AI0-001 Practice Test 25 — 15 Questions

Question 1

Refer to the exhibit. What is the recall of the model?

Accepted Answer

0.8. Recall is calculated as True Positives divided by (True Positives + False Negatives). From the confusion matrix, True Positives = 80 and False Negatives = 20, so recall = 80 / (80 + 20) = 80 / 100 = 0.8. This measures the model's ability to correctly identify all actual positive cases.

Answer

0.72

Answer

0.7

Answer

0.73

Question 2

A startup develops an AI recruiting tool that screens resumes. After deployment, they receive a complaint from a candidate who claims the system rejected them due to age discrimination. The startup has no formal AI governance process. They want to quickly assess and remediate the issue. The dataset includes age as a feature. What should they do first?

Accepted Answer

Conduct a bias analysis to measure the model's impact on different age groups. The first step in addressing a potential bias issue is to conduct a bias analysis to measure the model's impact on different age groups. This allows the startup to quantify the extent of any discriminatory behavior before taking remediation steps, ensuring that actions are data-driven and targeted. Without this analysis, any subsequent fix (like removing age) might be premature or ineffective, and could even introduce new biases.

Answer

Apologize to the candidate and offer a manual review of their resume

Answer

Immediately remove age from the feature set and retrain the model

Answer

Ignore the complaint because age is a legitimate business requirement

Question 3

A hospital deploys an AI diagnostic assistant that analyzes medical images. The system has been in use for six months, and radiologists have reported that the AI is increasingly confident in its predictions, but sometimes misses rare conditions. The AI ethics board is concerned about overreliance and potential harm from false negatives. They want to implement a governance framework that ensures appropriate human oversight. The hospital has a limited IT budget. What is the best approach?

Accepted Answer

Implement a human-in-the-loop process where the AI flags low-confidence or rare condition predictions for mandatory radiologist review. A human-in-the-loop process that triggers mandatory radiologist review only for low-confidence or rare-condition predictions directly addresses the risk of overreliance and false negatives without overwhelming the limited IT budget. This targeted oversight ensures that the AI's increasing confidence does not lead to missed rare conditions, while still allowing routine high-confidence predictions to proceed efficiently. The approach balances safety and resource constraints by focusing human attention where the AI is most likely to err.

Answer

Add a warning to the AI interface that says 'This tool may miss rare conditions'

Answer

Require all AI predictions to be reviewed by a radiologist before final diagnosis

Answer

Increase the AI's false positive threshold to reduce missed cases

Question 4

In the AI project lifecycle, which phase involves splitting the dataset into training, validation, and test sets while ensuring no data leakage?

Accepted Answer

Data preparation. Splitting the dataset into training, validation, and test sets is a core data preparation step that must be performed before any model training begins. This phase ensures that data leakage is prevented by keeping the test set completely isolated until final evaluation, which is critical for obtaining an unbiased estimate of model performance. In the AI project lifecycle, data preparation encompasses cleaning, transforming, and partitioning the data, making option A the correct phase.

Answer

Problem definition

Answer

Data acquisition

Answer

Model evaluation

Question 5

A healthcare organization uses an AI model to recommend treatment plans. The model was trained on data from a single hospital, and now treats patients from multiple demographics. Which ethical concern is most critical?

Accepted Answer

Fairness and bias in predictions. The model was trained on data from a single hospital, which likely has a homogeneous demographic profile. When deployed across multiple demographics, the model may produce biased or unfair predictions for underrepresented groups, making fairness and bias the most critical ethical concern. This directly violates the principle of distributive justice in AI ethics.

Answer

Accountability for treatment outcomes

Answer

Lack of transparency in model decisions

Answer

Privacy violations in training data

Question 6

A data scientist is preparing a dataset for a binary classification neural network. The dataset contains both numerical and categorical features, and some rows have identical entries. Which TWO preprocessing steps are most essential to improve model performance and avoid overfitting?

Accepted Answer

Removing duplicate records. Removing duplicate records (A) is essential because identical rows can artificially inflate the importance of certain patterns, leading the model to memorize noise rather than generalize. This directly reduces overfitting by ensuring the training set reflects true data distribution. Scaling numerical features (B) to zero mean and unit variance (standardization) is critical for neural networks as it prevents features with larger magnitudes from dominating gradient updates, enabling faster convergence and stable training.

Answer

Increasing the batch size

Answer

Applying PCA for dimensionality reduction

Answer

Using dropout regularization in the model

Question 7

A financial institution uses an AI model to approve loans. The model uses features including credit score and ZIP code. During an audit, it is discovered that the model has a high false positive rate for loan default predictions in certain ZIP codes. What should the institution do to address this?

Accepted Answer

Retrain the model with fairness constraints. Retraining the model with fairness constraints directly addresses the root cause of the bias—the model's learned correlations between ZIP code and default risk. Fairness constraints, such as demographic parity or equalized odds, are applied during training to ensure the model's predictions are not systematically skewed against certain groups. This approach preserves the predictive power of legitimate features while mitigating discriminatory outcomes, aligning with AI governance principles.

Answer

Remove the ZIP code feature from the model

Answer

Increase the decision threshold for those ZIP codes

Answer

Discontinue use of the model for those ZIP codes

Question 8

An image classification model misclassifies a stop sign as a speed limit sign after a few pixels are altered. What is the most effective defense against such attacks?

Accepted Answer

Adversarial training. Adversarial training is the most effective defense because it explicitly incorporates adversarial examples—like the perturbed stop sign—into the model's training data. By training on both clean and adversarially altered images, the model learns to be robust against small, malicious perturbations that cause misclassification. This directly addresses the root cause of the vulnerability, unlike other options that only mitigate symptoms or ignore the attack vector.

Answer

Use a larger validation dataset

Answer

Reduce the input image resolution

Answer

Increase the model's complexity

Question 9

A data analyst needs to select two appropriate unsupervised learning techniques for clustering unlabeled data. (Choose two.)

Accepted Answer

Hierarchical clustering. Hierarchical clustering is an unsupervised learning technique that groups unlabeled data points into a tree-like structure (dendrogram) based on similarity, without requiring predefined cluster counts. It is appropriate for clustering tasks where the data lacks labels, making it a correct choice for this question.

Answer

Linear regression

Answer

Support vector machine

Answer

Decision tree

Question 10

A company deploys a deep learning model for real-time image classification. After deployment, they notice high inference latency exceeding the 100ms SLA. Which action would most likely reduce latency without significantly impacting accuracy?

Accepted Answer

Apply model quantization. Model quantization reduces the precision of the model's weights and activations (e.g., from 32-bit floating point to 8-bit integer), which significantly decreases memory bandwidth and computational requirements during inference. This directly lowers latency without fundamentally altering the model's learned representations, so accuracy degradation is typically minimal (often <1-2%).

Answer

Add more training data to improve model robustness

Answer

Replace the model with a simpler logistic regression model

Answer

Increase batch size for inference

Question 11

An LLM-based chatbot is being deployed for customer support. The security team wants to prevent the bot from generating toxic or harmful responses. Which defense is MOST appropriate?

Accepted Answer

Output filtering and guardrails. Output filtering and guardrails can block harmful content before it reaches the user. Input validation sanitizes inputs, red teaming identifies vulnerabilities, and rate limiting prevents abuse but not toxic content.

Answer

Input validation and sanitization

Answer

Rate limiting on API requests

Answer

Red teaming the AI system

Question 12

An AI practitioner is fine-tuning a large language model for a domain-specific task using a small labeled dataset (500 examples). They have limited GPU memory. Which technique is MOST suitable?

Accepted Answer

QLoRA (Quantized Low-Rank Adaptation). QLoRA (Quantized Low-Rank Adaptation) is the most suitable technique because it combines 4-bit quantization of the base model with low-rank adapter modules, drastically reducing GPU memory usage while still allowing fine-tuning on a small dataset. This approach preserves the model's pre-trained knowledge and avoids catastrophic forgetting, which is critical when only 500 labeled examples are available.

Answer

Full fine-tuning of all model parameters

Answer

Instruction tuning with the full dataset

Answer

Retrieval-Augmented Generation (RAG) without fine-tuning

Question 13

A security analyst notices that an AI model used for facial recognition is returning unusually high confidence scores for certain individuals while consistently misidentifying others. Which type of attack is most likely occurring?

Accepted Answer

Model inversion attack. A model inversion attack allows an adversary to reconstruct training data or infer sensitive attributes from the model's outputs. In this scenario, the unusually high confidence scores for certain individuals and misidentification of others indicate that the attacker is exploiting the model's internal representations to extract information about the training data, leading to biased or overconfident predictions for specific classes.

Answer

Data poisoning

Answer

Evasion attack

Answer

Model extraction attack

Question 14

A security team discovers that an AI-based anomaly detection system frequently misclassifies benign network traffic as malicious when the source IP is from a specific geographic region. Which type of AI vulnerability is most likely being exploited?

Accepted Answer

Adversarial evasion. The scenario describes an AI-based anomaly detection system that misclassifies benign traffic from a specific geographic region as malicious. This is a classic example of an adversarial evasion attack, where an attacker crafts inputs (in this case, network traffic) that appear benign to human analysts but cause the AI model to misclassify them. The geographic bias suggests the attacker is exploiting the model's learned decision boundary, likely by manipulating features such as source IP or packet timing to evade detection.

Answer

Data poisoning

Answer

Model inversion

Answer

Membership inference

Question 15

Which TWO are characteristics of supervised learning?

Accepted Answer

Requires labeled data. Supervised learning requires labeled data because the model learns a mapping from input features to a known target variable. The correct answer B is fundamental: without labeled examples, the algorithm cannot calculate a loss function to adjust its weights during training.

Answer

Does not require target variable

Answer

Uses reinforcement signals

Answer

Learns to cluster data