AI0-001 Practice Test 48 — 15 Questions

Question 1

Which TWO techniques should be considered when optimizing a deep learning model for deployment on edge devices with limited computational resources?

Accepted Answer

Model quantization. Model quantization (B) is correct because it reduces the numerical precision of weights and activations (e.g., from FP32 to INT8), shrinking model size and memory bandwidth while enabling faster integer arithmetic on resource-constrained edge hardware. Knowledge distillation (D) is correct because it trains a smaller 'student' model to mimic a larger 'teacher' model, yielding a compact network with far fewer parameters and FLOPs suitable for edge deployment. Adversarial training (A) is a robustness technique against adversarial examples and does not reduce compute or memory footprint. Using a GPU for inference (C) increases power, cost, and hardware requirements, which is counterproductive on constrained edge devices. Increasing the number of layers (E) enlarges the model and raises computational and memory demands, the opposite of optimization.

Answer

Apply adversarial training

Answer

Use a GPU for inference

Answer

Increase the number of layers

Question 2

Which TWO of the following are key characteristics of unsupervised learning?

Accepted Answer

It uses data without labeled responses. Unsupervised learning algorithms, such as k-means clustering or hierarchical clustering, operate exclusively on input data that has no labeled responses. The model must infer the underlying structure directly from the features without any ground-truth outputs to guide it, which is the defining characteristic of unsupervised learning.

Answer

It predicts a target variable based on input features

Answer

It requires a reward signal to learn optimal actions

Answer

It typically requires a separate validation set for tuning

Question 3

During testing a chatbot, the QA team observes that the bot sometimes responds with harmful content when given adversarial prompts. Which type of testing should be prioritised to catch these edge cases?

Accepted Answer

Red-teaming and adversarial testing. Red-teaming and adversarial testing are specifically designed to probe an AI system for vulnerabilities, including generating harmful or unsafe outputs from adversarial prompts. This approach simulates real-world attacks to uncover edge cases that standard functional tests miss, making it the correct priority for catching harmful content in a chatbot.

Answer

Unit tests for data pipeline functions

Answer

Regression testing on previously fixed bugs

Answer

Integration tests for API connectivity

Question 4

An e-commerce company uses a gradient boosting model to forecast daily sales. Recently, the model's predictions have become less accurate, showing a significant drop in R-squared on validation data. The data scientist checks for data drift but finds no significant changes in feature distributions. The model was trained on data from the past 24 months and is retrained monthly. Upon inspecting the feature importance, the data scientist notices that the top feature 'promotion_flag' has decreased in importance over time. What is the most likely cause of the performance degradation, and what should be done?

Accepted Answer

Concept drift has occurred; retrain the model more frequently with recent data only, or use an online learning approach. The model's performance degradation, despite no data drift in feature distributions, is likely due to concept drift—the relationship between features and the target variable has changed over time. The decreasing importance of 'promotion_flag' suggests that promotions no longer influence sales as they once did. Retraining more frequently with recent data or using online learning can help the model adapt to the new concept.

Answer

The model is overfitting to historical promotions; apply more regularization

Answer

The model's hyperparameters need tuning; perform a grid search

Answer

The promotion_flag feature is leaking future information; remove it

Question 5

A machine learning team is developing a model to predict server failure from telemetry data. They use a deep neural network with 3 hidden layers. After training, the model achieves 99% accuracy on training data but only 85% on validation data. Which technique should the team apply to reduce the generalization error?

Accepted Answer

Apply L2 regularization. The model exhibits high variance (overfitting) because it achieves 99% accuracy on training data but only 85% on validation data. L2 regularization (also known as weight decay) adds a penalty proportional to the squared magnitude of the weights to the loss function, which discourages the network from fitting noise in the training data and improves generalization. This directly reduces the gap between training and validation performance.

Answer

Increase the number of hidden layers

Answer

Increase the learning rate

Answer

Add more training data

Question 6

A developer wants to deploy a scikit-learn model as a REST API endpoint with minimal infrastructure management. Which cloud service is MOST appropriate?

Accepted Answer

Use AWS SageMaker to create a real-time endpoint. AWS SageMaker provides a fully managed service for deploying machine learning models as real-time endpoints with built-in scaling, monitoring, and automatic infrastructure management. It directly supports scikit-learn models via pre-built containers, eliminating the need for custom runtime setup or manual server configuration. This makes it the most appropriate choice for a developer seeking minimal infrastructure management.

Answer

Use AWS Lambda with a custom runtime

Answer

Deploy on an EC2 instance manually

Answer

Use Amazon ECS with manual Docker setup

Question 7

A small business launched a customer support chatbot powered by a pre-trained language model. The chatbot was fine-tuned on a dataset of past support tickets. For the first week, it performed well, accurately answering 85% of queries. After a routine software update that included a new version of the underlying language model library, the chatbot's accuracy dropped to 60% and it began giving nonsensical responses to some questions. The update did not change any code or configuration specific to the chatbot. The business has a backup of the previous environment. What is the MOST appropriate immediate action?

Accepted Answer

Roll back the software update to the previous version of the language model library.. The most appropriate immediate action is to roll back the software update to the previous version of the language model library (Option D). The accuracy drop and nonsensical responses are directly caused by the library update, which likely changed internal model behavior (e.g., tokenization, attention mechanisms, or default hyperparameters) without any code or configuration changes. Restoring the previous environment immediately resolves the issue and allows the business to investigate the library changes in a controlled manner.

Answer

Retrain the chatbot on the original dataset using the new library version.

Answer

Add more intents to the chatbot's configuration to cover the errors.

Answer

Increase the model's temperature parameter to 1.5 to encourage more varied responses.

Question 8

A data scientist is training a deep neural network for sentiment analysis. The training loss decreases steadily but the validation loss starts to increase after 10 epochs. What is the most likely cause and best corrective action?

Accepted Answer

Overfitting; apply dropout and early stopping. The scenario describes a classic case of overfitting: the training loss decreases steadily, indicating the model is learning the training data well, but the validation loss increases after 10 epochs, meaning the model is memorizing noise and patterns specific to the training set rather than generalizing. The best corrective action is to apply dropout (which randomly drops neurons during training to reduce co-adaptation) and early stopping (which halts training when validation performance degrades), both of which are standard regularization techniques for deep neural networks.

Answer

Underfitting; increase model complexity

Answer

Vanishing gradients; use ReLU activation

Answer

Data leakage; shuffle data before splitting

Question 9

A team is deploying a model on Kubernetes using Kubeflow. They want to automatically scale the number of inference pods based on request latency. Which Kubernetes-native feature should they configure?

Accepted Answer

Horizontal Pod Autoscaler (HPA) with custom metrics. The Horizontal Pod Autoscaler (HPA) with custom metrics is the correct choice because it allows scaling based on application-level metrics like request latency, not just CPU or memory. By configuring HPA to use a custom metric (e.g., from Prometheus or a metrics adapter), the team can automatically adjust the number of inference pods to maintain target latency thresholds, which is essential for responsive inference serving.

Answer

Kubeflow Pipelines component

Answer

Cluster Autoscaler

Answer

Vertical Pod Autoscaler (VPA)

Question 10

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

Accepted Answer

Hierarchical clustering. Hierarchical clustering (C) is correct because it is an unsupervised technique that groups unlabeled data by iteratively merging or splitting clusters based on a distance metric, producing a dendrogram without requiring predefined labels. K-means (E) is also correct because it is an unsupervised partitioning algorithm that assigns unlabeled points to k clusters by minimizing within-cluster variance. Linear regression (A) is a supervised method that predicts a continuous target from labeled data, so it is not a clustering technique. Support vector machine (B) is primarily a supervised classifier (and can be used for regression), not an unsupervised clustering method. Decision tree (D) is a supervised model for classification or regression on labeled data, so it does not fit the unlabeled clustering scenario.

Answer

Linear regression

Answer

Support vector machine

Answer

Decision tree

Question 11

A financial institution deploys an AI model for loan approval. To meet regulatory requirements under the EU AI Act for high-risk AI systems, they must ensure human oversight. Which implementation best satisfies the requirement for meaningful human intervention?

Accepted Answer

Provide a human reviewer with the ability to override the model's decision before finalization. The EU AI Act's human oversight requirement for high-risk systems demands that humans can effectively monitor, intervene, and override AI outputs. Providing a human reviewer with the ability to override the model's decision before it is finalized directly implements meaningful human intervention at the decision point, which is the core of Article 14 oversight obligations.

Answer

Audit model decisions quarterly for bias

Answer

Allow users to appeal decisions through a customer service hotline

Answer

Display a confidence score for each decision

Question 12

A data scientist is preparing a dataset for a text classification model. To prevent train/test leakage, which THREE practices should they follow?

Accepted Answer

Use time-based splitting for temporal data. Option B is correct because for temporal data, a time-based split (e.g., training on earlier timestamps and testing on later ones) prevents future information from leaking into the training set, which a random split would allow. Option C is correct because performing the train/test split before any cleaning, normalization, or other preprocessing ensures that statistics and transformations are learned only from the training data and not influenced by the test set. Option E is correct because removing duplicate samples and keeping all text from the same document in a single split prevents identical or near-identical content from appearing in both training and test sets, which would inflate performance estimates. Option A does not belong because shuffling the entire dataset before splitting is not inherently leakage-preventing and can actually cause leakage with temporal or grouped data. Option D does not belong because applying feature scaling to the entire dataset before splitting leaks test-set statistics (mean, variance) into training, which is a classic preprocessing leakage error.

Answer

Shuffle the entire dataset before splitting to ensure randomness

Answer

Apply feature scaling to the entire dataset before splitting

Question 13

Which TWO of the following are common threats to AI model security?

Accepted Answer

Data poisoning. Data poisoning (B) is a core AI-specific threat in which an attacker corrupts the training dataset so the model learns incorrect or malicious behavior, degrading accuracy or implanting backdoors. Adversarial examples (C) are also a fundamental AI model threat, where inputs are deliberately perturbed with small, often imperceptible changes that cause the model to misclassify or produce attacker-chosen outputs. By contrast, SQL injection (A) targets database-driven applications through malicious query strings, not the model itself, and DDoS (D) is an availability attack that floods network or service resources rather than manipulating model behavior. Phishing (E) is a social-engineering attack against users' credentials or trust, not a direct threat to AI model integrity or inference.

Answer

SQL injection

Answer

Distributed denial-of-service (DDoS)

Answer

Phishing attacks

Question 14

Which THREE of the following are key principles of AI ethics as defined by major frameworks?

Accepted Answer

Transparency. Transparency (A) is a core AI-ethics principle because major frameworks require that AI systems' capabilities, limitations, data use, and decision logic be disclosed and explainable to affected stakeholders. Accountability (C) is likewise fundamental, as frameworks such as the OECD AI Principles and UNESCO Recommendation demand that responsibility for AI outcomes be clearly assigned to identifiable humans or organizations, with mechanisms for redress. Fairness (E) is a third key principle, requiring that AI systems avoid unjust bias and discrimination and treat individuals and groups equitably across the lifecycle. By contrast, scalability (B) is an engineering and performance concern about handling growing workloads, and latency (D) is a technical metric of response delay; neither is an ethical principle in AI frameworks.

Answer

Scalability

Answer

Latency

Question 15

Refer to the exhibit. The training log shows losses and accuracies over 5 epochs. What is the most likely problem?

Accepted Answer

Overfitting. The training log shows high training accuracy (e.g., 99%) but low validation accuracy (e.g., 60%) across epochs, with the validation loss increasing after an initial drop. This divergence indicates the model has memorized the training data rather than learning generalizable patterns, which is the hallmark of overfitting.

Answer

Data leakage

Answer

Underfitting

Answer

Vanishing gradient