(SEM VI) THEORY EXAMINATION 2023-24 ARTIFICIAL INTELLIGENCE

ARTIFICIAL INTELLIGENCE – KME062

Section-wise Important Questions & Ready Answers

SECTION A

(Attempt all – 2 marks each)

(a) Agent vs Rational Agent

An agent is an entity that perceives its environment through sensors and acts upon that environment using actuators. A rational agent is one that selects actions that maximize its performance measure based on the percept sequence and available knowledge. Rationality is concerned with doing the right thing, not necessarily with knowing everything.

(b) Artificial Intelligence and Its Importance

Artificial Intelligence is the branch of computer science that focuses on creating systems capable of performing tasks that normally require human intelligence such as learning, reasoning, and decision-making. Modern engineering focuses on AI due to its ability to automate complex processes, improve efficiency, reduce human error, and enable intelligent systems in healthcare, manufacturing, and robotics.

(c) Logic and Semantics in AI

Logic in AI provides formal rules for reasoning and inference, while semantics deals with the meaning of symbols and statements. Together, they help AI systems represent knowledge and draw correct conclusions from given facts.

(d) Propositional Logic and Its Disadvantage

Propositional logic represents knowledge using propositions that are either true or false. Its main disadvantage is limited expressiveness, as it cannot represent relationships between objects or handle quantifiers like “for all” or “there exists”.

(e) Utility Theory and Its Advantages

Utility theory assigns numerical values to outcomes to represent preferences. In AI, it helps agents make rational decisions under uncertainty by choosing actions that maximize expected utility.

(f) Learning with Observation – Advantages and Disadvantages

Learning with observation allows an agent to learn by watching others, reducing the need for trial-and-error. However, it depends heavily on the quality of observed behavior and may not generalize well to new situations.

(g) Perceptron and Its Relation to AI Networks

A perceptron is the simplest neural network model that performs binary classification. It forms the foundation of artificial neural networks by mimicking the functioning of biological neurons.

(h) Function and Generalization Advantage

A function maps inputs to outputs. Generalization allows a learned function to perform well on unseen data, which is more powerful than averaging because it captures underlying patterns rather than just mean values.

(i) Roles of NLP in AI

Natural Language Processing enables machines to understand, interpret, and generate human language. It plays a key role in chatbots, translation systems, sentiment analysis, and voice assistants.

(j) Robotics and Benefits of AI in Robotics

Robotics involves designing machines capable of performing physical tasks. AI enhances robotics by enabling perception, learning, autonomy, and decision-making, leading to smarter and adaptive robots.

SECTION B

(Attempt any three – 10 marks each)

1. Parameters for Selection of an Agent

An agent is selected based on parameters such as performance measure, environment characteristics, percepts, and actuators. For example, a vacuum-cleaning robot agent must consider cleanliness, energy efficiency, and navigation accuracy to act rationally.

2. Syntax and Semantics of First Order Logic (FOL)

Syntax of FOL defines symbols such as constants, variables, predicates, and quantifiers. Semantics assigns meaning to these symbols and determines truth values. FOL is more expressive than propositional logic and is widely used for knowledge representation.

3. Bayesian Network and Decision Making

A Bayesian network is a probabilistic graphical model representing variables and their dependencies. It helps in decision-making under uncertainty by updating probabilities based on evidence, such as diagnosing diseases based on symptoms.

4. Reinforcement Learning and Its Applications

Reinforcement learning enables an agent to learn optimal behavior through rewards and penalties. Unlike inductive learning, it focuses on sequential decision-making. Applications include robotics, game playing, and autonomous systems.

5. Communicating Agents and Their Advantages

Communicating agents share information to coordinate actions efficiently. Intelligent agents contribute to AI by enabling cooperation, adaptability, and distributed problem-solving.

SECTION C

Q3(a) Learning Agent and Its Components

A learning agent improves its performance over time. It consists of a performance element, learning element, critic, and problem generator. These components help the agent learn from experience and adapt to its environment.

Q3(b) Search Algorithms and Heuristic Function in DFS

Search algorithms explore solution spaces to solve problems. Heuristic functions guide the search by estimating the cost to reach the goal. In DFS, heuristics help prioritize promising paths, reducing unnecessary exploration.

Q4(a) Propositional Theorem Proving

Propositional theorem proving determines whether a conclusion logically follows from premises using inference rules such as resolution. It is used in automated reasoning systems.

Q4(b) Knowledge Representation in Logical Agents

Logical agents store knowledge in the form of logical sentences. By reasoning over this knowledge, they can make intelligent decisions, such as diagnosing faults based on known symptoms.

Q5(a) Planning in AI and Its Types

Planning involves generating a sequence of actions to achieve goals. Types include state-space planning, partial-order planning, and hierarchical planning. Forward planning starts from the initial state, while backward planning works from the goal state.

Q5(b) Probability Theory and Bayes’ Rule

Probability theory deals with uncertainty. Bayes’ rule updates probabilities based on new evidence and is widely used in medical diagnosis, where doctors estimate disease likelihood based on symptoms and test results.

Q6(a) Learning, Supervised and Unsupervised Learning

Learning is the process of improving performance through experience. Supervised learning uses labeled data, while unsupervised learning identifies hidden patterns without labeled outputs.

Q6(b) Learning in Neural Networks

Neural networks learn by adjusting weights using algorithms such as backpropagation. For example, a network trained on handwritten digits learns to classify numbers accurately.

Q7(a) Robot and Its Architecture

A robot consists of sensors, controller, actuators, and power supply. AI-based robots can adapt, learn, and operate autonomously, transforming conventional machines into intelligent systems.

Q7(b) Image Processing Features and Vision Comparison

Image processing includes enhancement, segmentation, and feature extraction. Vision helps robots navigate environments and manipulate objects accurately.