Generative AI and Amazon Bedrock (AIF-C01)

Question 1

A company needs to implement a generative AI solution to create marketing content. They want to test models from Anthropic, AI21 Labs, and Amazon Titan without managing servers or infrastructure. Which AWS service is most suitable?

Accepted Answer

Amazon Bedrock

Answer

Amazon SageMaker JumpStart

Answer

AWS Lambda

Answer

Amazon Q

Question 2

An LLM-based customer service chatbot invents answers about the company's return policies. Which technique is the most cost-effective and efficient to fix this using updated official documentation?

Accepted Answer

Retrieval-Augmented Generation (RAG)

Answer

Model Re-training (Pre-training)

Answer

Fine-tuning

Answer

Increasing model temperature

Question 3

You are using Amazon Bedrock to generate creative names for a startup. Currently, the model returns very common and repetitive names. Which change in the inference parameter would increase variety and creativity?

Accepted Answer

Increase Temperature

Answer

Decrease Temperature

Answer

Set Top-K to 1

Answer

Decrease Max Tokens

Question 4

A company wants to ensure that no Personally Identifiable Information (PII) of its customers is accidentally sent to the Bedrock model during chat interactions. What feature should they configure?

Accepted Answer

PII Filters in Guardrails for Amazon Bedrock

Answer

Knowledge Bases for Amazon Bedrock

Answer

Model Evaluation

Answer

Provisioned Throughput

Question 5

Which 'Prompt Engineering' technique improves the model's mathematical reasoning by explicitly asking it: 'Think step-by-step'?

Accepted Answer

Chain-of-Thought (CoT) Prompting

Answer

Zero-shot Prompting

Answer

Few-shot Prompting

Answer

Prompt Injection

Question 6

Which AWS service natively integrates with 'Knowledge Bases for Amazon Bedrock' to act as the vector database storing embeddings?

Accepted Answer

Amazon OpenSearch Serverless

Answer

Amazon S3

Answer

Amazon DynamoDB Standard

Answer

Amazon Redshift

Question 7

A company wants to adapt a Bedrock model to learn a new language or specific technical jargon, using a large corpus of unlabeled data. Which customization method is appropriate?

Accepted Answer

Continued Pre-training

Answer

Fine-tuning

Answer

Prompt Engineering

Answer

RAG

Question 8

Which Amazon Bedrock component allows orchestrating complex tasks, like booking appointments or sending emails, by invoking APIs and reasoning about the necessary steps?

Accepted Answer

Agents for Amazon Bedrock

Answer

Knowledge Bases

Answer

Guardrails

Answer

Playgrounds

Question 9

In the Amazon Bedrock pricing model, how is the usage of text models (LLMs) typically charged on-demand?

Accepted Answer

Per number of input tokens processed and output tokens generated.

Answer

Per instance hour deployed.

Answer

Per API request, regardless of length.

Answer

Flat monthly fee.

Question 10

Which AWS model is specifically optimized for generating software code from natural language comments?

Accepted Answer

Amazon Q Developer (based on code FMs)

Answer

Amazon Titan Image Generator

Answer

Amazon Polly

Answer

Amazon Comprehend