A company is creating a model to label credit card transactions. The company has a large volume of sample transaction data to train the model. Most of the transaction data is unlabeled. The data does not contain confidential information.
The company needs to obtain labeled sample data to fine-tune the model.
Which solutions will meet these requirements? (Choose two.)
A. Run batch inference jobs on the unlabeled data.
B. Run an Amazon SageMaker AI training job that uses the PyTorch Distributed library to label data.
C. Use an Amazon SageMaker Ground Truth labeling job with Amazon Mechanical Turk workers.
D. Use an optical character recognition model trained on labeled samples to label unlabeled samples.
E. Run an Amazon SageMaker AI labeling job.
Show Answer
Correct Answer: C, E
Explanation: The requirement is to obtain labels for a large unlabeled dataset with no confidentiality concerns. Amazon SageMaker Ground Truth can use Amazon Mechanical Turk workers to generate human labels for non-confidential data. Amazon SageMaker labeling jobs are the managed mechanism for creating labeled datasets. Batch inference does not create labels without an already trained model, distributed PyTorch is for training rather than labeling, and OCR is unrelated to credit card transaction labeling.
Question 64
Which statement describes a generative AI use case for multimodal models?
A. Deploy multiple scalable and cost-effective versions of a model.
B. Process large amounts of data to train multiple models.
C. Write code in multiple programming languages.
D. Process different data types, such as images, audio, and video.
Show Answer
Correct Answer: D
Explanation: Multimodal generative AI models are designed to understand and generate across multiple data modalities, such as text, images, audio, and video. Processing different data types is the defining use case of multimodal models. The other options describe scalability, training infrastructure, or code generation rather than multimodal capability.
Question 65
What is continues pre-training?
A. The process of fine-tuning a pre-trained language model on labeled data for a specific task
B. The process of providing unlabeled data to a pre-trained language model to improve the model’s domain knowledge
C. The process of training a language model from the beginning on a specific dataset
D. The process of evaluating the performance of a pre-trained language model on a test set
Show Answer
Correct Answer: B
Explanation: Continued pre-training (also called domain-adaptive or further pre-training) takes an already pre-trained language model and trains it further on additional unlabeled text to adapt or improve its knowledge for a specific domain before any supervised fine-tuning. Fine-tuning on labeled data is a separate step.
Question 66
Which type of ML technique provides the MOST explainability?
A. Linear regression
B. Support vector machines
C. Random cut forest (RCF)
D. Neural network
Show Answer
Correct Answer: A
Explanation: Linear regression is generally the most explainable among the listed models because its coefficients directly quantify the effect of each input feature on the predicted output. Support vector machines, random cut forests, and neural networks are typically less interpretable due to their more complex decision mechanisms.
Question 67
A company wants to use AI for budgeting. The company made one budget manually and one budget by using an AI model. The company compared the budgets to evaluate the performance of the AI model. The AI model budget produced incorrect numbers.
Which option represents the AI model’s problem?
A. Hallucinations
B. Safety
C. Interpretability
D. Cost
Show Answer
Correct Answer: A
Explanation: The AI model generated incorrect numerical results for a budgeting task. Producing plausible-looking but factually incorrect or fabricated outputs is characteristic of AI hallucinations. The issue described is not about safety, interpretability, or cost.
Question 68
A company wants to use foundational models (FMs) to develop and deploy an AI model.
Which AWS service or resource will meet these requirements with the LEAST development effort?
A. Amazon Bedrock
B. Amazon SageMaker AI
C. Amazon Bedrock PartyRock
D. Amazon Q Developer
Show Answer
Correct Answer: A
Explanation: Amazon Bedrock is the managed AWS service for building and deploying generative AI applications using foundation models through an API without managing infrastructure or training models, providing the least development effort. Amazon SageMaker AI is more general-purpose ML development, PartyRock is a no-code prototyping playground rather than a deployment service, and Amazon Q Developer is an AI coding assistant.
Question 69
Which option is a disadvantage of using generative AI models in production systems?
A. Possible high accuracy and reliability
B. Deterministic and consistent behavior
C. Negligible computational resource requirements
D. Hallucinations and inaccuracies
Show Answer
Correct Answer: D
Explanation: A key disadvantage of deploying generative AI in production is that models can hallucinate, producing incorrect or fabricated information with confidence. The other options describe advantages or inaccurate claims: generative AI is not inherently deterministic, does not have negligible computational requirements, and high accuracy/reliability is not a disadvantage.
Question 70
An AI Practitioner is using an LLM-as-a-judge in Amazon Bedrock to evaluate the quality of agent responses in a production environment. The AI practitioner wants to apply a built-in metric that assesses how thoroughly the agent responses address all parts of each prompt or question.
Which metric will meet these requirements?
A. Recall-Oriented Understudy for Gisting Evaluation (ROUGE)
B. Completeness
C. Following instructions
D. Refusal
Show Answer
Correct Answer: B
Explanation: The built-in LLM-as-a-judge metric 'Completeness' evaluates whether a response fully addresses all parts of the user's prompt or question and covers the requested information without omitting important elements. ROUGE is a text-overlap metric, 'Following instructions' focuses on adherence to instructions, and 'Refusal' evaluates appropriate refusal behavior.
Question 71
A financial company stores patterns of fraudulent behavior in a database. The company uses this data to conduct investigations.
The company wants to use a graph-based ML solution to develop an AI tool that helps with these investigations.
Which AWS service will meet these requirements?
A. Amazon OpenSearch Service
B. Amazon Aurora
C. Amazon Neptune
D. Amazon MemoryDB
Show Answer
Correct Answer: C
Explanation: Amazon Neptune is AWS's managed graph database service and supports graph analytics and graph machine learning, making it well suited for fraud detection and investigation use cases where relationships between entities are critical. OpenSearch is a search/analytics engine, Aurora is a relational database, and MemoryDB is an in-memory key-value database.
Question 72
A real estate company is developing an ML model to predict house prices by using sales and marketing data. The company wants to use feature engineering to build a model that makes accurate predictions.
Which approach will meet these requirements?
A. Understand patterns by providing data visualization.
B. Tune the model’s hyperparameters.
C. Create or select relevant features for model training.
D. Collect data from multiple sources.
Show Answer
Correct Answer: C
Explanation: Feature engineering is the process of creating, transforming, and selecting relevant features that improve a machine learning model's predictive performance. Understanding patterns with visualization, hyperparameter tuning, and collecting more data are useful ML activities but are not feature engineering itself.
$19
Get all 443 questions with detailed answers and explanations
Instant download HTML + PDF delivered the moment payment clears.
Secure Stripe checkout we never see or store your card details.
7-day refund if files are defective see our refund policy.