51. What is the Minimax algorithm in AI?
- An algorithm that finds the minimum number of moves to win a game
- A decision-making algorithm used in two-player games that maximizes the score for the maximizing player while minimizing the score for the minimizing player, assuming both play optimally
- An optimization algorithm that minimizes the maximum error in a neural network
- A search algorithm that alternates between minimum and maximum depth limits
Answer : B Explanation: The Minimax algorithm is used in adversarial game-playing AI for two-player zero-sum games like chess, checkers, and tic-tac-toe. It builds a game tree and evaluates all possible moves. The MAX player tries to maximize their score; the MIN player tries to minimize it. The algorithm assumes both players play perfectly. Alpha-Beta Pruning (already in your Q27) is an optimization of Minimax that eliminates branches that cannot affect the final decision, dramatically reducing the number of nodes to evaluate. Minimax is the foundation of classical game-playing AI.
52. What is a knowledge graph in AI?
- A flowchart diagram showing how an AI model processes information step by step
- A structured network of real-world entities and the relationships between them, represented as subject-predicate-object triples, used to store and query interconnected knowledge
- A graph that visualizes the accuracy of a machine learning model during training
- A mathematical graph used to optimize the weights of a neural network
Answer : B Explanation: A Knowledge Graph is a semantic network that represents real-world entities (people, places, concepts) and their relationships as interconnected nodes and edges using triples (e.g., “Einstein — bornIn — Ulm”). Knowledge graphs enable AI systems to understand context and relationships between concepts. Google’s Knowledge Graph powers rich search results. Wikidata is a large open knowledge graph. In AI, knowledge graphs are used for question answering, recommendation systems, and enhancing LLMs with structured factual knowledge (as in Retrieval Augmented Generation — RAG).
53. What is Breadth-First Search (BFS) in AI and when is it used?
- A search algorithm that explores the deepest node before exploring siblings
- An uninformed search algorithm that explores all neighbour nodes at the current depth level before moving to nodes at the next depth level, guaranteeing the shortest path in unweighted graphs
- A heuristic search algorithm that uses an evaluation function to guide exploration
- A search method that explores nodes based on their cost from the start node only
Answer : B Explanation: Breadth-First Search (BFS) is an uninformed (blind) search algorithm that explores a graph level by level from the root node. It uses a queue (FIFO) data structure. BFS is complete (always finds a solution if one exists) and optimal for unweighted graphs (finds the shortest path in terms of number of edges). However, it requires significant memory because it stores all nodes at the current level. BFS is used in GPS navigation for finding shortest routes, web crawling, peer-to-peer networks, and social network analysis for finding degrees of separation.
54. What is fuzzy logic in Artificial Intelligence?
- A logic system that produces random outputs to simulate human uncertainty
- A form of logic that handles degrees of truth between 0 and 1 rather than strict binary true/false values, allowing reasoning with imprecise or uncertain information
- A type of machine learning that uses blurry image datasets for training
- An AI technique that deliberately introduces errors to make systems more robust
Answer : B Explanation: Fuzzy Logic, introduced by Lotfi Zadeh in 1965, extends classical binary logic (true/false) to handle degrees of truth. Values range continuously between 0 (completely false) and 1 (completely true) — so “tall” might be 0.8 true for a 6-foot person. This allows AI systems to reason with vague, imprecise, or uncertain information the way humans naturally do. Fuzzy logic is used in washing machines (adjusting cycle based on load), air conditioners (temperature control), automotive systems (anti-lock brakes), and financial forecasting systems.
55. What is a Genetic Algorithm in AI?
- An algorithm that analyzes DNA sequences from biological databases
- A search and optimization algorithm inspired by natural evolution that uses selection, crossover, and mutation to evolve solutions toward an optimal outcome
- A machine learning algorithm that generates new features from existing genetic data
- A type of neural network that mimics gene expression patterns in living organisms
Answer : B Explanation: Genetic Algorithms (GAs) are optimization techniques inspired by Darwinian natural selection and genetics. They work by: (1) Creating a population of candidate solutions (chromosomes), (2) Evaluating fitness of each solution, (3) Selecting the fittest solutions for reproduction, (4) Applying crossover (combining two parents) and mutation (random changes) to create new offspring, (5) Repeating until an optimal solution is found. GAs are used for optimization problems where exhaustive search is impractical — including engineering design, scheduling, neural architecture search, and game strategy optimization.
56. What is an inference engine in AI?
- The hardware component that powers AI computations in a server
- The component of an expert system that applies logical rules to the knowledge base to derive new facts, draw conclusions, and make decisions
- A system that infers missing training data from existing datasets
- A software tool used to measure the response time of an AI model
Answer : B Explanation: The Inference Engine is the “reasoning” component of an expert system or knowledge-based AI. It applies logical rules stored in the knowledge base to deduce new information and arrive at conclusions. There are two main inference strategies: Forward Chaining (data-driven — starts from known facts and works toward a goal), and Backward Chaining (goal-driven — starts from a goal and works backward to find supporting facts). The inference engine is what distinguishes a knowledge-based AI system from a simple database lookup, enabling genuine reasoning capability.
57. What is transfer learning in AI?
- The process of transferring AI models from one server to another for deployment
- A machine learning technique where a model trained on one task is reused as the starting point for a model on a different but related task, saving significant training time and data
- A method for transferring knowledge between human experts and AI systems
- A technique for copying neural network weights between identical model architectures
Answer : B Explanation: Transfer Learning allows a model trained on a large dataset for one task to be fine-tuned for a related task with less data and computation. For example, ResNet trained on ImageNet (1.4M images) can be fine-tuned for medical image classification with only thousands of images. In NLP, pre-trained LLMs like BERT and GPT are fine-tuned for specific tasks (sentiment analysis, Q&A). Transfer learning dramatically reduces the need for large task-specific datasets and training compute, making AI accessible for domains with limited data. It is one of the most practically important techniques in modern AI.
58. What is Bayesian learning in AI?
- A machine learning approach based on trial-and-error with Bayesian reward signals
- A probabilistic approach to learning that uses Bayes’ theorem to update the probability of a hypothesis as new evidence is observed
- An optimization technique that uses Bayesian statistics to reduce neural network size
- A data augmentation method that generates synthetic training examples using Bayes’ theorem
Answer : B Explanation: Bayesian learning applies Bayes’ theorem — P(H|E) = P(E|H) × P(H) / P(E) — to update the probability of a hypothesis (H) given evidence (E). It starts with a prior probability (initial belief), updates it with observed data (likelihood), and produces a posterior probability. Bayesian approaches are used in Naive Bayes classifiers, Bayesian networks (graphical models for probabilistic reasoning), spam filtering, medical diagnosis, and A/B testing. Bayesian methods are powerful when working with uncertainty and limited data, providing not just predictions but confidence levels.
59. What is the Breadth First Search (BFS) vs Depth First Search (DFS) key difference in AI search?
- BFS uses a stack; DFS uses a queue for storing nodes during search
- BFS explores level by level using a queue and is optimal for shortest path; DFS explores as deep as possible using a stack and uses less memory but may not find the optimal path
- DFS is always faster than BFS for all graph types and problem sizes
- BFS and DFS produce identical search results for all problem types
Answer : B Explanation: BFS and DFS are both uninformed (blind) search strategies. BFS uses a Queue (FIFO) — explores all nodes at depth d before any node at depth d+1. It is complete and optimal (finds shallowest solution) but memory-intensive. DFS uses a Stack (LIFO or recursion) — explores as deep as possible before backtracking. It uses much less memory (only stores the current path) but can get lost in infinite branches and is not optimal. Your existing Q6 confirms DFS uses less memory. BFS/DFS comparison is one of the most fundamental and frequently tested AI search topics.
60. What is an Artificial Intelligence Agent in AI?
- A human expert who teaches an AI system how to perform tasks
- An autonomous entity that perceives its environment through sensors and takes actions through actuators to achieve its goals
- A software tool used to monitor the performance of AI applications in production
- A component of a neural network that processes individual data samples
Answer : B Explanation: An AI Agent is any entity that perceives its environment and takes actions to achieve specific goals. The PEAS framework describes agents: Performance measure (how success is evaluated), Environment (what the agent operates in), Actuators (how the agent acts), and Sensors (how the agent perceives). Types of agents include: Simple Reflex Agents (react to current perception), Model-Based Agents (maintain internal world model), Goal-Based Agents (act toward specific goals), Utility-Based Agents (maximize expected utility), and Learning Agents (improve through experience).
