AmazonBedrockAndGenAI

Question 1

Generative AI is a subset of ____.

Answer 1

Deep Learning

Question 2

____ is used to generate new data that is similar to the data it was trained on.

Answer 2

Generative AI

Question 3

In order for AI to generate data, we must rely on a ____ model.

Answer 3

foundation

Question 4

Foundation models are ____ on a wide variety of input data which costs tens of millions of dollars.

Answer 4

trained

Question 5

____ models take in larges around of input data in order to generate new content.

Answer 5

Foundation

Question 6

____ models are a type of AI designed to generate coherent human-like text.

Answer 6

Large Language

Question 7

Large Language models are trained on large amounts of ____.

Answer 7

text data (books, articles, websites, other textual data)

Question 8

You normally interact with LLMs by giving a ____.

Answer 8

prompt

Question 9

The generated text by a LLM may be different for every user that uses the same prompt. What is this kind of output called?

Answer 9

Non-deterministic

Question 10

____ is the AWS fully managed service used to build Generative AI applications.

Answer 10

Amazon Bedrock

Question 11

True/False: Amazon Bedrock ony makes available foundation models developed by Amazon.

Answer 11

False: models are available from a number of other companys.

Question 12

Amazon Bedrock makes a ____ of the FM, available only to you, which you can futher ____ with your own data.

Answer 12

copy / fine-tune

Question 13

True/False: None of your data leaves your account or is used to train the FM.

Answer 13

t

Question 14

_____ models support a variety of input and outputs.

Answer 14

Multimodal

Question 15

Amazon ____ are high-performing foundation models from AWS that support image, text, multimodal model choices via a fully managed API.

Answer 15

Titan

Question 16

____ is the converting of raw text into a sequence of tokens.

Answer 16

Tokenizaton

Question 17

____ tokenization is where text is split into individual words.

Answer 17

word-based

Question 18

____ tokenization is where some words can be split as well.

Answer 18

subword

Question 19

A ____ is the number of tokens a LLM can consider when generating text.

Answer 19

context window

Question 20

The larger the context window a GenAI model has, the more ____ and ____ it has.

Answer 20

information / conherence

Question 21

Large GenAI context windows require more ____.

Answer 21

memory and processing

Question 22

When you are considering which GenAI model to use, the ____ is the first factor to consider.

Answer 22

context window

Question 23

Vectors have a high ____ to capture many features for one input token, such as semantic meaning, synactic role and sentiment.

Answer 23

dimensionality

Question 24

You can use ____ models to power search applications they are easily searchable due to nearst-neighbor (kNN)

Answer 24

embedding

Question 25

Words that have a semantic relationship have similar ____ in avector database.

Answer 25

embeddings

Question 26

You can use ____ of word embeddings to 2D to help visually see relationships with more nuance.

Answer 26

dimensionality reduction

Question 27

Amazon Bedrock provides ____ to control the interaction between users and Foundation Models (FMs)

Answer 27

guardrails

Question 28

Guardrails can be used to filter ____ and ____ content.

Answer 28

undesirable and harmful

Question 29

True/False: Guardrails in Amazon Bedrock can remove personally identifiable information (PII).

Answer 29

t

Question 30

True/False: Guardrails in Amazon Bedrock cannot reduce hallucinations in your models.

Answer 30

False: it can reduce

Question 31

When creating a model guardrail, what is the capability called that allows you to validate if the model responses are grounded and factually correct based on the information provided in the reference source, and block responses that are below the defined threshold?

Answer 31

Grounding

Question 32

When creating a model guardrail, what is the capability called that allows you to validate if the model responses are relevant to the user’s query and block responses that are below the defined threshold?

Answer 32

Relevance

Question 33

With ____, the model can manage and carry out various multi-step tasks.

Answer 33

agents

Question 34

Agents can perform task ____, by ensuring they are done in the correct order and that information is passed correctly between tasks.

Answer 34

coordination

Question 35

Agents are configured to perform specific ____.

Answer 35

pre-defined action groups

Question 36

Agents can leverage ____ to retrieve information when necessary.

Answer 36

RAG - Retrieval Augmented Generation

Question 37

____ logging allows you to send your model logs to Cloudwatch or S3.

Answer 37

Model Invocation

Question 38

True/False: Model Invocation Logging can only send text logs to CloudWatch/S3.

Answer 38

False: text, images and embeddings can be included

Question 39

You can further analyze your Bedrock model logs with ____.

Answer 39

CloudWatch Logs Insights

Question 40

Amazon Bedrock can publish ____ to CloudWatch, including ContentFilteredCount that allows you to see if ____ are functioning.

Answer 40

metrics / Guardrails

Question 41

You can build CloudWatch ____ on top of CloudWatch metrics.

Answer 41

alarms