ARTIFICIAL INTELLIGENCE EXAM – PRACTICE QUESTIONS AND CORRECT ANSWERS
(VERIFIED ANSWERS) PLUS RATIONALES 2026 Q&A | INSTANT DOWNLOAD PDF.
Core Domains
Machine Learning Frameworks and Algorithms
Natural Language Processing and Linguistics
Computer Vision and Pattern Recognition
Neural Network Architectures and Deep Learning
AI Ethics, Governance, and Bias Mitigation
Data Engineering and Lifecycle Management
Reinforcement Learning and Robotics
Statistical Inference and Probability
Introduction
This comprehensive practice assessment is designed to evaluate a candidate's mastery of the
multifaceted field of Artificial Intelligence. The exam encompasses foundational theoretical
principles, advanced algorithmic implementation, and the critical ethical considerations required
in modern professional practice. By utilizing a blend of conceptual inquiries and complex,
scenario-based challenges, the assessment ensures that practitioners possess the necessary
skills for real-world decision-making and technical problem-solving. Each question is structured
to test the candidate’s ability to apply academic knowledge to practical industry dilemmas,
ensuring they are prepared for the rigorous demands of AI deployment, regulatory compliance,
and system optimization.
SECTION ONE: QUESTIONS 1–100
1. Which technique is primarily used to prevent overfitting in a deep neural network by
randomly "dropping out" neurons during the training phase?
A. Batch Normalization
B. Stochastic Gradient Descent
,🟢 C. Dropout Regularization
D. L2 Regularization
🔴 RATIONALE: Dropout is a regularization technique where randomly selected neurons are
ignored during training, forcing the remaining neurons to learn more robust features and reducing
the model's reliance on specific paths.
2. In the context of Natural Language Processing (NLP), what does the "Transformer"
architecture rely on to process sequences in parallel rather than sequentially?
🟢 A. Self-Attention Mechanism
B. Recurrent Hidden States
C. Convolutional Kernels
D. Long Short-Term Memory Cells
🔴 RATIONALE: Unlike RNNs, Transformers use self-attention to weight the importance of
different words in a sequence simultaneously, allowing for massive parallelization and better
handling of long-range dependencies.
3. A company develops a facial recognition system that performs significantly worse on
individuals with darker skin tones. Which ethical concern does this represent?
A. Data Privacy
B. Accountability
🟢 C. Algorithmic Bias
D. Transparency
🔴 RATIONALE: Algorithmic bias occurs when a model produces systemically prejudiced results
due to unrepresentative training data or flawed assumptions in the machine learning process.
4. Which search algorithm expands the shallowest nodes first and is guaranteed to find the
shortest path in an unweighted graph?
,A. Depth-First Search
🟢 B. Breadth-First Search
C. A* Search
D. Greedy Best-First Search
🔴 RATIONALE: Breadth-First Search (BFS) explores all neighbor nodes at the present depth
before moving to nodes at the next depth level, ensuring the first time the goal is reached, it is via
the shortest path.
5. What is the primary purpose of the "Activation Function" in an artificial neural network?
A. To initialize weights to non-zero values
🟢 B. To introduce non-linearity into the model
C. To calculate the error between predicted and actual values
D. To normalize input features to a common scale
🔴 RATIONALE: Without non-linear activation functions (like ReLU or Sigmoid), a multi-layer
neural network would behave like a single-layer linear regression model, regardless of its depth.
6. In Reinforcement Learning, the trade-off between trying new actions and sticking to
known high-reward actions is known as:
🟢 A. Exploration vs. Exploitation
B. Reward vs. Punishment
C. Policy vs. Value Function
D. Markov vs. Non-Markovian
🔴 RATIONALE: Exploration involves searching for new information about the environment,
while exploitation involves using existing knowledge to maximize immediate rewards.
7. Which metric is most appropriate for evaluating a classification model where the cost of a
"False Negative" is extremely high, such as in medical cancer screening?
, A. Accuracy
B. Precision
🟢 C. Recall
D. F1-Score
🔴 RATIONALE: Recall (Sensitivity) measures the ability of a model to find all relevant cases
within a dataset. In medical contexts, missing a positive case (False Negative) is usually more
dangerous than a False Positive.
8. A data scientist is working with a dataset where the features have vastly different ranges
(e.g., age 0-100 and annual income 0-200,000). What step should be taken?
A. Dimensionality Reduction
🟢 B. Feature Scaling
C. Data Augmentation
D. Hyperparameter Tuning
🔴 RATIONALE: Feature scaling (Normalization or Standardization) ensures that features with
larger numerical ranges do not dominate the objective function during the optimization process.
9. Under the European Union's GDPR, what is the "Right to Explanation" in the context of
automated decision-making?
🟢 A. The right for individuals to receive meaningful information about the logic involved in AI
decisions.
B. The requirement for AI models to be open-source.
C. The mandate that all AI decisions must be reviewed by a human.
D. The prohibition of using personal data for AI training.
🔴 RATIONALE: The GDPR provides citizens the right to understand how an automated system
arrived at a decision that significantly affects them, promoting transparency in AI.
10. What type of learning involves training a model on a labeled dataset where the "ground
truth" is provided for every input?
(VERIFIED ANSWERS) PLUS RATIONALES 2026 Q&A | INSTANT DOWNLOAD PDF.
Core Domains
Machine Learning Frameworks and Algorithms
Natural Language Processing and Linguistics
Computer Vision and Pattern Recognition
Neural Network Architectures and Deep Learning
AI Ethics, Governance, and Bias Mitigation
Data Engineering and Lifecycle Management
Reinforcement Learning and Robotics
Statistical Inference and Probability
Introduction
This comprehensive practice assessment is designed to evaluate a candidate's mastery of the
multifaceted field of Artificial Intelligence. The exam encompasses foundational theoretical
principles, advanced algorithmic implementation, and the critical ethical considerations required
in modern professional practice. By utilizing a blend of conceptual inquiries and complex,
scenario-based challenges, the assessment ensures that practitioners possess the necessary
skills for real-world decision-making and technical problem-solving. Each question is structured
to test the candidate’s ability to apply academic knowledge to practical industry dilemmas,
ensuring they are prepared for the rigorous demands of AI deployment, regulatory compliance,
and system optimization.
SECTION ONE: QUESTIONS 1–100
1. Which technique is primarily used to prevent overfitting in a deep neural network by
randomly "dropping out" neurons during the training phase?
A. Batch Normalization
B. Stochastic Gradient Descent
,🟢 C. Dropout Regularization
D. L2 Regularization
🔴 RATIONALE: Dropout is a regularization technique where randomly selected neurons are
ignored during training, forcing the remaining neurons to learn more robust features and reducing
the model's reliance on specific paths.
2. In the context of Natural Language Processing (NLP), what does the "Transformer"
architecture rely on to process sequences in parallel rather than sequentially?
🟢 A. Self-Attention Mechanism
B. Recurrent Hidden States
C. Convolutional Kernels
D. Long Short-Term Memory Cells
🔴 RATIONALE: Unlike RNNs, Transformers use self-attention to weight the importance of
different words in a sequence simultaneously, allowing for massive parallelization and better
handling of long-range dependencies.
3. A company develops a facial recognition system that performs significantly worse on
individuals with darker skin tones. Which ethical concern does this represent?
A. Data Privacy
B. Accountability
🟢 C. Algorithmic Bias
D. Transparency
🔴 RATIONALE: Algorithmic bias occurs when a model produces systemically prejudiced results
due to unrepresentative training data or flawed assumptions in the machine learning process.
4. Which search algorithm expands the shallowest nodes first and is guaranteed to find the
shortest path in an unweighted graph?
,A. Depth-First Search
🟢 B. Breadth-First Search
C. A* Search
D. Greedy Best-First Search
🔴 RATIONALE: Breadth-First Search (BFS) explores all neighbor nodes at the present depth
before moving to nodes at the next depth level, ensuring the first time the goal is reached, it is via
the shortest path.
5. What is the primary purpose of the "Activation Function" in an artificial neural network?
A. To initialize weights to non-zero values
🟢 B. To introduce non-linearity into the model
C. To calculate the error between predicted and actual values
D. To normalize input features to a common scale
🔴 RATIONALE: Without non-linear activation functions (like ReLU or Sigmoid), a multi-layer
neural network would behave like a single-layer linear regression model, regardless of its depth.
6. In Reinforcement Learning, the trade-off between trying new actions and sticking to
known high-reward actions is known as:
🟢 A. Exploration vs. Exploitation
B. Reward vs. Punishment
C. Policy vs. Value Function
D. Markov vs. Non-Markovian
🔴 RATIONALE: Exploration involves searching for new information about the environment,
while exploitation involves using existing knowledge to maximize immediate rewards.
7. Which metric is most appropriate for evaluating a classification model where the cost of a
"False Negative" is extremely high, such as in medical cancer screening?
, A. Accuracy
B. Precision
🟢 C. Recall
D. F1-Score
🔴 RATIONALE: Recall (Sensitivity) measures the ability of a model to find all relevant cases
within a dataset. In medical contexts, missing a positive case (False Negative) is usually more
dangerous than a False Positive.
8. A data scientist is working with a dataset where the features have vastly different ranges
(e.g., age 0-100 and annual income 0-200,000). What step should be taken?
A. Dimensionality Reduction
🟢 B. Feature Scaling
C. Data Augmentation
D. Hyperparameter Tuning
🔴 RATIONALE: Feature scaling (Normalization or Standardization) ensures that features with
larger numerical ranges do not dominate the objective function during the optimization process.
9. Under the European Union's GDPR, what is the "Right to Explanation" in the context of
automated decision-making?
🟢 A. The right for individuals to receive meaningful information about the logic involved in AI
decisions.
B. The requirement for AI models to be open-source.
C. The mandate that all AI decisions must be reviewed by a human.
D. The prohibition of using personal data for AI training.
🔴 RATIONALE: The GDPR provides citizens the right to understand how an automated system
arrived at a decision that significantly affects them, promoting transparency in AI.
10. What type of learning involves training a model on a labeled dataset where the "ground
truth" is provided for every input?
