Back to Subject

Unit 6: Artificial Intelligence and Data Science - Practice Quiz

CAP1007 — Fundamentals Of Information Technology 60 Questions
0 Correct 0 Wrong 60 Left
0/60

1 What does the abbreviation AI stand for?

definition and scope of AI Easy
A. Artificial Intelligence
B. Automated Interface
C. Applied Informatics
D. Advanced Integration

2 Which of the following best describes Artificial Intelligence?

definition and scope of AI Easy
A. A type of computer network cable
B. A method for building physical hardware only
C. Machines performing tasks that normally require human intelligence
D. A tool used exclusively for printing documents

3 The term "Artificial Intelligence" was first coined at a 1956 conference held at which university?

history and evolution of AI Easy
A. Cambridge
B. Dartmouth
C. Stanford
D. Oxford

4 Which test was proposed to check whether a machine can exhibit intelligent behavior similar to a human?

history and evolution of AI Easy
A. Logic Test
B. Turing Test
C. Boolean Test
D. Binary Test

5 Machine Learning is a subset of which broader field?

introduction to machine learning Easy
A. Computer Networking
B. Artificial Intelligence
C. Database Administration
D. Web Design

6 What is the main idea behind Machine Learning?

introduction to machine learning Easy
A. Systems replace all computer hardware
B. Systems learn from data and improve over time
C. Systems only follow fixed rules forever
D. Systems print reports automatically

7 Which type of machine learning uses labeled data for training?

introduction to machine learning Easy
A. Reinforcement learning
B. Reactive learning
C. Unsupervised learning
D. Supervised learning

8 Artificial neural networks are inspired by which biological structure?

neural networks Easy
A. The human brain
B. The digestive system
C. The human heart
D. The skeletal system

9 What is the basic unit of a neural network called?

neural networks Easy
A. Cluster
B. Register
C. Pixel
D. Neuron (node)

10 What does NLP stand for in the context of AI?

natural language processing Easy
A. Natural Language Processing
B. New Logic Programming
C. Network Layer Protocol
D. Numerical Linear Processing

11 Which of the following is a common application of Natural Language Processing?

natural language processing Easy
A. Chatbots and language translation
B. Cooling computer hardware
C. Charging batteries
D. Cutting metal sheets

12 What does Generative AI primarily do?

generative AI Easy
A. Cools down servers
B. Measures network speed
C. Creates new content such as text, images, or audio
D. Deletes files automatically

13 Which of the following is an example of a Generative AI tool?

generative AI Easy
A. Windows Calculator
B. ChatGPT
C. Microsoft Excel
D. WinRAR

14 Data Science mainly focuses on extracting what from data?

Data science: definition and importance Easy
A. Sound waves
B. Physical materials
C. Useful insights and knowledge
D. Electric power

15 Why is Data Science important for organizations?

Data science: definition and importance Easy
A. It helps make better data-driven decisions
B. It replaces office furniture
C. It slows down computer processing
D. It removes the need for electricity

16 Which of the following is one of the well-known "V"s used to describe Big Data?

big data Easy
A. Velocity of light
B. Vision
C. Volume
D. Voltage

17 The term Big Data generally refers to datasets that are:

big data Easy
A. Only stored on paper
B. Made only of images
C. Too large and complex for traditional tools to handle
D. Always smaller than one kilobyte

18 Which of the following is a common method of collecting data?

Data collection methods Easy
A. Formatting a hard disk
B. Surveys and questionnaires
C. Painting walls
D. Watering plants

19 Data collected directly by a researcher for a specific purpose is called:

primary vs. secondary data Easy
A. Secondary data
B. Tertiary data
C. Primary data
D. Backup data

20 Data obtained from already existing sources like books or reports is called:

primary vs. secondary data Easy
A. Live data
B. Secondary data
C. Raw data
D. Primary data

21 A system that plays chess by evaluating millions of board positions but cannot recognize a spoken word is best described as an example of which type of AI?

definition and scope of AI Medium
A. Super AI
B. General (Strong) AI
C. Narrow (Weak) AI
D. Reactive-free AI

22 The period during which AI research funding and interest sharply declined due to unmet expectations is commonly referred to as the:

history and evolution of AI Medium
A. AI Boom
B. AI Winter
C. Turing Era
D. Expert Freeze

23 Which test, proposed in 1950, evaluates a machine's ability to exhibit intelligent behavior indistinguishable from a human?

history and evolution of AI Medium
A. The Boolean Test
B. The Bayes Test
C. The Dartmouth Test
D. The Turing Test

24 A model is trained on emails labeled as "spam" or "not spam" so it can classify new emails. This is an example of:

introduction to machine learning Medium
A. Supervised learning
B. Reinforcement learning
C. Unsupervised learning
D. Transfer clustering

25 A customer segmentation task groups shoppers by purchasing behavior without any predefined labels. Which learning type applies?

introduction to machine learning Medium
A. Reinforcement learning
B. Unsupervised learning
C. Semi-labeled regression
D. Supervised learning

26 A model performs very well on training data but poorly on new, unseen data. This problem is called:

introduction to machine learning Medium
A. Overfitting
B. Regularization
C. Underfitting
D. Normalization

27 In an artificial neural network, what is the primary purpose of an activation function?

neural networks Medium
A. To label the output classes
B. To introduce non-linearity into the model
C. To collect raw input data
D. To store the training dataset

28 A neural network that contains many hidden layers between the input and output layers is commonly called a:

neural networks Medium
A. Deep neural network
B. Linear regressor
C. Decision tree
D. Shallow perceptron

29 Which component of a neuron controls the strength of the connection between two nodes and is adjusted during training?

neural networks Medium
A. Label
B. Token
C. Epoch
D. Weight

30 Breaking the sentence "AI is useful" into the individual units ["AI", "is", "useful"] is an NLP step known as:

natural language processing Medium
A. Normalization
B. Tokenization
C. Vectorization scoring
D. Compilation

31 An application that analyzes product reviews to decide whether each is positive or negative is performing:

natural language processing Medium
A. Sentiment analysis
B. Image segmentation
C. Speech synthesis
D. Data compression

32 Which of the following best distinguishes generative AI from traditional discriminative models?

generative AI Medium
A. It creates new content rather than only classifying existing data
B. It requires no training data at all
C. It cannot process text or images
D. It only sorts data into fixed categories

33 When a language model produces confident but factually incorrect information, this behavior is commonly called:

generative AI Medium
A. Overfitting
B. Tokenization
C. Backpropagation
D. Hallucination

34 Data science is best described as an interdisciplinary field that primarily combines:

Data science: definition and importance Medium
A. Only spreadsheet formatting and typing
B. Statistics, computer science, and domain knowledge
C. Graphic design and video editing
D. Hardware repair and networking cables

35 In the classic "3 Vs" of big data, the term Velocity refers to:

big data Medium
A. The number of different data formats
B. The total size of stored data
C. The speed at which data is generated and processed
D. The accuracy of the data values

36 A dataset containing tweets, images, videos, and sensor logs together demonstrates which characteristic of big data?

big data Medium
A. Veracity
B. Volume
C. Velocity
D. Variety

37 A researcher uses an online questionnaire to gather opinions directly from users for a new study. This is an example of:

Data collection methods Medium
A. Data transformation
B. Data cleaning
C. Primary data collection through a survey
D. Secondary data from a published report

38 Which scenario clearly uses secondary data?

primary vs. secondary data Medium
A. Running your own experiment and recording results
B. Analyzing census figures published by a government agency
C. Observing traffic and manually counting cars
D. Interviewing customers in person for feedback

39 A dataset column for "age" contains the value 250 and several blank cells. Which two data cleaning issues are present?

basics of data cleaning and transformation Medium
A. Overfitting and underfitting
B. Tokenization and stemming
C. Encryption and compression
D. Outliers and missing values

40 Converting all values in a numeric feature to a common to range is a transformation step known as:

basics of data cleaning and transformation Medium
A. Deduplication
B. Tokenization
C. Imputation
D. Normalization

41 A system uses a large lookup table to respond to every possible input with a pre-defined output, showing no capacity to generalize to unseen inputs. Under the standard definition of AI as systems that can learn, reason, and adapt, how should this system be classified?

definition and scope of AI Hard
A. It is an example of artificial general intelligence
B. It is a machine learning system because it maps inputs to outputs
C. It is a strong AI because it can respond to any input correctly
D. It is not genuinely intelligent because it lacks generalization and adaptation, relying purely on stored rules

42 The period roughly from the 1970s to 1980s and again in the late 1980s is often called an 'AI winter.' What was the primary cause of these downturns?

history and evolution of AI Hard
A. The invention of the transistor made all prior AI obsolete
B. A collapse in funding and interest after AI systems failed to meet inflated expectations and hit computational and data limits
C. A total ban on AI research imposed by governments worldwide
D. The discovery that neural networks could never be trained

43 A model achieves 99% accuracy on training data but only 62% on test data. Which combination of description and remedy is most accurate?

introduction to machine learning Hard
A. Overfitting; apply regularization, gather more data, or reduce model complexity
B. Underfitting; increase model complexity substantially
C. High bias; remove regularization entirely
D. Data leakage; the labels are missing from training

44 You are given a dataset with 10,000 samples where only 50 belong to the positive class (fraud). A classifier predicts 'not fraud' for every sample. What accuracy does it achieve, and why is accuracy misleading here?

introduction to machine learning Hard
A. accuracy, but it is misleading because the classifier never detects the minority class
B. accuracy, which reflects perfect fraud detection
C. accuracy, which is acceptable for imbalanced data
D. accuracy, because the classes are balanced

45 In a deep feedforward network using sigmoid activations, gradients become extremely small in early layers during backpropagation, slowing learning. What is this problem and a common mitigation?

neural networks Hard
A. Vanishing gradients; use ReLU activations and normalized initialization to preserve gradient magnitude
B. Overfitting; increase the learning rate
C. Underfitting; remove all activation functions
D. Exploding gradients; add more sigmoid layers

46 Why does a multilayer neural network require non-linear activation functions between layers?

neural networks Hard
A. They are only needed in the output layer to produce probabilities
B. They make the network train faster by reducing the number of weights
C. They guarantee the network will never overfit the training data
D. Without them, a stack of linear layers collapses into a single equivalent linear transformation, unable to model non-linear relationships

47 A traditional bag-of-words model classifies the sentences 'The service was good, not bad' and 'The service was bad, not good' identically. What core NLP limitation does this expose?

natural language processing Hard
A. It requires the sentences to be translated first
B. It cannot handle words longer than five characters
C. It ignores word order and syntactic context, so it cannot capture meaning that depends on sequence
D. It only works on numeric data, not text

48 In word embeddings such as Word2Vec, the relationship can be observed. What property of the embedding space does this demonstrate?

natural language processing Hard
A. Semantic and syntactic relationships are encoded as consistent linear directions in vector space
B. Embeddings only capture spelling similarity, not meaning
C. The model has memorized every sentence in its training corpus
D. Words are stored as unique integer IDs with no relationships

49 A large language model confidently produces a fabricated citation to a non-existent research paper. What is this phenomenon, and what is its root cause?

generative AI Hard
A. Overfitting; the model memorized the wrong citation
B. Data leakage; the citation appeared in the test set
C. Hallucination; the model predicts statistically plausible token sequences without any grounded verification of factual truth
D. Gradient explosion during text generation

50 Which statement best distinguishes discriminative models from generative models?

generative AI Hard
A. Generative models cannot be used for any classification task
B. Discriminative models learn to separate classes, while generative models learn or to produce new samples
C. Discriminative models generate new data, while generative models only classify
D. Both learn identical distributions but differ only in speed

51 A team builds a highly accurate churn-prediction model but never translates it into any business action or intervention. From a data science value perspective, what is the key failure?

Data science: definition and importance Hard
A. The model accuracy was too high to be trusted
B. The team used too little data to train the model
C. Insights were never operationalized into decisions, so the analytics produced no actionable value
D. Data science requires no connection to business outcomes

52 A company streams sensor readings arriving in inconsistent formats at millions of events per second and needs real-time analysis. Which combination of the 'V's of big data is MOST directly stressed?

big data Hard
A. Veracity and value only, since the data is trustworthy
B. Only volume, since the total size is large
C. None of the V's apply to streaming data
D. Velocity and variety, since data arrives fast and in heterogeneous formats

53 A researcher conducts an online voluntary survey and generalizes the results to the entire national population. What is the most serious methodological flaw?

Data collection methods Hard
A. Surveys always produce perfectly random samples
B. Online surveys guarantee causal conclusions
C. Selection bias, because voluntary online respondents are not representative of the whole population
D. The sample is too large to analyze correctly

54 An analyst reuses a government census dataset collected years ago for a new study on current housing trends. Which limitation of secondary data is MOST relevant?

primary vs. secondary data Hard
A. Secondary data is always more expensive than primary data
B. The analyst automatically owns the copyright to the census data
C. Secondary data cannot be used in any research
D. The data may be outdated and was collected for a different purpose, reducing relevance to the current question

55 A startup interviews 200 of its own customers directly to understand product satisfaction. A competitor's published market report is also consulted. Which classification is correct?

primary vs. secondary data Hard
A. Both are secondary data because they were analyzed later
B. Both sources are primary data because both concern the market
C. The interviews are primary data; the market report is secondary data
D. The interviews are secondary data; the report is primary data

56 A dataset has a numeric 'income' column with 15% missing values that are missing not at random (higher earners tend not to report). Simply replacing them with the column mean is risky because:

basics of data cleaning and transformation Hard
A. The mean cannot be computed when values are missing
B. Mean imputation ignores the systematic pattern of missingness and biases the distribution toward the center
C. Mean imputation always increases variance beyond the true value
D. Missing values can never be imputed under any circumstances

57 A feature 'age' ranges from 0 to 100 while 'salary' ranges from 0 to 200,000. A distance-based algorithm like k-NN performs poorly. What transformation directly addresses this?

basics of data cleaning and transformation Hard
A. Feature scaling (normalization or standardization) so both features contribute comparably to distance
B. One-hot encoding both numeric columns
C. Converting salary to a categorical string
D. Removing the age column entirely

58 A retailer wants to know not just what happened and why, but what action to take to maximize revenue next quarter. Which type of analytics does this require?

introduction to data analytics Hard
A. Predictive analytics alone, which only forecasts outcomes
B. Diagnostic analytics, which only explains past causes
C. Prescriptive analytics, which recommends actions using optimization and simulation
D. Descriptive analytics, which only summarizes past data

59 An analyst finds that ice cream sales and drowning incidents are strongly correlated and concludes ice cream causes drownings. What is the flaw, and what likely explains the correlation?

introduction to data analytics Hard
A. Two variables can never be correlated by chance
B. The correlation coefficient was calculated incorrectly
C. Drowning must cause ice cream sales instead
D. Correlation is not causation; a confounding variable (hot weather) drives both

60 In k-fold cross-validation with , which statement is TRUE about how the data is used?

introduction to machine learning Hard
A. The model is trained once on 80% and never validated on the rest
B. All 5 folds are used simultaneously as both training and test data
C. Each sample is used only for training and never for validation
D. The data is split into 5 folds; each fold serves as the validation set once while the other 4 train the model, and results are averaged
Previous MCQ
Unit 5