M+ Candidate Selection: Elements of AI Exam

Candidate Information (Mandatory)

Name:

Email ID:

Mobile Number: Enter a valid phone number (e.g., 10-15 digits)

Exam Instructions:

This exam consists of 50 multiple-choice questions based on the Elements of AI course.
You have a total of **60 minutes** to complete the exam once started.
**Scoring:**
- Each correct answer: **+2 marks**
- Each wrong answer: **-0.5 marks** (Negative Marking)
- Unanswered questions: **0 marks**
Your camera will be activated to ensure fair conduct of the exam.
The timer will start immediately after you click "Start Exam".

Time Remaining: 60:00

PROCTORING ACTIVE: Your activity is being monitored via webcam.

Camera status: Initializing...

1. Which of the following best defines Artificial Intelligence (AI)?

A computer that can perfectly mimic human emotion. A machine that can perform tasks that typically require human intelligence. Any computer program written in Python. A database that stores vast amounts of information.

2. What is the name of the test designed to determine if a machine can exhibit intelligent behavior indistinguishable from that of a human?

The Dartmouth Conference The Lovelace Test The Turing Test The Intelligence Quotient (IQ) Test

3. What is **Weak AI** (or Narrow AI)?

AI that can handle any intellectual task a human can. AI designed and trained to perform a specific, limited task. AI that is inherently less powerful than traditional computing. AI that can pass the Turing Test.

4. What subfield of AI focuses on enabling systems to learn from data without being explicitly programmed?

Expert Systems Machine Learning (ML) Symbolic Logic Computational Linguistics

5. In **Supervised Learning**, what must the training data contain?

Only unlabeled data points. Rewards and penalties. Input features paired with the correct output label (ground truth). Data from two different classes only.

6. **Clustering** is a technique primarily associated with which type of machine learning?

Supervised Learning Reinforcement Learning Unsupervised Learning Deep Learning

7. What is the goal of an agent in **Reinforcement Learning**?

To find hidden structures in unlabeled data. To maximize the accumulated reward. To classify data into two distinct categories. To fit a line to a set of data points.

8. Predicting whether a customer **will or will not** click on an ad is an example of a _____ problem.

Regression Clustering Classification Dimensionality Reduction

9. A **Perceptron** is the simplest form of what type of AI structure?

Decision Tree Expert System Neural Network Knowledge Base

10. **Deep Learning** is a subset of Machine Learning that is characterized by:

The use of classical statistical methods. Neural networks with many (multiple) hidden layers. Systems that require very little training data. Algorithms that rely on symbolic logic.

11. In AI search algorithms, what does **state space** refer to?

The total memory used by the algorithm. The set of all possible configurations/situations an agent can be in. The final, goal position only. The number of steps taken.

12. Which search algorithm explores all nodes at the present depth level before moving on to nodes at the next depth level?

Depth-First Search (DFS) Breadth-First Search (BFS) A* Search Hill Climbing

13. A **heuristic function** (h(n)) is used in search algorithms to:

Determine if the goal state is reached. Estimate the cost from the current state to the goal state. Keep track of visited nodes. Calculate the actual cost from the start state.

14. **A* search** is considered an informed search algorithm because it uses:

Random chance. Both the cost so far and a heuristic estimate. Only the cost so far. Only the shortest path.

15. Which of the following is NOT a characteristic of a well-defined AI problem?

Initial state. Set of possible actions. An ambiguous goal test. Path cost function.

16. When training an ML model, the data is typically split into three sets. Which set is used for the final unbiased evaluation?

Training Set Testing Set Validation Set Hyperparameter Set

17. What is **Overfitting** in Machine Learning?

The model performs poorly on both training and test data. The model learns the training data and noise too well, performing poorly on new data. The training data is too small to be useful. The model is too simple to capture the underlying patterns.

18. Which term refers to the inputs used to predict the output in a supervised learning model?

Labels Loss Functions Features Hyperparameters

19. Which metric is most appropriate for evaluating a **Regression** model?

Accuracy Precision Mean Squared Error (MSE) F1-Score

20. In an image classification task, what is the **label**?

The image file size. The category/name of the object in the image (e.g., "cat", "dog"). The time it takes to process the image. The pixel values of the image.

21. What component of a neural network determines the output of a neuron given its input?

The Loss Function The Activation Function The Learning Rate The Backpropagation Algorithm

22. In a feedforward neural network, the flow of information is:

Input Layer → Output Layer → Hidden Layer Input Layer → Hidden Layers → Output Layer Output Layer → Hidden Layers → Input Layer Circular (Recurrent)

23. What does the term **"weights"** represent in a neural network?

The number of layers. The strength of the connection between neurons. The size of the training dataset. The maximum possible score.

24. The process of adjusting the weights and biases in a neural network based on the error is called:

Forward Propagation Feature Scaling Backpropagation Hyperparameter Tuning

25. Convolutional Neural Networks (CNNs) are primarily used for which type of task?

Time-series forecasting Image and video analysis Text generation Reinforcement learning games

26. What is the main concern regarding **algorithmic bias** in AI systems?

It makes the training process too slow. It can lead to unfair or discriminatory outcomes against certain groups. It requires too many neurons in the network. It only happens in unsupervised learning models.

27. The ability of an AI system to be understood by humans, explaining *why* it made a certain decision, is known as:

General AI Data Augmentation Explainable AI (XAI) Quantum Computing

28. AI applications in job automation primarily affect which sector?

Only highly technical research positions. Only manual labor positions. Roles involving repetitive or data-driven tasks (both manual and intellectual). Only roles requiring creativity.

29. Why is **data privacy** a critical ethical consideration for AI?

Because AI models cannot function without private data. Because large-scale AI often relies on collecting and analyzing sensitive personal data. Because private data makes the model overfit. Because it violates all current laws.

30. The concept that AI should be developed and used in a way that respects human values and rights is often referred to as:

Algorithmic Complexity AI Ethics and Governance Expert System Logic Data Splitting

31. Who is considered one of the founding fathers of AI and coined the term "Artificial Intelligence" in 1955?

Alan Turing John McCarthy Marvin Minsky Geoffrey Hinton

32. An **Expert System** is an AI program designed to:

Learn new skills independently. Mimic the decision-making ability of a human expert. Predict stock market prices using regression. Generate creative text and images.

33. The principle that computers can process information to simulate human thought is known as the:

Computational Theory of Mind Law of Thermodynamics Occam's Razor Principle of Least Action

34. Which of these is a common application of **Natural Language Processing (NLP)**?

Object recognition in photos Speech translation Predicting a house price Controlling a robotic arm

35. The **Singularity** is a hypothetical point in time where:

All AI is shut down permanently. AI exceeds human intelligence and begins self-improvement. The first functional computer is built. Neural networks are limited to one layer.

36. What is the primary purpose of **cross-validation** in machine learning?

To speed up the model training time. To ensure the model is evaluated on all parts of the data, reducing reliance on a single split. To convert numerical data to categorical data. To reduce the total number of features.

37. A **Decision Tree** model makes predictions by:

Calculating a weighted sum of inputs. Following a series of conditional (if/then) rules based on feature values. Moving an agent in an environment to maximize reward. Grouping unlabeled data points.

38. If a classification model has **high precision** but **low recall**, what does that mean?

It correctly finds most of the actual positive cases. It makes many false positive predictions. When it predicts positive, it is usually correct, but it misses many actual positive cases. It performs equally well on all types of data.

39. Which algorithm is most commonly used for clustering?

Linear Regression K-Nearest Neighbors (KNN) K-Means Support Vector Machine (SVM)

40. What is a **hyperparameter** in machine learning?

A parameter whose value is learned by the model during training. A parameter whose value is set *before* the training process begins. The final output of the model. The size of the dataset.

41. **Computer Vision** is the field of AI that enables computers to:

Understand and generate human language. Interpret and act on visual information from the world. Solve mathematical optimization problems. Control physical robots in a factory.

42. Which type of neural network is best suited for processing sequential data like text or speech?

Feedforward Networks Convolutional Neural Networks (CNN) Recurrent Neural Networks (RNN) Generative Adversarial Networks (GAN)

43. **Generative AI** models, like those used for image creation or text writing, are primarily focused on:

Classifying existing data. Creating new, original content. Finding optimal paths in a maze. Calculating probability distributions.

44. What is a **chatbot**?

A type of supervised learning algorithm. A computer program designed to simulate conversation with human users. A specific metric for model performance. The dataset used for training an ML model.

45. **AlphaGo**, the program that defeated a human champion in Go, is a famous example of AI using which technique?

Unsupervised Clustering Linear Regression Reinforcement Learning Symbolic Logic

46. What does **GPU** stand for, and why is it essential for Deep Learning?

General Processing Unit; it handles sequential tasks efficiently. Graphics Processing Unit; it is highly efficient at parallel computation needed for training large NNs. Good Performance Unit; it guarantees a high accuracy score. Geographic Positioning Unit; it enables location-based AI.

47. What is the fundamental ethical concern addressed by the **European Union's AI Act**?

Limiting the cost of AI software. Establishing a regulatory framework based on the risk level of AI systems. Mandating the use of Python for all AI development. Forcing companies to share all their training data.

48. Which term describes the phenomenon where an AI system perpetuates or amplifies societal biases present in the training data?

Catastrophic Forgetting Algorithmic Redundancy Data Contamination Social Bias Amplification

49. A **Large Language Model (LLM)**, like the technology behind Gemini or ChatGPT, is primarily trained on:

Proprietary hardware blueprints. Massive amounts of text data. Financial market trading history. Low-resolution image datasets.

50. What is the primary role of an **Activation Function** in a neural network?

To calculate the loss. To introduce non-linearity into the model. To initialize the weights to zero. To connect every neuron to every other neuron.