AWS Certified AI Practitioner Practice Exam

What is an availability zone (AZ)?

1 pts

volume_mute

A physical facility that houses numerous AWS servers

A virtual division of AWS infrastructure designed to ensure fault tolerance

A firewall-like security group that controls access to AWS resources

A geographical region that covers more than one continent

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which cloud service model provides full applications to users without requiring them to manage infrastructure?

1 pts

volume_mute

Infrastructure as a service (IaaS)

Platform as a service (PaaS)

Software as a service (SaaS)

Virtualization as a service (VaaS)

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a defining characteristic of the hybrid cloud model?

1 pts

volume_mute

It exclusively uses private data centers for all workloads

It combines public and private cloud environments

It relies only on on-premises infrastructure

It uses a multitenant model for security

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key advantage of AWS regions?

1 pts

volume_mute

They eliminate the need for availability zones (AZs)

They allow customers to comply with data residency requirements

They are only available in Asia

The provide unlimited computing power without redundancy

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which AWS service allows developers to deploy applications without managing servers?

1 pts

volume_mute

Amazon EC2

AWS Lambda

Amazon RDS

Amazon CloudFront

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which of the following best describes Amazon S3?

1 pts

volume_mute

A relational database service for structured data storage

A scalable object storage service designed for durability and availability

A high performance compute service for running applications

A content delivery network (CDN) for accelerating web traffic

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the main benefit of using a confusion matrix to evaluate a machine learning (ML) model?

1 pts

volume_mute

It shows the training time of the model

It helps analyze false positives and false negatives

It calculates the total number of data points used in training

It eliminates the need for additional performance metrics

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

A bank wants to detect fraudulent transactions in real time. What type of inference should they use?

1 pts

volume_mute

Batch inference

Asynchronous inference

Real-time inference

On-demand inference

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a major challenge when using high-dimensional datasets in machine learning (ML)?

1 pts

volume_mute

It reduces the need for model tuning

It increases computational costs and complexity

It makes models interpret data more efficiently

It ensures better accuracy for all ML tasks

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

An organization notices that its machine learning (ML) model is showing less accurate results over time due to changing customer behavior. What issue is this?

1 pts

volume_mute

Model overfitting

Hyperparameter tuning issue

Feature engineering error

Drift

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the goal for using Amazon SageMaker Model Monitor?

1 pts

volume_mute

To train deep-learning models faster

To automatically deploy machine learning (ML) models

To detect issues such as data drift and concept drift

To fine-tune pretrained models

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

A business wants to improve efficiency of their data processing pipeline by automating feature extraction and transformation. Which AWS tool is best suited for this?

1 pts

volume_mute

Amazon Rekognition

Amazon Textract

Amazon SageMake Data Wangler

AWS Glue

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

A software company wants to analyze customer feedback to determine wheter reviews are positive, neutral, or negative. Which AWS service should they use?

1 pts

volume_mute

Amazon Textract

Amazon Comprehend

AWS Lambda

Amazon SageMaker Feature Store

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the primary purpose of hyperparameter tuning in machine learning (ML)?

1 pts

volume_mute

To create new training data

To adjust model parameters for improved performance

To convert categorical data into numerical form

To speed up the training process

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

How do diffusion models work?

1 pts

volume_mute

They add and remove noise in various steps

They use competing neural networks

They only create text

They only create sound

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the primary purpose of fine-tuning a foundation model (FM)?

1 pts

volume_mute

To make a model more general-purpose

To customize the model for a specific domain

To lower the latency

To train the model from scratch

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

How does an encoder work in a transformer model?

1 pts

volume_mute

It creates synthetic data

It fine-tunes hyperparameters

It evaluates model predictions

It processes input sequences and extracts meaningful representations

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which of the following describes a limit on how much a large language model (LLM) can process at a time?

1 pts

volume_mute

Overfitting

Bias

Context windows

The discriminator

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What's the reason for using reinforcement learning from human feedback (RLHF) in generative AI models?

1 pts

volume_mute

To improve the alignment the model's responses with human preferences

To lower training costs

To replace deep learning models

To reduce the latency of the model

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key benefit of a multimodal foundation model (FM)?

1 pts

volume_mute

It requires much less data for the training

It can process and create different types of content like text, images, and videos

it is more explainable than text-based models

It does not require GPUs

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Of the choices below, which is a main reason why large language models consume significant compute resources?

1 pts

volume_mute

They use retrieval-augmented generation (RAG)

They use complex linear algebra

They process billions of parameters to generate accurate responses

They rely on human feedback and evaluation for all responses

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Why can RAG help lower hallucinations in AI models?

1 pts

volume_mute

It has guardrails for certain types of prompts

It searches a proprietary database to enhance the response

It disables the probability system

It relies only on human supervision

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Amon these choices, what is a core capability of Amazon Lex?

1 pts

volume_mute

It analyzes data for fraud detection

It allows for video analysis

It supports intents and slot filling to create conversational AI experiences

It automates the extraction of text from scanned documents

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What can you do with Amazon Comprehend?

1 pts

volume_mute

Analyze text to extract insights such as for sentiment, key phrases, and entities

Transcribe audio recordings into text

Convert text into humanlike speech

Translate text into multiple languages

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a major capability of Amazon Rekognition?

1 pts

volume_mute

The translation of spoken language into text

The use of optical character recognition

The generation of AI-powered chatbot responses

The ability to recognize faces, objects, and scenes in images and videos

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Among these options, which is a technique used in natural language processing (NLP) preprocessing?

1 pts

volume_mute

Automatically correcting all grammatical errors in a sentence

Translating text into another language before processing

Lemmatization, which reduces words to their root form

Applying deepfake detection to verify text authenticity

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Why is stopword removal used in natural language processing (NLP) preprocessing?

1 pts

volume_mute

Stopwords add important meaning to a text

It helps remove uncommon words to focus only on frequently used terms

Stopwords are removed to reduce text length to speed up processing

Removing stopwords like the and is helps improve efficiency without losing essential meaning

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

When is AI not always the best solution for a business use case?

1 pts

volume_mute

AI models do not work well in cloud-based environments

AI requiers human oversight at all times

AI can only be used for simple automation tasks

AI solutions can be complex and costly when a simpler approach may suffice

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the Compare Models feature in the Bedrock playground?

1 pts

volume_mute

It allows for using image and text responses

It provides for making longer responses for two different models

It combines responses from two models

It allows for evaluating the side-by-side responses from two different models

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Why would you fine-tune a model in Bedrock?

1 pts

volume_mute

To reduce compute costs

To improve latency

To improve response accuracy for domain specific scenarios

To use images

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key advantage of using open source foundation models (FMs) in AWS Bedrock?

1 pts

volume_mute

They are always free to use

They support real-time inference by default

They offer transparency, customization, and community innovation

They provide the codebase and dataset

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What's a benefit of using a distilled model in Bedrock?

1 pts

volume_mute

It reduces compute requirements, making it suitable for edge devices

It increases the context window

It increases the creativity of the responses

It guarantees higher accuracy than larger models.

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key benefit of multiagent collaboration in Bedrock?

1 pts

volume_mute

It relies on distilled models

It provides for 100% accuracy

It allows different agents to specialize in specific tasks, leading to better problem solving

It enables users to edit model weights and biases

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key benefit of batch processing when using Bedrock?

1 pts

volume_mute

There is low latency

The context window is unlimited

There is higher accuracy

Lower costs

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which technique improves the consistency and clarity of prompts when used repeatedly?

1 pts

volume_mute

Few-shot prompting

Zero-shot prompting

Chain-of-thought prompting

Prompt templates

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the purpose of chain-of-thought (CoT) prompting?

1 pts

volume_mute

To specify for the large language model (LLM) which examples to use

To limit the response of the large language model (LLM)

To allow for step-by-step reasoning for complex problems

To adjust the large language model (LLM) parameters

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which type of prompting involves providing several examples to help the model learn a pattern?

1 pts

volume_mute

Zero-shot prompting

Chain-of-thought (CoT) prompting

Few-shot prompting

Template prompting

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is model poisoning?

1 pts

volume_mute

This allows a model to process personal health data

This is when biased or malicious data is injected into the training process

This when a model lacks proper licensing

This is where there are conflicting instructions in a single prompt

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is exposure as it relates to security risks with foundation models (FMs)?

1 pts

volume_mute

The use of the model for consumer apps

The loss of GPU performance during training

The unintentional inclusion of sensitive data in training sets

The failure to generate output within token limits

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is jailbreaking of a foundation model (FM)?

1 pts

volume_mute

Running a model without GPU acceleration

Resetting a model's API token for expanding the access

Deleting a model's training history

Tricking the model into bypassing safety and ethical restrictions

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which of the following best describes AI governance?

1 pts

volume_mute

A set of marketing strategies for AI adoption

Policies and oversight structures to guide ethical AI development

Legal ownership of AI-generated content

Techniques for building large language models (LLMs)

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What does controllability in AI focus on?

1 pts

volume_mute

Reducing computational costs

Building faster AI systems

Aligning AI actions with human intent and oversight

Removing human involvement

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which of the following is a business advantage of adopting responsible AI practices?

1 pts

volume_mute

Enhanced trust and improved brand reputation

Reduced need for data collection

Increased dependence on human intervention

Elimination of all algorithm errors

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the role of Amazon SageMaker Clarify in responsible AI?

1 pts

volume_mute

Detecting harmful URLs in user prompts

Managing compute costs during training

Identifying and explaining bias in data and models

Encrypting training data

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is Amazon Augmented AI (A2I) primarily used for?

1 pts

volume_mute

Accelerating training of neural networks

Creating synthetic data from scratch

Managing GPUs in SageMaker

Integrating human reviews into AI workflows

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key advantage of reinforcement learning from human feedback (RLHF)?

1 pts

volume_mute

It removes the need for model retraining

It ensures AI systems remain static

It helps models align outputs with human preferences and judgements

It eliminates the need for data labeling

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a compliance risk caused by AI systems developing emergent capabilities?

1 pts

volume_mute

They strictly follow preprogrammed features with no surprises

They may behave unpredictably in ways not planned by the original designers

They automatically file compliance reports as they change

They prevent the need for any human oversight

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is a key characteristic of a regulated workload?

1 pts

volume_mute

It only applies to gaming or entertainment systems

It only focuses on making systems faster, not safer

It only requires encryption of stored data

It must follow specific compliance rules due to legal, industry, or safety concerns

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Why is data logging important in AI systems?

1 pts

volume_mute

it helps speed up model training by skipping error checks

It captures inputs, outputs, and system events, supporting debugging, monitoring, and transparency

It prevents the need for data residency compliance

It automatically guarantees 100% model accuracy

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which AWS service provides a detailed history of resource configuration changes and relationships to support compliance auditing?

1 pts

volume_mute

AWS Config

AWS Trusted Advisor

Amazon Inspector

AWS Artifact

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

In the Generative AI Security Scoping Matrix, which scope involves using publicly available generative AI tools without backend access or customization?

1 pts

volume_mute

Scope 3: Pretrained models

Scope 4: Fine-tuned models

Scope 5: Self-trained models

Scope 1: Consumer applications

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

Which AWS service uses machine learning (ML) to detect and classify sensitive data like personally identifiable information (PII) or protected health information (PHI) across your environment?

1 pts

volume_mute

AWS Verified Permissions

AWS Shield Advanced

Amazon Macie

Amazon SageMaker Role Manager

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

What is the purpose of a model card in the context of generative AI?

1 pts

volume_mute

To automatically retrain models with new data

To manage access controls across cloud infrastructure

To document the data sources, intended uses, risks, and limitations of a model

Top optimize compute resources for faster model deployment

note_alt Add notes

flag Flag

volume_mute

Correct Answer

Explanation

In a generative AI systems, who always controls user data, regardless of application scope?

1 pts

volume_mute

The application provider

The cloud service provider

The data annotation team

The customer or end user

note_alt Add notes

flag Flag

volume_mute