Artificial Intelligence (AI) MCQ Questions and Answers

Artificial Intelligence (AI) MCQ Questions and Answers

71. What is the role of AI in healthcare?

  1. AI in healthcare is limited to appointment scheduling and administrative tasks only
  2. AI is used in healthcare for medical image analysis, drug discovery, disease prediction, personalized treatment, clinical decision support, and robotic surgery
  3. AI has not yet been approved for any medical applications and remains in research only
  4. AI in healthcare is used exclusively for managing electronic health records

Answer : B
Explanation: AI is transforming healthcare across multiple dimensions. Medical Imaging: AI detects cancer, diabetic retinopathy, and fractures in X-rays/MRIs with radiologist-level accuracy. Drug Discovery: AI accelerates molecule screening and drug design (DeepMind’s AlphaFold solved the protein folding problem). Disease Prediction: ML models predict patient readmission and sepsis risk. Personalized Medicine: AI tailors treatment plans based on genomic data. Robotic Surgery: AI-assisted robots perform minimally invasive procedures. Natural Language Processing: extracts clinical insights from medical records and literature. AI is increasingly FDA-approved for clinical applications.

72. What is Optical Character Recognition (OCR) in AI?

  1. A computer vision technique that optimizes the color rendering of images on screens
  2. An AI technology that converts different types of documents — such as scanned paper documents, PDFs, or images — into machine-readable and editable text
  3. A network security technique that recognizes and blocks unauthorized optical connections
  4. A deep learning method for generating synthetic characters in animated films

Answer : B
Explanation: OCR (Optical Character Recognition) is an AI technology that converts images of text into machine-readable digital text. Modern OCR uses deep learning (CNNs and LSTMs) to achieve high accuracy even with handwritten text, different fonts, and poor image quality. Applications include: digitizing historical documents, automated data extraction from invoices and receipts, passport and ID scanning at borders, enabling searchable PDFs, reading license plates, and making documents accessible to screen readers. Tesseract (open-source) and Google Vision API are popular OCR tools.

73. What is the concept of AI Ethics?

  1. A set of programming guidelines that AI systems must follow during code execution
  2. The branch of ethics that addresses the moral questions, principles, and responsibilities involved in the development, deployment, and impact of AI systems on society
  3. A legal framework that defines penalties for misuse of AI technology
  4. The study of AI systems that can make ethical decisions in philosophical thought experiments

Answer : B
Explanation: AI Ethics addresses the moral principles and societal responsibilities surrounding AI development and use. Key concerns include: Bias and Fairness (AI systems perpetuating or amplifying existing biases), Transparency and Explainability (understanding how AI makes decisions — XAI), Privacy (data collection and surveillance), Accountability (who is responsible when AI causes harm), Job Displacement (automation replacing human workers), Autonomous Weapons (AI in warfare), and Existential Risk (long-term concerns about superintelligent AI). Organizations like UNESCO, IEEE, and the EU have published AI ethics guidelines to address these challenges.

74. What is Explainable AI (XAI)?

  1. An AI system that verbally explains its capabilities to users before they start interacting
  2. A set of methods and techniques that enable humans to understand, interpret, and trust the decisions and predictions made by AI models — especially complex “black box” models
  3. A type of AI that can only make decisions that are simple enough for humans to verify manually
  4. An AI documentation standard requiring developers to explain their model’s architecture

Answer : B
Explanation: Explainable AI (XAI) makes AI decision-making transparent and interpretable. “Black box” models like deep neural networks achieve high accuracy but are difficult to interpret. XAI methods include: LIME (Local Interpretable Model-agnostic Explanations) — explains individual predictions, SHAP (SHapley Additive exPlanations) — assigns importance scores to each feature, Attention Visualization — shows which parts of input the model focused on, and Saliency Maps — highlights image regions driving a CNN’s decision. XAI is crucial for high-stakes applications in healthcare, finance, and criminal justice where decisions must be justifiable.

75. What is Simulation in the context of AI?

  1. The process of simulating a real AI model using a simplified mathematical approximation
  2. The use of computer-based models to imitate real-world processes or systems, allowing AI to train and test in safe virtual environments before deployment in the real world
  3. A technique for simulating human brain activity using specialized AI hardware
  4. The process of creating simulated training data using generative AI models

Answer : B
Explanation: Simulation allows AI agents to learn and be tested in virtual environments without real-world risks or costs. Reinforcement Learning agents (like game-playing AI and robotic controllers) are trained in simulations before deployment. The “sim-to-real gap” (differences between simulated and real environments) is a key research challenge. Popular simulation environments include OpenAI Gym (now Gymnasium), MuJoCo (physics-based robotics), CARLA (autonomous driving), and Minecraft (general AI research). Simulation is also used in AI safety research to test agent behavior in dangerous scenarios.

76. What is the concept of AI Bias?

  1. The tendency of AI models to favor one computing platform over another
  2. Systematic errors in AI model outputs that result from prejudiced training data or flawed model design, leading to unfair or discriminatory decisions against certain groups
  3. A mathematical term for the constant value added to a neuron’s weighted sum
  4. The preference of an AI system for certain hardware configurations during training

Answer : B
Explanation: AI Bias occurs when AI systems produce systematically unfair outcomes, often reflecting or amplifying biases present in training data. Types include: Data Bias (training data not representative of the real world), Algorithmic Bias (model design choices that disadvantage certain groups), and Confirmation Bias (AI reinforcing existing stereotypes). Real-world examples: facial recognition systems with lower accuracy for darker skin tones, credit scoring algorithms discriminating by zip code, and hiring algorithms biased against women. Addressing AI bias requires diverse training data, fairness metrics, bias audits, and inclusive development teams.

77. What is the difference between AI Planning and Search in AI?

  1. Planning finds the goal state; Search determines the initial state of a problem
  2. AI Search finds a path from initial to goal state by exploring a search space; AI Planning constructs a sequence of actions (a plan) to achieve a goal given knowledge of actions and their effects
  3. Planning is used for simple problems; Search is used only for complex optimization
  4. Both Planning and Search produce identical outputs for all problem types

Answer : B
Explanation: AI Search and Planning are related but distinct. Search explores a state space to find a path to a goal — like BFS, DFS, or A*. Planning is more structured: given an initial state, goal state, and a set of possible actions with preconditions and effects, a planner constructs a sequence of actions that will achieve the goal. The STRIPS planning language (State Transition as Rules for Intelligent Planning System) is a classic formalism. Planning is essential for robotics task execution, dialogue systems, and scheduling. The Planning Domain Definition Language (PDDL) is the standard notation for AI planning problems.

78. What is Anomaly Detection in AI?

  1. A technique for detecting grammatical anomalies in natural language text
  2. An AI technique that identifies data points, events, or patterns that deviate significantly from normal expected behavior — commonly used for fraud detection and network security
  3. A computer vision method for detecting anomalies in camera hardware during image capture
  4. A quality control technique that flags AI models with lower-than-expected accuracy

Answer : B
Explanation: Anomaly Detection identifies unusual patterns that do not conform to expected behavior. Applications include: Fraud Detection (flagging unusual credit card transactions), Cybersecurity (detecting network intrusions and malware), Industrial IoT (identifying equipment failures before they occur), Medical Diagnosis (spotting abnormal readings in patient vitals), and Quality Control (detecting defective products on assembly lines). Methods include statistical approaches (Z-score, IQR), machine learning (Isolation Forest, One-Class SVM), and deep learning (autoencoders trained on normal data that produce high reconstruction error for anomalies).

79. What is the Sigmoid activation function and where is it used?

  1. An activation function that outputs either 0 or 1 with no intermediate values
  2. A smooth S-shaped activation function that maps any input to a value between 0 and 1, used in binary classification output layers and historically in hidden layers
  3. An activation function that outputs the input value unchanged for all positive inputs
  4. A function that generates random sigmoid curves for data augmentation in neural networks

Answer : B
Explanation: The Sigmoid function σ(x) = 1/(1+e⁻ˣ) produces an S-shaped curve that maps any real input to a value between 0 and 1. This makes it ideal for binary classification output layers where the output represents a probability. However, Sigmoid has significant drawbacks for hidden layers: it causes the vanishing gradient problem (gradients become very small for large or small inputs, slowing training), and its outputs are not zero-centered. For these reasons, ReLU and its variants have largely replaced Sigmoid in hidden layers, though Sigmoid remains standard for binary output neurons and in LSTM gates.

80. What is the ReLU activation function and why is it widely used?

  1. A function that randomly activates a fixed percentage of neurons during each training step
  2. Rectified Linear Unit — an activation function that outputs the input directly if positive, and zero otherwise (f(x) = max(0, x)), widely used because it is computationally efficient and helps mitigate the vanishing gradient problem
  3. A function that returns the real-valued output of a complex Euclidean linear computation
  4. An activation function used exclusively in the output layer of regression neural networks

Answer : B
Explanation: ReLU (Rectified Linear Unit) is defined as f(x) = max(0, x) — it outputs x if x > 0, and 0 otherwise. ReLU is the most widely used activation function in deep learning hidden layers because: it is computationally simple, does not suffer from vanishing gradients for positive values, and produces sparse activations (some neurons output 0), making the network more efficient. The main weakness is “dying ReLU” (neurons stuck outputting 0). Variants addressing this include Leaky ReLU (f(x) = max(0.01x, x)), PReLU, and ELU. ReLU is the default choice for most deep learning hidden layers.

Pages: 1 2 3 4 5 6 7 8 9 10