA to the open list.
• Set g(A)=0g(A)=0 and f(A)=h(A)f(A)=h(A).
2. Expand Node AA:
 • Evaluate all neighbors of AA.
• Update g(n)g(n) and f(n)f(n) for each neighbor.
3. Choose the Next Node:
• Select the node with the lowest f(n)f(n) value from the open list.
4. Repeat:
• Continue expanding nodes and updating costs until the goal node is reached.
5. Backtrack:
• Trace the path from the goal node back to the start node using parent
pointers.

Q6. Elaborate the IDA* algorithm with advantage and disadvantage.

IDA Algorithm:*
Iterative Deepening A* (IDA*) is a graph search algorithm that combines the space efficiency
of Depth-First Search (DFS) with the optimality of the A* algorithm. It performs iterative
deepening using f(n)=g(n)+h(n)f(n)=g(n)+h(n), where g(n)g(n) is the cost to reach the current
node and h(n)h(n) is the heuristic estimate to the goal.
Steps:
1. Perform a DFS with a cost limit based on f(n)f(n).
2. Incrementally increase the cost limit and re-run DFS until the goal is found.
3. Ensure the heuristic function is admissible (it never overestimates the true cost).
Advantages:
1. Space Efficiency: Requires O(bd)O(bd) space, where bb is the branching factor
and dd is the depth of the solution.
2. Optimality: Guarantees the optimal solution if the heuristic is admissible.
3. Combines Strengths: Combines the best features of A* (optimality) and DFS (space
efficiency).
Disadvantages:
1. Time Complexity: It may re-expand nodes multiple times, leading to a higher
runtime.
2. Performance Dependency: Heavily dependent on the quality of the heuristic.
 Q7. Difference between A* and IDA* with points.

Aspect A* IDA*

Space Complexity O(bd)O(bd) O(b⋅d)O(b⋅d)

Time Complexity Lower due to fewer node re-expansions Higher as nodes are re-expanded.

Optimality Optimal with an admissible heuristic Optimal with an admissible heuristic.

Traversal Method Breadth-First Search-like behavior Depth-First Search-like behavior.

Suitable for memory-constrained

Use Case Suitable for memory-rich systems. systems.

Q8. Describe the AO* algorithm with step by step solution for a problem.
AO Algorithm:*
The AO* (And-Or*) algorithm is used for traversing and searching graphs that represent
problems with sub-goals. It handles problems where solutions may involve combining
multiple branches (AND arcs) or choosing between branches (OR arcs).
Steps:
1. Initialization: Start from the root node and mark it as OPEN.
2. Expansion: Expand the most promising node.
3. Selection: Choose AND/OR branches based on the graph structure.
4. Path Evaluation: Calculate the cost of selected paths.
5. Update: Backtrack and update the parent nodes with the new path costs.
6. Repeat: Continue until the solution graph satisfies all goal conditions.
Example: Consider a problem where a task AA requires sub-tasks BB and CC (AND arc), or
sub-task DD (OR arc).
1. Expand AA, identifying the sub-tasks B+CB+C and DD.
2. Calculate the cost of completing B+CB+C versus DD.
3. Choose the lower-cost path.

Q9. Elaborate the alpha-Beta algorithm with advantage and disadvantage.

Alpha-Beta Pruning:
Alpha-Beta is an optimization for the Minimax algorithm in decision-making, reducing the
number of nodes evaluated in a search tree. It maintains two bounds:
 • Alpha: Best value the maximizing player can guarantee.
• Beta: Best value the minimizing player can guarantee.
Steps:
1. Traverse the tree using Minimax.
2. Skip branches that cannot influence the final decision.
3. Update alpha and beta values at each node.
4. Continue until all necessary nodes are evaluated.
Advantages:
1. Efficiency: Reduces the number of nodes explored, improving performance.
2. Same Outcome: Produces the same result as Minimax.
3. Scalability: Enables deeper exploration in time-constrained scenarios.
Disadvantages:
1. Dependency on Order: The efficiency depends on the order of node evaluation.
2. Implementation Complexity: Requires careful tracking of alpha and beta values.

Q10. Explain the different issues in knowledge representation.

Knowledge Representation in AI:
Knowledge representation deals with how to represent information about the world in a
form that a computer system can utilize to solve problems. Major issues include:
1. Representational Adequacy:
• Ability to represent all relevant information.
• Example: Capturing facts, concepts, and relationships.
2. Inferential Adequacy:
• Ability to derive new facts from the represented knowledge.
• Example: Logical inference in predicate calculus.
3. Inferential Efficiency:
• Optimizing the reasoning process to derive conclusions quickly.
• Example: Using indexing or hierarchical structures.
4. Acquisition and Learning:
• The ability to acquire new knowledge and refine existing knowledge.
 • Example: Machine learning integration.
5. Flexibility and Generality:
• The representation should accommodate diverse domains.
• Example: Ontologies and semantic networks.
6. Handling Uncertainty:
• Representing and reasoning with incomplete or uncertain data.
• Example: Probabilistic reasoning systems.

Q11. Explain different approaches of knowledge representation

Knowledge representation (KR) is the method used to encode knowledge about the world
into a format usable by AI systems for reasoning and decision-making. There are several
approaches:
1. Logical Representation:
• Description: Uses formal logic to represent facts, rules, and relationships.
• Example: Predicate logic.
• "All humans are mortal" can be written
as ∀x:Human(x)⇒Mortal(x)∀x:Human(x)⇒Mortal(x).
• Advantages:
• Precise and unambiguous.
• Supports inference through logical reasoning.
• Disadvantages: Computationally expensive for large datasets.
2. Semantic Networks:
• Description: A graph-based representation with nodes (concepts) and edges
(relationships).
• Example: "A dog is an animal" is represented as two nodes (Dog, Animal) with
an "is-a" link.
• Advantages:
• Intuitive visualization.
• Suitable for hierarchical relationships.
• Disadvantages: Limited in representing complex rules or uncertainty.
3. Frames:
 • Description: Knowledge is structured as objects (frames) with attributes
(slots) and values.
• Example: A "car" frame might have slots for "model," "engine," and "color."
• Advantages:
• Supports default reasoning (e.g., a slot can have a default value).
• Modular and reusable.
• Disadvantages: Inefficient for representing dynamic or procedural
knowledge.
4. Production Rules:
• Description: Knowledge is encoded as "IF-THEN" rules.
• Example: "IF it is raining, THEN carry an umbrella."
• Advantages:
• Easy to implement.
• Suitable for decision-making tasks.
• Disadvantages: Can lead to inefficiency with a large rule base.
5. Ontologies:
• Description: Formal frameworks defining concepts, relationships, and rules in
a domain.
• Example: An ontology for biology might define terms like "organism," "cell,"
and "gene."
• Advantages:
• Facilitates sharing and reusing knowledge across systems.
• Disadvantages: Complex to design and maintain.

Q12. Explain the forward and backward reasoning with diagram.

Forward Reasoning:
• Description: Starts with known facts and applies inference rules to derive new facts
until the goal is reached.
• Example: Diagnosing a disease based on symptoms.
• Known facts: Fever, cough.
• Inference: "IF fever AND cough, THEN flu."
Backward Reasoning:
• Description: Starts with a goal and works backward to determine which facts and
rules can support the goal.
• Example: Verifying a hypothesis in diagnostics.
• Goal: Determine if a patient has the flu.
• Traceback: "Does the patient have symptoms like fever and cough?"
Diagram:
Forward Reasoning: Facts → Rules → Derived Facts → Goal Backward Reasoning: Goal →
Rules → Known Facts

Q13. Explain First Order or Predicate Logic with example.

Predicate Logic:
Predicate logic extends propositional logic by dealing with objects, their properties, and
relationships. It includes:
1. Objects: Entities (e.g., John, Car).
2. Predicates: Properties or relationships (e.g., Likes(John, Pizza)).
3. Quantifiers:
• Universal (∀∀): "For all."
• Existential (∃∃): "There exists."
Example:
1. Sentence: "All humans are mortal."
• Representation: ∀x(Human(x)⇒Mortal(x))∀x(Human(x)⇒Mortal(x)).
2. Sentence: "Some humans are doctors."
• Representation: ∃x(Human(x)∧Doctor(x))∃x(Human(x)∧Doctor(x)).

Q14. Represent the following sentence into first order logic.

a. Lipton is a tea.
b. Lata is a child who drink tea.
c. Ruma dislike lata
d. Ruma dislike children who drink tea.
a. Lipton is a tea.
Tea(Lipton)Tea(Lipton)
b. Lata is a child who drinks tea.
Child(Lata)∧Drinks(Lata,Tea)Child(Lata)∧Drinks(Lata,Tea)
c. Ruma dislikes Lata.
Dislikes(Ruma,Lata)Dislikes(Ruma,Lata)
d. Ruma dislikes children who drink tea.
∀x(Child(x)∧Drinks(x,Tea)⇒Dislikes(Ruma,x))∀x(Child(x)∧Drinks(x,Tea)⇒Dislikes(Ruma,x))

Q15. Discuss partial order planning giving suitable example.

Partial Order Planning (POP):
POP is a planning strategy that does not impose a strict sequence on actions. Instead, it
allows actions to remain unordered if they are not dependent on each other, leading to
flexibility in execution.
Steps in Partial Order Planning:
1. Define Goals: Identify the desired outcomes.
2. Select Actions: Choose actions that satisfy specific goals.
3. Order Actions: Impose order only when required (based on dependencies).
4. Resolve Conflicts: Adjust the plan to prevent conflicts between actions.
Example:
Goal: Prepare a sandwich and a cup of tea.
1. Actions:
• A1A1: Get bread.
• A2A2: Get butter.
• A3A3: Spread butter on bread.
• A4A4: Boil water.
• A5A5: Brew tea.
2. Partial Order Plan:
• A1A1 and A2A2 can be performed in any order.
• A3A3 must follow A1A1 and A2A2.
• A4A4 must precede A5A5.
Q16. Explain the hierarchical planning with suitable example.
Hierarchical Planning:
Hierarchical planning decomposes complex tasks into simpler subtasks or levels. The
planning process moves from higher-level abstractions to lower-level details, ensuring
systematic progress.
Steps in Hierarchical Planning:
1. Define the Goal: Outline the primary objective.
2. Decompose into Subtasks: Break down the goal into smaller, manageable tasks.
3. Plan Each Subtask: Develop plans for each subtask individually.
4. Integrate Plans: Combine subtask plans into a cohesive solution.
Example:
Goal: Organize a Birthday Party.
1. Top-Level Task:
• Organize Birthday Party.
2. Subtasks:
• A1A1: Book Venue.
• A2A2: Arrange Food.
• A3A3: Send Invitations.
3. Detailed Subtasks for A2A2:
• A2.1A2.1: Decide Menu.
• A2.2A2.2: Hire Caterer.
This approach ensures efficient task management by focusing on individual components
before integration.

Q17. Describe the various challenges in artificial intelligence planningAnd how we handle
these types of challenges.
Challenges in AI Planning:
1. State Explosion:
• Problem: The number of possible states increases exponentially with problem
size.
• Solution: Use heuristics to prioritize promising states and prune irrelevant
ones.
 2. Uncertainty:
• Problem: The environment may have incomplete or uncertain information.
• Solution: Employ probabilistic planning or decision-theoretic approaches.
3. Dynamic Environments:
• Problem: Changes in the environment can invalidate precomputed plans.
• Solution: Use adaptive or real-time planning techniques.
4. Resource Constraints:
• Problem: Limited time, computation, or physical resources can affect
planning.
• Solution: Optimize plans using resource-aware algorithms.
5. Action Dependencies:
• Problem: Dependencies between actions may complicate sequencing.
• Solution: Use dependency graphs or constraint satisfaction techniques.
Handling Strategies:
• Hierarchical Decomposition: Break down complex plans into simpler subtasks.
• Reinforcement Learning: Adapt plans based on feedback from the environment.
• Machine Learning: Enhance prediction accuracy for uncertain outcomes.

Q18. Explain the various application of artificial intelligence planning.

Applications of AI Planning:
1. Robotics:
• Navigation and task execution in autonomous robots.
• Example: Path planning for delivery robots.
2. Logistics and Supply Chain Management:
• Optimizing routes and schedules.
• Example: AI-driven systems for Amazon’s delivery logistics.
3. Healthcare:
• Personalized treatment planning for patients.
• Example: Scheduling surgeries and resource allocation.
4. Game Development:
 • AI-driven character behavior and game strategy.
• Example: AI opponents in chess and strategy games.
5. Manufacturing:
• Workflow automation and resource optimization in production lines.
• Example: Planning assembly sequences in car manufacturing.
6. Space Exploration:
• Mission planning and autonomous operation for spacecraft.
• Example: NASA's Mars rovers use AI planning for navigation and experiments.

Q19. Draw the planning graph with proper indication of action.

A planning graph consists of alternating levels of states (propositions) and actions. Each level
represents:
• Proposition Level: Possible facts or conditions.
• Action Level: Actions applicable given the current state.
Key Features:
1. Nodes: Represent actions or propositions.
2. Edges: Connect actions to their preconditions and effects.
3. Mutex Links: Indicate mutually exclusive actions or states.
Example: Planning to Bake a Cake
1. Initial State:
• Ingredients and utensils available.
2. Goal State:
• Cake is baked.
Graph Representation:
1. Proposition Level 0: Ingredients are available.
2. Action Level 1: Mix ingredients.
3. Proposition Level 2: Batter is ready.
4. Action Level 3: Bake batter.
5. Proposition Level 4: Cake is baked.
(Note: I can create an image of the graph if you need a visual representation.)
Q20. How is reinforcement learning work Give the suitable example.
Reinforcement Learning (RL):
Reinforcement Learning is a machine learning paradigm where an agent learns to make
decisions by interacting with its environment and receiving feedback in the form of rewards
or penalties.
Key Components:
1. Agent: Learns and takes actions.
2. Environment: The system the agent interacts with.
3. States (SS): The current situation of the agent.
4. Actions (AA): Choices available to the agent.
5. Reward (RR): Feedback indicating success or failure of an action.
Working Steps:
1. Initialization: Define the state space, action space, and reward function.
2. Exploration vs. Exploitation: The agent explores new actions or exploits known
rewarding actions.
3. Update Policy: Improve the strategy (policy) based on rewards.
4. Iteration: Continue interacting with the environment to refine the policy.
Example: Training a Robot to Walk
1. States: Robot's current position.
2. Actions: Move left leg, move right leg, balance, etc.
3. Reward: Positive points for stable walking, negative points for falling.
4. Policy Update: The robot learns which sequences of actions maximize its reward over
time.

Q21. Mention some example of neural network in brief.

Neural Networks:
Neural networks are a class of algorithms inspired by the structure and function of the
human brain. They consist of layers of interconnected nodes (neurons) that process data
to perform tasks like pattern recognition, classification, and prediction.
Examples of Neural Networks:
1. Feedforward Neural Network (FNN):
 • Use Case: Image classification and text recognition.
• Example: Recognizing handwritten digits using the MNIST dataset.
2. Convolutional Neural Network (CNN):
• Use Case: Image and video processing.
• Example: Object detection in self-driving cars.
3. Recurrent Neural Network (RNN):
• Use Case: Sequence prediction and time-series analysis.
• Example: Language translation and stock price prediction.
4. Generative Adversarial Network (GAN):
• Use Case: Image generation and style transfer.
• Example: Generating realistic human faces.
5. Deep Belief Network (DBN):
• Use Case: Dimensionality reduction and feature learning.
• Example: Preprocessing data for downstream tasks.

Q22. Differentiate between reinforcement learning and deep learning with application.

Aspect Reinforcement Learning (RL) Deep Learning (DL)

Learning through rewards and penalties to Extracting features and patterns from large
Purpose make decisions. datasets.

Learning Supervised, unsupervised, or semi-

Paradigm Trial-and-error based. supervised.

Data Interaction with the environment, no need for Requires labeled/unlabeled data for
Dependency labeled data. training.

Representation learning and function

Key Focus Decision-making and policy learning. approximation.

Image recognition, NLP, and speech

Applications Robotics, game playing, autonomous vehicles. processing.

Example Applications:
• RL: Training a robot to navigate a maze.
 • DL: Identifying objects in an image.

Q23. Describe the inductive learning.why you should use inductive learning in artificial
intelligence.
Inductive Learning:
Inductive learning is the process of generalizing from specific examples to form general rules or
patterns. It involves learning a model from observed data and making predictions for unseen cases.

Steps in Inductive Learning:

1. Observe training data.

2. Identify patterns or relationships.

3. Formulate general hypotheses or rules.

4. Test on new data for prediction accuracy.

Example:
Given examples of animals labeled as mammals or non-mammals, an inductive learning model learns
rules (e.g., "If it has fur, it's a mammal").

Why Use Inductive Learning in AI?

1. Generalization Ability: Helps models predict outcomes for new, unseen data.

2. Data-Driven Approach: Automatically discovers patterns without explicit programming.

3. Wide Applicability: Useful in classification, regression, and clustering tasks.

4. Adaptability: Models improve as more data becomes available.

Q24. Write down the weak and strong artificial intelligence example.
1. Weak AI:
Weak AI, also known as Narrow AI, is designed to perform a specific task without
consciousness or general intelligence.
Examples:
1. Siri and Alexa: Virtual assistants that perform tasks like setting reminders or
answering questions.
2. Recommendation Systems: Netflix or Spotify recommending content based
on user preferences.
3. Autonomous Cars: Self-driving vehicles recognizing traffic signs and obstacles.
Strong AI:
Strong AI, or General AI, would possess human-like reasoning and problem-solving abilities,
capable of understanding and learning any intellectual task.
Examples (Hypothetical for now):
4. Human-Level Robots: Machines that can autonomously perform tasks
requiring human cognition, like reasoning or creativity.
5. AI Teachers: Systems capable of adapting teaching styles to individual
learners without prior programming.
(Note: Strong AI has not yet been realized; current examples are theoretical.)

Tutorial Sheet 1

1. What do you understand by Artificial Intelligence?

Artificial Intelligence (AI):

AI refers to the simulation of human intelligence in machines programmed to think,
learn, reason, and perform tasks autonomously. It involves creating systems capable
of decision-making, pattern recognition, problem-solving, and adapting based on
experience.

Key Features:
1. Learning: Machines can learn from data using algorithms (e.g., supervised or
unsupervised learning).
2. Reasoning: The ability to apply logic and rules to arrive at conclusions.
3. Problem-Solving: Formulating solutions for specific challenges.
4. Perception: Interpreting data from the environment, such as images, sound, or
text.
5.
Examples:
• Self-driving cars (autonomous navigation).
• Virtual assistants like Siri or Alexa.
• Fraud detection in financial systems.

2. Give some real-world applications of AI.

1. Healthcare:
• AI in medical imaging for detecting diseases.
• Personalized medicine recommendations.
 2. Finance:
• Fraud detection using machine learning.
• Predictive analytics for stock market trends.
•
3. Retail:
• Dynamic pricing models.
• Personalized shopping experiences with recommendation engines.

4. Transportation:
• Autonomous vehicles.
• Traffic management systems.

5. Education:
• AI tutors for personalized learning.
• Automating administrative tasks.

6. Customer Service:
• AI chatbots for 24/7 customer support.
• Automated ticket resolution systems.

3. How Artificial intelligence, Machine Learning, and Deep Learning

differ from each other?

3. What are the types of AI?

AI can be classified into the following types:
1. Based on Capability:
o Narrow AI (Weak AI): Focuses on a single task.
▪ Example: Siri, spam filtering.
o General AI (Strong AI): Mimics human intelligence across multiple tasks. (Still
theoretical.)
o Super AI: Hypothetical AI surpassing human intelligence.
 2. Based on Functionality:
o Reactive Machines: Perform specific tasks without memory (e.g., Deep Blue
chess computer).
o Limited Memory: Can use past data for decision-making (e.g., self-driving
cars).
o Theory of Mind: Understands emotions and social cues (future
development).
o Self-Aware AI: Possesses self-awareness (hypothetical).

5. What are the different domains/Subsets of AI?

1. Machine Learning (ML): Enables systems to learn from data and improve over time.
• Example: Spam detection in emails.
2. Natural Language Processing (NLP): Helps machines understand and generate human
language.
• Example: Google Translate.
3. Computer Vision: Processes and interprets visual data like images and videos.
• Example: Facial recognition.
4. Robotics: Develops intelligent robots to perform tasks.
• Example: Boston Dynamics’ robots.
5. Expert Systems: Mimics decision-making abilities of a human expert.
• Example: Medical diagnostic tools.
6. Reinforcement Learning (RL): Learns optimal actions through rewards.
• Example: AlphaGo playing Go.

6. What are the types of Machine Learning?

1. Supervised Learning:
o Description: Learns from labeled data to make predictions.
o Example: Predicting house prices.
2. Unsupervised Learning:
o Description: Finds hidden patterns in unlabeled data.
o Example: Customer segmentation in marketing.
3. Reinforcement Learning:
 o Description: Learns to perform actions by maximizing rewards.
o Example: Training robots for navigation.
4. Semi-Supervised Learning:
o Description: Combines labeled and unlabeled data for training.
o Example: Image classification with limited annotations.
o

7. What is Deep Learning, and how is it used in real-world?

Deep Learning:
Deep learning is a subset of machine learning that uses artificial neural networks to simulate
the way humans learn. These networks are capable of handling vast amounts of data with
multiple layers for feature extraction and decision-making.
Applications in the Real World:
1. Image Recognition: Detecting objects in photos (e.g., Google Photos).
2. Speech Recognition: Converting speech to text (e.g., Siri, Alexa).
3. Healthcare: Diagnosing diseases from medical scans.
4. Autonomous Vehicles: Identifying objects and making driving decisions.
5. Fraud Detection: Identifying suspicious activities in financial transactions.

8. Which programming language is used for AI?

Several programming languages are used for AI development, including:
1. Python:
o Why? Popular for its simplicity and rich libraries (e.g., TensorFlow, PyTorch,
Scikit-learn).
o Applications: ML, NLP, and computer vision.
2. R:
o Why? Excellent for statistical analysis and data visualization.
o Applications: Predictive modeling and data analysis.
3. Java:
o Why? Scalable and platform-independent.
o Applications: AI-powered mobile and enterprise applications.
 4. C++:
o Why? High performance and control over system resources.
o Applications: AI in robotics and game development.
5. Prolog:
o Why? Tailored for logic programming and knowledge representation.
o Applications: Expert systems and natural language processing.

Tutorial Sheet 2

9. What is the intelligent agent in AI, and where are they used?
Intelligent Agent:
An intelligent agent is an entity in AI that perceives its environment through sensors and acts
upon it using actuators to achieve specific goals. It makes decisions based on observations
and a built-in model.
Characteristics:
1. Autonomy: Operates independently.
2. Reactive: Responds to environmental changes.
3. Goal-Oriented: Acts to achieve defined objectives.
4. Learning Ability: Improves performance over time.
Types of Intelligent Agents:
1. Simple Reflex Agents: Respond directly to percepts.
2. Model-Based Agents: Maintain an internal state to handle partially observable
environments.
3. Goal-Based Agents: Act to achieve specific goals.
4. Utility-Based Agents: Maximize overall satisfaction (utility).
Applications:
• Autonomous Vehicles: Perceive roads, obstacles, and traffic to navigate safely.
• Virtual Assistants: Siri, Alexa, and Google Assistant.
• Robotics: Industrial robots performing tasks like assembly and welding.
• Healthcare: AI systems diagnosing diseases or recommending treatments.
10. How is machine learning related to AI?
Relationship Between AI and ML:
• Machine Learning (ML) is a subset of Artificial Intelligence (AI). AI encompasses the
broader goal of creating machines that simulate human intelligence, while ML
focuses on enabling machines to learn from data.
• ML is one of the primary techniques used to build AI systems.
Key Concepts of ML in AI:
1. Training Data: AI systems use ML algorithms to identify patterns in data.
2. Predictive Models: ML allows AI to make predictions or decisions without explicit
programming.
3. Adaptability: AI systems adapt to new data through ML techniques like supervised,
unsupervised, and reinforcement learning.
Example:
An AI chatbot uses ML to analyze user queries and respond appropriately.

11. What is the use of computer vision in AI?

Computer Vision in AI:
Computer Vision is a field of AI that enables machines to interpret and understand visual
information from the world, such as images or videos.
Key Functions:
1. Image Recognition: Identifying objects, people, or scenes in an image.
2. Object Detection: Locating objects within images or video frames.
3. Facial Recognition: Identifying or verifying individuals using facial features.
4. Video Analysis: Tracking and analyzing motion or activities in videos.
Applications:
1. Healthcare: Diagnosing diseases from X-rays and MRIs.
2. Retail: Automated checkout systems like Amazon Go.
3. Security: Monitoring through surveillance cameras with real-time alerts.
4. Autonomous Vehicles: Recognizing pedestrians, traffic lights, and road signs.

12. Explain the minimax algorithm along with the different terms.
Minimax Algorithm:
Minimax is a decision-making algorithm used in game theory and AI for finding the optimal
strategy in two-player games, such as chess or tic-tac-toe.
Key Terms:
1. Maximizer: Aims to maximize the score, representing one player.
2. Minimizer: Aims to minimize the score, representing the opponent.
3. Utility Function: Measures the desirability of a game state.
4. Game Tree: A tree-like representation of all possible moves in a game.
Steps:
1. Construct the game tree.
2. Assign utility values to terminal states.
3. Propagate values back up the tree:
o Maximizer chooses the maximum value.
o Minimizer chooses the minimum value.
Example:
• In tic-tac-toe, Minimax evaluates all possible moves for both players and selects the
optimal one.

13. What is game theory? How is it important in AI?

Game Theory:
Game theory is the study of strategic interactions between rational decision-makers. It
analyzes how participants make decisions to maximize their payoffs.
Key Concepts in Game Theory:
1. Players: Decision-makers in the game.
2. Strategies: Plans of action available to players.
3. Payoff: Rewards or outcomes based on chosen strategies.
4. Equilibrium: A state where no player can improve their payoff unilaterally.
Importance in AI:
1. Multi-Agent Systems: Helps design strategies in systems involving multiple agents,
such as auctions or negotiations.
2. Reinforcement Learning: Used in competitive environments like games or
autonomous driving.
 3. Economic Modeling: AI applications in predicting market behaviors.
4. Security: Planning and executing countermeasures in cybersecurity.

14. What are some misconceptions about AI?

Common Misconceptions:
1. AI Can Think Like Humans:
o Reality: AI mimics specific tasks and lacks human-like consciousness or
emotions.
2. AI Will Replace All Jobs:
o Reality: AI automates repetitive tasks but also creates new opportunities.
3. AI Is Always Correct:
o Reality: AI systems can be biased or inaccurate if trained on flawed data.
4. AI Is the Same as Machine Learning:
o Reality: Machine learning is a subset of AI, focusing on data-driven learning.
5. AI Is Autonomous and Needs No Supervision:
o Reality: Most AI systems require human oversight, especially in critical
applications.

15. What are the different software platforms for AI development?

Popular AI Development Platforms:
1. TensorFlow:
o Open-source library for deep learning and ML.
o Supported by Google.
o Used for NLP, computer vision, and predictive analytics.
2. PyTorch:
o Python-based framework for deep learning.
o Developed by Facebook.
o Preferred for research and rapid prototyping.
3. Scikit-learn:
o Python library for ML algorithms.
 o Used for regression, classification, and clustering.
4. Keras:
o High-level neural network API built on TensorFlow.
o Simplifies building and training deep learning models.
5. Microsoft Azure AI:
o Cloud-based tools and services for AI and ML.
o Offers pre-trained models and infrastructure for training custom models.
6. IBM Watson:
o Suite of AI tools for business applications.
o Specializes in NLP and conversational AI.

16. How can AI be used in fraud detection?

AI in Fraud Detection: AI systems detect fraudulent activities by analyzing large volumes of
transaction data, identifying patterns, and flagging anomalies.
Steps in AI-Driven Fraud Detection:
1. Data Collection: Gather transaction data from various sources.
2. Feature Extraction: Identify relevant indicators (e.g., transaction time, location).
3. Model Training: Train ML models to differentiate between normal and fraudulent
behavior.
4. Real-Time Monitoring: Use trained models to monitor transactions continuously.
5. Alert Generation: Flag suspicious activities for review.
Applications:
1. Banking: Detecting credit card fraud or identity theft.
2. E-commerce: Identifying fake reviews or suspicious accounts.
3. Insurance: Flagging fraudulent claims.